chore: Clean up repo

feat(llm, gui): More Inference Providers

README.md

@@ -17,7 +17,7 @@
 The <b>world state lives outside the model</b>, characters act through <b>intents that get validated</b> and applied by engine code. Each character's knowledge, memory, and emotional state are subjective and partial by construction.
 
 <p align="center">
-  <img src="./web/docs/src/assets/img/puppet.webp" />
+  <img src="./web/docs/src/assets/img/puppet.webp" alt="This pixel-art style image, set against a black background with the title 'The Puppet Master Paradox' at the top, depicts two identical, stylized puppets with white, round heads and orange patterned bodies facing each other. Between them hovers a small, glowing, four-pointed star, while each puppet has an orange-outlined speech bubble above it: the left one states, 'Good evening. Welcome to my humble bakery!', and the right one replies, 'Nice to meet you Assassin Bob. Wait... wha-', concluding with a small white 'X' icon in the bottom right corner." />
 </p>
 
 Single-agent or single-context systems (AI Dungeon and its descendants) prompt one model to _be_ the world and everyone in it. That breaks in predictable ways over long sessions:
@@ -41,7 +41,7 @@ Omnia answers every one of these failures with the same move: **pull the thing t
 ## What this buys you
 
 <p align="center">
-  <img src="./web/docs/src/assets/img/features.webp" />
+  <img src="./web/docs/src/assets/img/features.webp" alt="This pixel-art image features a grid of six distinct panels, each with an orange-outlined, jagged border, illustrating different concepts: 'Emergent Deceit' shows a person hiding a sword behind their back while offering a rose to another person; 'Divergent Perceptions' depicts two figures sitting at a table with speech bubbles labeled 'A' and 'B'; 'Player Agnostic' shows three figures engaged in different activities—watering a plant, standing still, and juggling—under the plumbob symbol from the sims; 'State Validated Agency' displays a stylized symbol of a person partially inside a portal crossed out by a large red 'X'; 'Deterministic Time' shows four circular panels representing a day-night cycle connected by arrows and clock icons; and 'Bring Your Own Model' features a large, central omnia icon surrounded by various LLM model logos including chatgpt, mistral, deepseek, claude, etc****" />
 </p>
 
 The payoff is scenario complexity that **uni-agent systems structurally cannot represent, no matter how good the model gets**.

apps/gui/next-env.d.ts

@@ -1,6 +1,6 @@
 /// <reference types="next" />
 /// <reference types="next/image-types/global" />
-import "./.next/dev/types/routes.d.ts";
+import "./.next/types/routes.d.ts";
 
 // NOTE: This file should not be edited
 // see https://nextjs.org/docs/app/api-reference/config/typescript for more information.

apps/gui/src/app/actions.ts

@@ -7,8 +7,10 @@ import type { SimSnapshot } from "@/lib/simulation";
 import {
   ProviderManager,
   ModelProviderInstance,
-  AVAILABLE_PROVIDERS,
+  getAvailableProviders as listAvailableProviders,
   ModelProviderMeta,
+  ModelLister,
+  ModelInfo,
 } from "@omnia/llm";
 
 function resolveScenarioPath(relative: string): string {
@@ -254,6 +256,7 @@ export async function createProviderInstance(
   modelName?: string,
   type: "generative" | "embedding" = "generative",
   maxContext?: number,
+  endpointUrl?: string,
 ): Promise<ModelProviderInstance> {
   return ProviderManager.create(
     name,
@@ -262,6 +265,7 @@ export async function createProviderInstance(
     modelName,
     type,
     maxContext,
+    endpointUrl,
   );
 }
 
@@ -281,6 +285,7 @@ export async function updateProviderInstance(
   modelName?: string,
   type: "generative" | "embedding" = "generative",
   maxContext?: number,
+  endpointUrl?: string,
 ): Promise<void> {
   ProviderManager.update(
     id,
@@ -290,6 +295,7 @@ export async function updateProviderInstance(
     modelName,
     type,
     maxContext,
+    endpointUrl,
   );
 }
 
@@ -305,7 +311,7 @@ export async function setProviderMapping(
 }
 
 export async function getAvailableProviders(): Promise<ModelProviderMeta[]> {
-  return AVAILABLE_PROVIDERS;
+  return listAvailableProviders();
 }
 
 export async function regenerateEmbeddings(
@@ -313,3 +319,32 @@ export async function regenerateEmbeddings(
 ): Promise<void> {
   await simulationManager.regenerateAllEmbeddings(newProviderInstanceId);
 }
+
+/**
+ * Fetch available models for a provider given its credentials.
+ * Used when creating a new instance (before it's saved to the DB).
+ */
+export async function fetchAvailableModels(
+  providerName: string,
+  apiKey: string,
+  endpointUrl?: string,
+): Promise<ModelInfo[]> {
+  return ModelLister.listModels(providerName, apiKey, endpointUrl);
+}
+
+/**
+ * Fetch available models for an existing saved provider instance.
+ * The API key is retrieved from the DB server-side — never sent to the client.
+ */
+export async function fetchAvailableModelsForInstance(
+  instanceId: string,
+): Promise<ModelInfo[]> {
+  const instances = ProviderManager.list();
+  const inst = instances.find((i) => i.id === instanceId);
+  if (!inst) return [];
+  return ModelLister.listModels(
+    inst.providerName,
+    inst.apiKey,
+    inst.endpointUrl,
+  );
+}

apps/gui/src/app/builder/page.tsx

@@ -1,18 +1,5 @@
-"use client";
+import { BuilderView } from "@/components/builder/BuilderView";
 
 export default function BuilderPage() {
-  return (
-    <div className="flex-1 overflow-y-auto w-full">
-      <div className="mx-auto max-w-[800px] px-10 py-12">
-        <h1 className="mb-6 text-headline-lg text-primary animate-fade-in">
-          Scenario Builder
-        </h1>
-        <div className="border border-border/30 bg-card p-6 shadow-[2px_2px_0_0_var(--border)] min-h-[300px] flex flex-col items-center justify-center">
-          <p className="text-body-md text-muted-foreground font-mono text-center">
-            Scenario builder interface coming soon...
-          </p>
-        </div>
-      </div>
-    </div>
-  );
+  return <BuilderView />;
 }

apps/gui/src/app/page.tsx

@@ -1,5 +1,5 @@
-import { DashboardView } from "@/components/play/DashboardView";
+import { HomeView } from "@/components/home/HomeView";
 
 export default function Home() {
-  return <DashboardView />;
+  return <HomeView />;
 }

apps/gui/src/components/builder/BuilderView.tsx

@@ -0,0 +1,18 @@
+"use client";
+
+export function BuilderView() {
+  return (
+    <div className="flex-1 overflow-y-auto w-full">
+      <div className="mx-auto max-w-[800px] px-10 py-12">
+        <h1 className="mb-6 text-headline-lg text-primary animate-fade-in">
+          Scenario Builder
+        </h1>
+        <div className="border border-border/30 bg-card p-6 shadow-[2px_2px_0_0_var(--border)] min-h-[300px] flex flex-col items-center justify-center">
+          <p className="text-body-md text-muted-foreground font-mono text-center">
+            Scenario builder interface coming soon...
+          </p>
+        </div>
+      </div>
+    </div>
+  );
+}

apps/gui/src/components/config/ConfigView.tsx

@@ -103,7 +103,7 @@ export function ConfigView() {
 
   return (
     <div className="flex-1 overflow-y-auto w-full relative">
-      <div className="relative z-10 mx-auto max-w-[800px] px-10 py-12">
+      <div className="relative z-10 mx-auto max-w-[1024px] px-10 py-12">
         <h1 className="mb-6 text-headline-lg text-primary animate-fade-in">
           Configuration
         </h1>

apps/gui/src/components/config/ProviderInstancesConfig.tsx

@@ -1,14 +1,20 @@
 "use client";
 
-import { useEffect, useState } from "react";
+import { useEffect, useRef, useState } from "react";
 import {
   createProviderInstance,
   updateProviderInstance,
   setActiveProviderInstance,
   regenerateEmbeddings,
   deleteProviderInstance,
+  fetchAvailableModels,
+  fetchAvailableModelsForInstance,
 } from "@/app/actions";
-import type { ModelProviderInstance, ModelProviderMeta } from "@omnia/llm";
+import type {
+  ModelInfo,
+  ModelProviderInstance,
+  ModelProviderMeta,
+} from "@omnia/llm";
 import { Button } from "@/components/ui/button";
 import { Input } from "@/components/ui/input";
 import { Label } from "@/components/ui/label";
@@ -22,6 +28,14 @@ import {
   SelectTrigger,
   SelectValue,
 } from "@/components/ui/select";
+import {
+  Combobox,
+  ComboboxContent,
+  ComboboxEmpty,
+  ComboboxInput,
+  ComboboxItem,
+  ComboboxList,
+} from "@/components/ui/combobox";
 import {
   Card,
   CardHeader,
@@ -38,6 +52,17 @@ import {
 } from "@/components/ui/item";
 import { Empty, EmptyTitle, EmptyDescription } from "@/components/ui/empty";
 import { cn } from "@/lib/utils";
+import { RefreshCwIcon } from "lucide-react";
+
+const providerLogoMap: Record<string, string> = {
+  anthropic: "/claude_logo.webp",
+  "google-genai": "/gemini_logo.webp",
+  openrouter: "/openrouter_logo.webp",
+  ollama: "/ollama_logo.webp",
+  deepseek: "/deepseek_logo.webp",
+  openai: "/openai_logo.webp",
+  groq: "/groq_logo.webp",
+};
 
 interface ProviderInstancesConfigProps {
   instances: ModelProviderInstance[];
@@ -64,9 +89,65 @@ export function ProviderInstancesConfig({
     "generative",
   );
   const [editMaxContext, setEditMaxContext] = useState<number>(32768);
+  const [editEndpointUrl, setEditEndpointUrl] = useState("");
   const [loading, setLoading] = useState(false);
   const [error, setError] = useState("");
 
+  // Model listing state
+  const [availableModels, setAvailableModels] = useState<ModelInfo[]>([]);
+  const [modelsLoading, setModelsLoading] = useState(false);
+  const debounceTimer = useRef<ReturnType<typeof setTimeout> | null>(null);
+
+  // Fetch models for an existing saved instance
+  const fetchModelsForExistingInstance = async (instanceId: string) => {
+    setModelsLoading(true);
+    setAvailableModels([]);
+    try {
+      const models = await fetchAvailableModelsForInstance(instanceId);
+      setAvailableModels(models);
+    } catch {
+      setAvailableModels([]);
+    } finally {
+      setModelsLoading(false);
+    }
+  };
+
+  // Fetch models for a new instance (using live key/endpoint from form)
+  const fetchModelsForNewInstance = (
+    provider: string,
+    apiKey: string,
+    endpointUrl: string,
+  ) => {
+    if (debounceTimer.current) clearTimeout(debounceTimer.current);
+
+    const isOllama = provider === "ollama";
+    const hasCredentials = isOllama
+      ? endpointUrl.trim().length > 0
+      : apiKey.trim().length > 8; // don't spam API with partial keys
+
+    if (!hasCredentials) {
+      setAvailableModels([]);
+      return;
+    }
+
+    debounceTimer.current = setTimeout(async () => {
+      setModelsLoading(true);
+      setAvailableModels([]);
+      try {
+        const models = await fetchAvailableModels(
+          provider,
+          isOllama ? "none" : apiKey,
+          isOllama ? endpointUrl : undefined,
+        );
+        setAvailableModels(models);
+      } catch {
+        setAvailableModels([]);
+      } finally {
+        setModelsLoading(false);
+      }
+    }, 600);
+  };
+
   useEffect(() => {
     if (selectedInstanceId === null) {
       setEditName("");
@@ -76,6 +157,8 @@ export function ProviderInstancesConfig({
       setEditIsActive(false);
       setEditType("generative");
       setEditMaxContext(32768);
+      setEditEndpointUrl("");
+      setAvailableModels([]);
     } else if (selectedInstanceId === "new") {
       setEditName("");
       const defaultProvider = "google-genai";
@@ -86,6 +169,8 @@ export function ProviderInstancesConfig({
       setEditModel(pMeta?.defaultModel || "gemini-2.5-flash");
       setEditIsActive(false);
       setEditMaxContext(32768);
+      setEditEndpointUrl("");
+      setAvailableModels([]);
     } else {
       const inst = instances.find((i) => i.id === selectedInstanceId);
       if (inst) {
@@ -109,10 +194,22 @@ export function ProviderInstancesConfig({
             ? inst.maxContext
             : 32768,
         );
+        setEditEndpointUrl(inst.endpointUrl || "");
+        setAvailableModels([]);
+        // Auto-fetch models for existing instances
+        fetchModelsForExistingInstance(selectedInstanceId);
       }
     }
+    // eslint-disable-next-line react-hooks/exhaustive-deps
   }, [selectedInstanceId, instances, availableProviders]);
 
+  // Re-fetch models when provider/key/endpoint changes on new instance form
+  useEffect(() => {
+    if (selectedInstanceId !== "new") return;
+    fetchModelsForNewInstance(editProvider, editKey, editEndpointUrl);
+    // eslint-disable-next-line react-hooks/exhaustive-deps
+  }, [editProvider, editKey, editEndpointUrl, selectedInstanceId]);
+
   const handleProviderChange = (providerId: string | null) => {
     if (!providerId) return;
     setEditProvider(providerId);
@@ -122,6 +219,7 @@ export function ProviderInstancesConfig({
         ? pMeta?.defaultEmbeddingModel || ""
         : pMeta?.defaultModel || "",
     );
+    setAvailableModels([]);
   };
 
   const handleTypeChange = (type: "generative" | "embedding") => {
@@ -134,6 +232,14 @@ export function ProviderInstancesConfig({
     );
   };
 
+  const handleRefreshModels = () => {
+    if (selectedInstanceId && selectedInstanceId !== "new") {
+      fetchModelsForExistingInstance(selectedInstanceId);
+    } else {
+      fetchModelsForNewInstance(editProvider, editKey, editEndpointUrl);
+    }
+  };
+
   const handleSave = async (e: React.FormEvent) => {
     e.preventDefault();
     if (!editName.trim()) {
@@ -149,7 +255,7 @@ export function ProviderInstancesConfig({
       let targetInstanceId = selectedInstanceId;
 
       if (selectedInstanceId === "new") {
-        if (!editKey.trim()) {
+        if (editProvider !== "ollama" && !editKey.trim()) {
           setError("API Key is required for new instances.");
           setLoading(false);
           return;
@@ -157,10 +263,13 @@ export function ProviderInstancesConfig({
         const created = await createProviderInstance(
           editName,
           editProvider,
-          editKey,
+          editProvider === "ollama" ? "none" : editKey,
           editModel || undefined,
           editType,
           editType === "generative" ? editMaxContext : 0,
+          editProvider === "ollama"
+            ? editEndpointUrl || "http://localhost:11434"
+            : undefined,
         );
         if (editIsActive) {
           await setActiveProviderInstance(created.id);
@@ -194,10 +303,13 @@ export function ProviderInstancesConfig({
           selectedInstanceId,
           editName,
           editProvider,
-          editKey || undefined,
+          editProvider === "ollama" ? "none" : editKey || undefined,
           editModel || undefined,
           editType,
           editType === "generative" ? editMaxContext : 0,
+          editProvider === "ollama"
+            ? editEndpointUrl || "http://localhost:11434"
+            : undefined,
         );
         if (editIsActive) {
           await setActiveProviderInstance(selectedInstanceId);
@@ -235,7 +347,7 @@ export function ProviderInstancesConfig({
   };
 
   return (
-    <section className="mb-8">
+    <section className="mb-8 flex min-h-[600px] flex-col">
       {error && (
         <div className="mb-4 rounded border-2 border-red-500 bg-red-50 px-3 py-2 text-sm text-red-700">
           {error}
@@ -243,7 +355,7 @@ export function ProviderInstancesConfig({
       )}
       <div
         className={cn(
-          "mt-4 grid min-h-[400px] grid-cols-1 gap-4 md:grid-cols-[30%_70%]",
+          "mt-4 grid flex-1 grid-cols-1 gap-4 md:grid-cols-[30%_70%]",
           loading && "pointer-events-none opacity-60",
           "transition-opacity duration-200",
         )}
@@ -277,14 +389,22 @@ export function ProviderInstancesConfig({
                     )}
                     onClick={() => setSelectedInstanceId(inst.id)}
                   >
-                    <ItemContent>
-                      <ItemTitle>{inst.name}</ItemTitle>
-                      <ItemDescription>{inst.providerName}</ItemDescription>
-                      <div className="flex flex-row gap-1.5">
-                        {inst.isActive && <Badge>Active</Badge>}
-                        <Badge variant="outline">
-                          {inst.type === "generative" ? "gen" : "embed"}
-                        </Badge>
+                    <ItemContent className="flex-row items-center gap-3">
+                      {providerLogoMap[inst.providerName] && (
+                        <img
+                          src={providerLogoMap[inst.providerName]}
+                          alt={inst.providerName}
+                          className="max-h-[38px] max-w-[38px] shrink-0 rounded-sm object-contain"
+                        />
+                      )}
+                      <div className="flex flex-col gap-1">
+                        <ItemTitle>{inst.name}</ItemTitle>
+                        <div className="flex flex-row gap-1.5">
+                          {inst.isActive && <Badge>Active</Badge>}
+                          <Badge variant="outline">
+                            {inst.type === "generative" ? "gen" : "embed"}
+                          </Badge>
+                        </div>
                       </div>
                     </ItemContent>
                   </Item>
@@ -329,6 +449,7 @@ export function ProviderInstancesConfig({
                 <div className="grid grid-cols-[2fr_3fr] gap-4">
                   <div className="flex flex-col gap-1.5">
                     <Label>Instance Type</Label>
+                    {/* TODO: Change this to choice card */}
                     <Select
                       value={editType}
                       onValueChange={(v) =>
@@ -336,11 +457,11 @@ export function ProviderInstancesConfig({
                       }
                       items={[
                         {
-                          label: "Generative (Text Completion)",
+                          label: "Generative (Text Generation)",
                           value: "generative",
                         },
                         {
-                          label: "Embedding (Vector generation)",
+                          label: "Embedding (Vector Embeddings)",
                           value: "embedding",
                         },
                       ]}
@@ -351,10 +472,10 @@ export function ProviderInstancesConfig({
                       <SelectContent>
                         <SelectGroup>
                           <SelectItem value="generative">
-                            Generative (Chat / Text Completion)
+                            Generative (Text Generation)
                           </SelectItem>
                           <SelectItem value="embedding">
-                            Embedding (Vector generation)
+                            Embedding (Vector Embeddings)
                           </SelectItem>
                         </SelectGroup>
                       </SelectContent>
@@ -398,30 +519,109 @@ export function ProviderInstancesConfig({
                   </span>
                 )}
 
-                <div className="flex flex-col gap-1.5">
-                  <Label htmlFor="formKey">API Key</Label>
-                  <Input
-                    id="formKey"
-                    type="password"
-                    value={editKey}
-                    onChange={(e) => setEditKey(e.target.value)}
-                    placeholder={
-                      selectedInstanceId === "new"
-                        ? "AIzaSy..."
-                        : "•••••••• (unchanged)"
-                    }
-                    required={selectedInstanceId === "new"}
-                  />
-                </div>
+                {editProvider !== "ollama" && (
+                  <div className="flex flex-col gap-1.5">
+                    <Label htmlFor="formKey">API Key</Label>
+                    <Input
+                      id="formKey"
+                      type="password"
+                      value={editKey}
+                      onChange={(e) => setEditKey(e.target.value)}
+                      placeholder={
+                        selectedInstanceId === "new"
+                          ? "AIzaSy..."
+                          : "•••••••• (unchanged)"
+                      }
+                      required={selectedInstanceId === "new"}
+                    />
+                  </div>
+                )}
+
+                {editProvider === "ollama" && (
+                  <div className="flex flex-col gap-1.5">
+                    <Label htmlFor="formEndpoint">Endpoint URL</Label>
+                    <Input
+                      id="formEndpoint"
+                      value={editEndpointUrl}
+                      onChange={(e) => setEditEndpointUrl(e.target.value)}
+                      placeholder="e.g. http://localhost:11434"
+                      required
+                    />
+                  </div>
+                )}
 
                 <div className="flex flex-col gap-1.5">
-                  <Label htmlFor="formModel">Model Name</Label>
-                  <Input
-                    id="formModel"
+                  <div className="flex items-center justify-between">
+                    <Label htmlFor="formModel">Model</Label>
+                    <button
+                      type="button"
+                      onClick={handleRefreshModels}
+                      disabled={modelsLoading}
+                      className="flex items-center gap-1 rounded px-1.5 py-0.5 text-xs text-muted-foreground transition-colors hover:bg-muted hover:text-foreground disabled:opacity-40"
+                      title="Refresh model list"
+                    >
+                      <RefreshCwIcon
+                        className={cn(
+                          "size-3",
+                          modelsLoading && "animate-spin",
+                        )}
+                      />
+                      {modelsLoading
+                        ? "Fetching…"
+                        : availableModels.length > 0
+                          ? `${availableModels.length} models`
+                          : "Fetch models"}
+                    </button>
+                  </div>
+                  <Combobox
                     value={editModel}
-                    onChange={(e) => setEditModel(e.target.value)}
-                    placeholder="e.g. gemini-2.5-flash, gemini-2.5-pro"
-                  />
+                    onValueChange={(v) => {
+                      if (v) setEditModel(v);
+                    }}
+                    items={availableModels.map((m) => m.id)}
+                  >
+                    <ComboboxInput
+                      id="formModel"
+                      placeholder={
+                        modelsLoading
+                          ? "Fetching models…"
+                          : availableModels.length > 0
+                            ? "Search or select a model…"
+                            : "e.g. gemini-2.5-flash"
+                      }
+                      disabled={modelsLoading}
+                      showClear={false}
+                      value={editModel}
+                      onChange={(e) =>
+                        setEditModel((e.target as HTMLInputElement).value)
+                      }
+                      className="w-full"
+                    />
+                    <ComboboxContent>
+                      <ComboboxEmpty>
+                        {modelsLoading
+                          ? "Fetching models…"
+                          : "No models found. Type a custom model name above."}
+                      </ComboboxEmpty>
+                      <ComboboxList>
+                        {(modelId: string) => {
+                          const model = availableModels.find(
+                            (m) => m.id === modelId,
+                          );
+                          return (
+                            <ComboboxItem key={modelId} value={modelId}>
+                              <span className="flex-1 truncate">{modelId}</span>
+                              {model?.ownedBy && (
+                                <span className="ml-2 shrink-0 text-xs text-muted-foreground">
+                                  {model.ownedBy}
+                                </span>
+                              )}
+                            </ComboboxItem>
+                          );
+                        }}
+                      </ComboboxList>
+                    </ComboboxContent>
+                  </Combobox>
                 </div>
 
                 {editType === "generative" && (

apps/gui/src/components/home/HomeView.tsx

@@ -34,7 +34,7 @@ import {
   SelectValue,
 } from "@/components/ui/select";
 
-export function DashboardView() {
+export function HomeView() {
   const router = useRouter();
   const [loading, setLoading] = useState(false);
   const [loadingData, setLoadingData] = useState(true);
@@ -188,7 +188,7 @@ export function DashboardView() {
 
   return (
     <div className="flex-1 overflow-y-auto w-full relative">
-      <div className="relative z-10 mx-auto max-w-[800px] px-10 py-12">
+      <div className="relative z-10 mx-auto max-w-[1024px] px-10 py-12">
         <div className="animate-fade-in">
           {/* Centered Big Logo */}
           <div className="flex flex-col items-center justify-center mb-10 pt-4">

apps/gui/src/components/ui/combobox.tsx

@@ -0,0 +1,300 @@
+"use client";
+
+import * as React from "react";
+import { Combobox as ComboboxPrimitive } from "@base-ui/react";
+import { CheckIcon, ChevronDownIcon, XIcon } from "lucide-react";
+
+import { cn } from "@/lib/utils";
+import { Button } from "@/components/ui/button";
+import {
+  InputGroup,
+  InputGroupAddon,
+  InputGroupButton,
+  InputGroupInput,
+} from "@/components/ui/input-group";
+
+const Combobox = ComboboxPrimitive.Root;
+
+function ComboboxValue({ ...props }: ComboboxPrimitive.Value.Props) {
+  return <ComboboxPrimitive.Value data-slot="combobox-value" {...props} />;
+}
+
+function ComboboxTrigger({
+  className,
+  children,
+  ...props
+}: ComboboxPrimitive.Trigger.Props) {
+  return (
+    <ComboboxPrimitive.Trigger
+      data-slot="combobox-trigger"
+      className={cn("[&_svg:not([class*='size-'])]:size-4", className)}
+      {...props}
+    >
+      {children}
+      <ChevronDownIcon className="pointer-events-none size-4 text-muted-foreground" />
+    </ComboboxPrimitive.Trigger>
+  );
+}
+
+function ComboboxClear({ className, ...props }: ComboboxPrimitive.Clear.Props) {
+  return (
+    <ComboboxPrimitive.Clear
+      data-slot="combobox-clear"
+      render={<InputGroupButton variant="ghost" size="icon-xs" />}
+      className={cn(className)}
+      {...props}
+    >
+      <XIcon className="pointer-events-none" />
+    </ComboboxPrimitive.Clear>
+  );
+}
+
+function ComboboxInput({
+  className,
+  children,
+  disabled = false,
+  showTrigger = true,
+  showClear = false,
+  ...props
+}: ComboboxPrimitive.Input.Props & {
+  showTrigger?: boolean;
+  showClear?: boolean;
+}) {
+  return (
+    <InputGroup className={cn("w-auto", className)}>
+      <ComboboxPrimitive.Input
+        render={<InputGroupInput disabled={disabled} />}
+        {...props}
+      />
+      <InputGroupAddon align="inline-end">
+        {showTrigger && (
+          <InputGroupButton
+            size="icon-xs"
+            variant="ghost"
+            render={<ComboboxTrigger />}
+            data-slot="input-group-button"
+            className="group-has-data-[slot=combobox-clear]/input-group:hidden data-pressed:bg-transparent"
+            disabled={disabled}
+          />
+        )}
+        {showClear && <ComboboxClear disabled={disabled} />}
+      </InputGroupAddon>
+      {children}
+    </InputGroup>
+  );
+}
+
+function ComboboxContent({
+  className,
+  side = "bottom",
+  sideOffset = 6,
+  align = "start",
+  alignOffset = 0,
+  anchor,
+  ...props
+}: ComboboxPrimitive.Popup.Props &
+  Pick<
+    ComboboxPrimitive.Positioner.Props,
+    "side" | "align" | "sideOffset" | "alignOffset" | "anchor"
+  >) {
+  return (
+    <ComboboxPrimitive.Portal>
+      <ComboboxPrimitive.Positioner
+        side={side}
+        sideOffset={sideOffset}
+        align={align}
+        alignOffset={alignOffset}
+        anchor={anchor}
+        className="isolate z-50"
+      >
+        <ComboboxPrimitive.Popup
+          data-slot="combobox-content"
+          data-chips={!!anchor}
+          className={cn(
+            "cn-menu-target cn-menu-translucent group/combobox-content relative max-h-(--available-height) w-(--anchor-width) max-w-(--available-width) min-w-[calc(var(--anchor-width)+var(--spacing-7,1.75rem))] origin-(--transform-origin) overflow-hidden rounded border-2 bg-popover text-popover-foreground shadow-md duration-100 data-[chips=true]:min-w-(--anchor-width) data-[side=bottom]:slide-in-from-top-2 data-[side=inline-end]:slide-in-from-left-2 data-[side=inline-start]:slide-in-from-right-2 data-[side=left]:slide-in-from-right-2 data-[side=right]:slide-in-from-left-2 data-[side=top]:slide-in-from-bottom-2 *:data-[slot=input-group]:m-1 *:data-[slot=input-group]:mb-0 *:data-[slot=input-group]:h-8 *:data-[slot=input-group]:border-input/30 *:data-[slot=input-group]:bg-input/30 *:data-[slot=input-group]:shadow-none data-open:animate-in data-open:fade-in-0 data-open:zoom-in-95 data-closed:animate-out data-closed:fade-out-0 data-closed:zoom-out-95",
+            className,
+          )}
+          {...props}
+        />
+      </ComboboxPrimitive.Positioner>
+    </ComboboxPrimitive.Portal>
+  );
+}
+
+function ComboboxList({ className, ...props }: ComboboxPrimitive.List.Props) {
+  return (
+    <ComboboxPrimitive.List
+      data-slot="combobox-list"
+      className={cn(
+        "no-scrollbar max-h-72 scroll-py-1 overflow-y-auto overscroll-contain p-1 data-empty:p-0",
+        className,
+      )}
+      {...props}
+    />
+  );
+}
+
+function ComboboxItem({
+  className,
+  children,
+  ...props
+}: ComboboxPrimitive.Item.Props) {
+  return (
+    <ComboboxPrimitive.Item
+      data-slot="combobox-item"
+      className={cn(
+        "relative flex w-full cursor-default items-center gap-2 rounded-sm py-1 pr-8 pl-1.5 text-sm outline-hidden select-none data-highlighted:bg-accent data-highlighted:text-accent-foreground not-data-[variant=destructive]:data-highlighted:**:text-accent-foreground data-disabled:pointer-events-none data-disabled:opacity-50 [&_svg]:pointer-events-none [&_svg]:shrink-0 [&_svg:not([class*='size-'])]:size-4",
+        className,
+      )}
+      {...props}
+    >
+      {children}
+      <ComboboxPrimitive.ItemIndicator
+        render={
+          <span className="pointer-events-none absolute right-2 flex size-4 items-center justify-center" />
+        }
+      >
+        <CheckIcon className="pointer-events-none" />
+      </ComboboxPrimitive.ItemIndicator>
+    </ComboboxPrimitive.Item>
+  );
+}
+
+function ComboboxGroup({ className, ...props }: ComboboxPrimitive.Group.Props) {
+  return (
+    <ComboboxPrimitive.Group
+      data-slot="combobox-group"
+      className={cn(className)}
+      {...props}
+    />
+  );
+}
+
+function ComboboxLabel({
+  className,
+  ...props
+}: ComboboxPrimitive.GroupLabel.Props) {
+  return (
+    <ComboboxPrimitive.GroupLabel
+      data-slot="combobox-label"
+      className={cn("px-2 py-1.5 text-xs text-muted-foreground", className)}
+      {...props}
+    />
+  );
+}
+
+function ComboboxCollection({ ...props }: ComboboxPrimitive.Collection.Props) {
+  return (
+    <ComboboxPrimitive.Collection data-slot="combobox-collection" {...props} />
+  );
+}
+
+function ComboboxEmpty({ className, ...props }: ComboboxPrimitive.Empty.Props) {
+  return (
+    <ComboboxPrimitive.Empty
+      data-slot="combobox-empty"
+      className={cn(
+        "hidden w-full justify-center py-2 text-center text-sm text-muted-foreground group-data-empty/combobox-content:flex",
+        className,
+      )}
+      {...props}
+    />
+  );
+}
+
+function ComboboxSeparator({
+  className,
+  ...props
+}: ComboboxPrimitive.Separator.Props) {
+  return (
+    <ComboboxPrimitive.Separator
+      data-slot="combobox-separator"
+      className={cn("-mx-1 my-1 h-px bg-border", className)}
+      {...props}
+    />
+  );
+}
+
+function ComboboxChips({
+  className,
+  ...props
+}: React.ComponentPropsWithRef<typeof ComboboxPrimitive.Chips> &
+  ComboboxPrimitive.Chips.Props) {
+  return (
+    <ComboboxPrimitive.Chips
+      data-slot="combobox-chips"
+      className={cn(
+        "flex min-h-8 flex-wrap items-center gap-1 rounded border-2 bg-input bg-clip-padding px-2.5 py-1 text-sm shadow-sm transition-colors focus-within:outline-2 focus-within:outline-offset-2 focus-within:outline-primary has-aria-invalid:border-destructive has-data-[slot=combobox-chip]:px-1",
+        className,
+      )}
+      {...props}
+    />
+  );
+}
+
+function ComboboxChip({
+  className,
+  children,
+  showRemove = true,
+  ...props
+}: ComboboxPrimitive.Chip.Props & {
+  showRemove?: boolean;
+}) {
+  return (
+    <ComboboxPrimitive.Chip
+      data-slot="combobox-chip"
+      className={cn(
+        "flex h-[calc(var(--spacing,0.25rem)*5.25)] w-fit items-center justify-center gap-1 rounded-sm border-2 bg-muted px-1.5 text-xs font-medium whitespace-nowrap text-foreground has-disabled:pointer-events-none has-disabled:cursor-not-allowed has-disabled:opacity-50 has-data-[slot=combobox-chip-remove]:pr-0",
+        className,
+      )}
+      {...props}
+    >
+      {children}
+      {showRemove && (
+        <ComboboxPrimitive.ChipRemove
+          render={<InputGroupButton variant="ghost" size="icon-xs" />}
+          className="-ml-1 opacity-50 hover:opacity-100"
+          data-slot="combobox-chip-remove"
+        >
+          <XIcon className="pointer-events-none" />
+        </ComboboxPrimitive.ChipRemove>
+      )}
+    </ComboboxPrimitive.Chip>
+  );
+}
+
+function ComboboxChipsInput({
+  className,
+  ...props
+}: ComboboxPrimitive.Input.Props) {
+  return (
+    <ComboboxPrimitive.Input
+      data-slot="combobox-chip-input"
+      className={cn("min-w-16 flex-1 outline-none", className)}
+      {...props}
+    />
+  );
+}
+
+function useComboboxAnchor() {
+  return React.useRef<HTMLDivElement | null>(null);
+}
+
+export {
+  Combobox,
+  ComboboxInput,
+  ComboboxContent,
+  ComboboxList,
+  ComboboxItem,
+  ComboboxGroup,
+  ComboboxLabel,
+  ComboboxCollection,
+  ComboboxEmpty,
+  ComboboxSeparator,
+  ComboboxChips,
+  ComboboxChip,
+  ComboboxChipsInput,
+  ComboboxTrigger,
+  ComboboxValue,
+  useComboboxAnchor,
+};

apps/gui/src/components/ui/input-group.tsx

@@ -0,0 +1,172 @@
+"use client";
+
+import * as React from "react";
+import { cva, type VariantProps } from "class-variance-authority";
+
+import { cn } from "@/lib/utils";
+import { Button } from "@/components/ui/button";
+import { Input } from "@/components/ui/input";
+import { Textarea } from "@/components/ui/textarea";
+
+function InputGroup({ className, ...props }: React.ComponentProps<"div">) {
+  return (
+    <div
+      data-slot="input-group"
+      role="group"
+      className={cn(
+        "group/input-group relative flex h-8 w-full min-w-0 items-center rounded border-2 bg-input shadow-sm transition-colors outline-none in-data-[slot=combobox-content]:focus-within:border-inherit in-data-[slot=combobox-content]:focus-within:ring-0 has-disabled:bg-input/50 has-disabled:opacity-50 has-[[data-slot=input-group-control]:focus-visible]:outline-2 has-[[data-slot=input-group-control]:focus-visible]:outline-offset-2 has-[[data-slot=input-group-control]:focus-visible]:outline-primary has-[[data-slot][aria-invalid=true]]:border-destructive has-[>[data-align=block-end]]:h-auto has-[>[data-align=block-end]]:flex-col has-[>[data-align=block-start]]:h-auto has-[>[data-align=block-start]]:flex-col has-[>textarea]:h-auto has-[>[data-align=block-end]]:[&>input]:pt-3 has-[>[data-align=block-start]]:[&>input]:pb-3 has-[>[data-align=inline-end]]:[&>input]:pr-1.5 has-[>[data-align=inline-start]]:[&>input]:pl-1.5",
+        className,
+      )}
+      {...props}
+    />
+  );
+}
+
+const inputGroupAddonVariants = cva(
+  "flex h-auto cursor-text items-center justify-center gap-2 py-1.5 text-sm font-medium text-muted-foreground select-none group-data-[disabled=true]/input-group:opacity-50 [&>kbd]:rounded-[calc(var(--radius)-5px)] [&>svg:not([class*='size-'])]:size-4",
+  {
+    variants: {
+      align: {
+        "inline-start":
+          "order-first pl-2 has-[>button]:ml-[-0.3rem] has-[>kbd]:ml-[-0.15rem]",
+        "inline-end":
+          "order-last pr-2 has-[>button]:mr-[-0.3rem] has-[>kbd]:mr-[-0.15rem]",
+        "block-start":
+          "order-first w-full justify-start px-2.5 pt-2 group-has-[>input]/input-group:pt-2 [.border-b]:pb-2",
+        "block-end":
+          "order-last w-full justify-start px-2.5 pb-2 group-has-[>input]/input-group:pb-2 [.border-t]:pt-2",
+      },
+    },
+    defaultVariants: {
+      align: "inline-start",
+    },
+  },
+);
+
+function InputGroupAddon({
+  className,
+  align = "inline-start",
+  ...props
+}: React.ComponentProps<"div"> & VariantProps<typeof inputGroupAddonVariants>) {
+  return (
+    <div
+      role="group"
+      data-slot="input-group-addon"
+      data-align={align}
+      className={cn(inputGroupAddonVariants({ align }), className)}
+      onClick={(e) => {
+        if ((e.target as HTMLElement).closest("button")) {
+          return;
+        }
+        e.currentTarget.parentElement?.querySelector("input")?.focus();
+      }}
+      {...props}
+    />
+  );
+}
+
+const inputGroupButtonVariants = cva(
+  "flex items-center gap-2 text-sm shadow-none",
+  {
+    variants: {
+      size: {
+        xs: "h-6 gap-1 rounded-[calc(var(--radius)-3px)] px-1.5 [&>svg:not([class*='size-'])]:size-3.5",
+        sm: "",
+        "icon-xs":
+          "size-6 rounded-[calc(var(--radius)-3px)] p-0 has-[>svg]:p-0",
+        "icon-sm": "size-8 p-0 has-[>svg]:p-0",
+      },
+    },
+    defaultVariants: {
+      size: "xs",
+    },
+  },
+);
+
+function InputGroupButton({
+  className,
+  type = "button",
+  variant = "ghost",
+  size = "xs",
+  render,
+  ...props
+}: Omit<React.ComponentProps<typeof Button>, "size" | "type"> &
+  VariantProps<typeof inputGroupButtonVariants> & {
+    type?: "button" | "submit" | "reset";
+    render?: React.ReactElement;
+  }) {
+  if (render) {
+    return React.cloneElement(render, {
+      className: cn(
+        inputGroupButtonVariants({ size }),
+        (render.props as any)?.className,
+        className,
+      ),
+      type,
+      ...props,
+    } as any);
+  }
+
+  return (
+    <Button
+      type={type}
+      data-size={size}
+      variant={variant}
+      className={cn(inputGroupButtonVariants({ size }), className)}
+      {...props}
+    />
+  );
+}
+
+function InputGroupText({ className, ...props }: React.ComponentProps<"span">) {
+  return (
+    <span
+      className={cn(
+        "flex items-center gap-2 text-sm text-muted-foreground [&_svg]:pointer-events-none [&_svg:not([class*='size-'])]:size-4",
+        className,
+      )}
+      {...props}
+    />
+  );
+}
+
+function InputGroupInput({
+  className,
+  ...props
+}: React.ComponentProps<"input">) {
+  return (
+    <Input
+      data-slot="input-group-control"
+      className={cn(
+        "flex-1 rounded-none border-0 bg-transparent shadow-none outline-none focus-visible:outline-none disabled:bg-transparent aria-invalid:outline-none dark:bg-transparent dark:disabled:bg-transparent",
+        className,
+      )}
+      {...props}
+    />
+  );
+}
+
+function InputGroupTextarea({
+  className,
+  ...props
+}: React.ComponentProps<"textarea">) {
+  return (
+    <Textarea
+      data-slot="input-group-control"
+      className={cn(
+        "flex-1 resize-none rounded-none border-0 bg-transparent py-2 shadow-none outline-none focus-visible:outline-none disabled:bg-transparent aria-invalid:outline-none dark:bg-transparent dark:disabled:bg-transparent",
+        className,
+      )}
+      {...props}
+    />
+  );
+}
+
+export {
+  InputGroup,
+  InputGroupAddon,
+  InputGroupButton,
+  InputGroupText,
+  InputGroupInput,
+  InputGroupTextarea,
+};

apps/gui/src/lib/simulation/alias-handoff.ts

@@ -0,0 +1,71 @@
+import { HandoffEngine, checkHandoffTrigger } from "@omnia/memory";
+import type { SimSession } from "./types";
+
+/**
+ * Runs the HandoffEngine for every agent entity that has accumulated enough
+ * buffer entries to warrant a handoff (compression to long-term memory).
+ */
+export async function runHandoffResolution(session: SimSession): Promise<void> {
+  const worldState = session.coreRepo.loadWorldState(session.worldInstanceId);
+  if (!worldState) throw new Error("World state lost");
+
+  const handoffEngine = new HandoffEngine(
+    session.handoffProvider,
+    session.embeddingProvider,
+    session.bufferRepo,
+    session.ledgerRepo,
+  );
+
+  const entities = Array.from(worldState.entities.values());
+  for (const entity of entities) {
+    if (!entity.isAgent) continue;
+
+    const bufferEntries = session.bufferRepo.listForOwner(entity.id);
+    const maxContext =
+      session.handoffProvider.maxContext !== undefined
+        ? session.handoffProvider.maxContext
+        : 32768;
+
+    const trigger = checkHandoffTrigger(
+      entity,
+      bufferEntries,
+      worldState.clock.get(),
+      maxContext,
+    );
+    if (trigger !== "none") {
+      await handoffEngine.runHandoff(
+        entity,
+        bufferEntries,
+        worldState.clock.get(),
+      );
+    }
+  }
+}
+
+/**
+ * For every agent that shares a location with another entity they haven't
+ * previously encountered, generates a first-person alias description and
+ * persists it on the viewing entity.
+ */
+export async function runAliasResolution(session: SimSession): Promise<void> {
+  const worldState = session.coreRepo.loadWorldState(session.worldInstanceId);
+  if (!worldState) throw new Error("World state lost");
+
+  const entities = Array.from(worldState.entities.values());
+  for (const viewer of entities) {
+    if (!viewer.isAgent) continue;
+    if (!viewer.locationId) continue;
+
+    for (const target of entities) {
+      if (viewer.id === target.id) continue;
+      if (
+        target.locationId === viewer.locationId &&
+        !viewer.aliases.has(target.id)
+      ) {
+        const alias = await session.aliasGenerator.generate(viewer, target);
+        viewer.aliases.set(target.id, alias);
+        session.coreRepo.saveEntity(viewer, worldState.id);
+      }
+    }
+  }
+}

apps/gui/src/lib/simulation/env.ts

@@ -0,0 +1,16 @@
+import dotenv from "dotenv";
+import path from "path";
+import fs from "fs";
+
+// Load .env from monorepo root or apps/gui/
+const cwd = process.cwd();
+const envCandidates = [
+  path.resolve(cwd, ".env"),
+  path.resolve(cwd, "../../.env"),
+];
+for (const c of envCandidates) {
+  if (fs.existsSync(c) && fs.statSync(c).isFile()) {
+    dotenv.config({ path: c });
+    break;
+  }
+}

apps/gui/src/lib/simulation/index.ts

@@ -0,0 +1,17 @@
+/**
+ * Barrel entry point for the simulation module.
+ *
+ * Consumers import from "@/lib/simulation" exactly as before — no import
+ * paths need to change anywhere in the codebase.
+ */
+import { SimulationManager } from "./simulation-manager";
+
+export const simulationManager = new SimulationManager();
+
+export type {
+  SimSnapshot,
+  EntityInfo,
+  LogEntry,
+  IntentInfo,
+  WaitingContext,
+} from "../simulation-types";

apps/gui/src/lib/simulation/provider-resolver.ts

@@ -0,0 +1,146 @@
+import {
+  MockLLMProvider,
+  MockEmbeddingProvider,
+  ProviderManager,
+  buildLLMProvider,
+  buildEmbeddingProvider,
+} from "@omnia/llm";
+import type {
+  ILLMProvider,
+  IEmbeddingProvider,
+  ModelProviderInstance,
+} from "@omnia/llm";
+
+// ---------------------------------------------------------------------------
+// Public types
+// ---------------------------------------------------------------------------
+
+export interface ResolvedProviders {
+  actorProvider: ILLMProvider;
+  validatorProvider: ILLMProvider;
+  decoderProvider: ILLMProvider;
+  timedeltaProvider: ILLMProvider;
+  handoffProvider: ILLMProvider;
+  embeddingProvider: IEmbeddingProvider;
+}
+
+export interface ProviderResolverOptions {
+  /**
+   * Pre-resolved generative instance to fall back to when ProviderManager has
+   * no active generative provider (e.g. when the caller already validated a
+   * specific provider during session creation).
+   */
+  fallbackInstance?: ModelProviderInstance | null;
+  /**
+   * When true, throws an Error if no provider can be resolved for a task.
+   * When false (default), falls back silently to MockLLMProvider / MockEmbeddingProvider.
+   */
+  required?: boolean;
+}
+
+// ---------------------------------------------------------------------------
+// Resolution logic
+// ---------------------------------------------------------------------------
+
+/**
+// Public API
+// ---------------------------------------------------------------------------
+
+/**
+ * Resolves all six LLM + embedding providers needed for a simulation session.
+ *
+ * Resolution order for each generative task:
+ *  1. Task-specific mapping from ProviderManager (via `mappings[task]`)
+ *  2. ProviderManager active generative instance
+ *  3. `fallbackInstance` (if supplied)
+ *  4. GOOGLE_API_KEY env var → auto-creates a temporary GeminiProvider
+ *  5. Throws (if `required`) or returns MockLLMProvider
+ */
+export function resolveProviders(
+  mappings: Record<string, string>,
+  options: ProviderResolverOptions = {},
+): ResolvedProviders {
+  const { fallbackInstance = null, required = false } = options;
+  const list = ProviderManager.list();
+  const activeGenerative =
+    ProviderManager.getActive("generative") ?? fallbackInstance ?? null;
+
+  const resolveGenerative = (task: string): ILLMProvider => {
+    const mappedId = mappings[task];
+    let inst: ModelProviderInstance | null = mappedId
+      ? (list.find((p) => p.id === mappedId) ?? null)
+      : null;
+
+    if (!inst || inst.type !== "generative") {
+      inst = activeGenerative;
+    }
+
+    if (!inst) {
+      const envKey = process.env.GOOGLE_API_KEY;
+      if (envKey) {
+        inst = ProviderManager.create(
+          "Default (Env)",
+          "google-genai",
+          envKey,
+          undefined,
+          "generative",
+        );
+      }
+    }
+
+    if (!inst) {
+      if (required) {
+        throw new Error(
+          `No active LLM Provider Instance found for task "${task}". Please configure a key in Settings first.`,
+        );
+      }
+      return new MockLLMProvider([]);
+    }
+
+    return buildLLMProvider(inst);
+  };
+
+  const resolveEmbedding = (): IEmbeddingProvider => {
+    const mappedId = mappings["embeddings"];
+    let inst: ModelProviderInstance | null = mappedId
+      ? (list.find((p) => p.id === mappedId) ?? null)
+      : null;
+
+    if (!inst || inst.type !== "embedding") {
+      inst = ProviderManager.getActive("embedding");
+    }
+
+    if (!inst) {
+      const envKey = process.env.GOOGLE_API_KEY;
+      if (envKey) {
+        inst = ProviderManager.create(
+          "Default Embed (Env)",
+          "google-genai",
+          envKey,
+          "gemini-embedding-001",
+          "embedding",
+        );
+      }
+    }
+
+    if (!inst) {
+      if (required) {
+        throw new Error(
+          `No active Embedding Provider Instance found. Please configure an embedding key in Settings first.`,
+        );
+      }
+      return new MockEmbeddingProvider(undefined);
+    }
+
+    return buildEmbeddingProvider(inst);
+  };
+
+  return {
+    actorProvider: resolveGenerative("actor-prose"),
+    validatorProvider: resolveGenerative("llm-validator"),
+    decoderProvider: resolveGenerative("intent-decoder"),
+    timedeltaProvider: resolveGenerative("timedelta"),
+    handoffProvider: resolveGenerative("handoff"),
+    embeddingProvider: resolveEmbedding(),
+  };
+}

apps/gui/src/lib/simulation/session-store.ts

@@ -0,0 +1,139 @@
+import Database from "better-sqlite3";
+import path from "path";
+import fs from "fs";
+import type { SimSession, SavedState } from "./types";
+import type { SimSnapshot } from "../simulation-types";
+
+export const DATA_DIR = path.resolve(process.cwd(), "data");
+
+// ---------------------------------------------------------------------------
+// Low-level read/write helpers
+// ---------------------------------------------------------------------------
+
+export function loadSessionState(
+  db: Database.Database,
+  id: string,
+): SavedState | null {
+  try {
+    db.prepare(
+      `CREATE TABLE IF NOT EXISTS gui_meta (
+        id TEXT PRIMARY KEY,
+        state_json TEXT
+      )`,
+    ).run();
+    const row = db
+      .prepare(`SELECT state_json FROM gui_meta WHERE id = ?`)
+      .get(id) as { state_json: string } | undefined;
+    return row ? (JSON.parse(row.state_json) as SavedState) : null;
+  } catch {
+    return null;
+  }
+}
+
+export function saveSession(session: SimSession): void {
+  const state: SavedState = {
+    scenarioName: session.scenarioName,
+    scenarioDescription: session.scenarioDescription,
+    turn: session.turn,
+    maxTurns: session.maxTurns,
+    entities: session.entities,
+    playerEntityId: session.playerEntityId,
+    entityIndex: session.entityIndex,
+    status: session.status,
+    error: session.error,
+    waitingEntity: session.waitingEntity,
+    aliasDoneForTurn: session.aliasDoneForTurn,
+    log: session.log,
+    providerMappings: session.providerMappings,
+  };
+
+  session.db
+    .prepare(
+      `CREATE TABLE IF NOT EXISTS gui_meta (
+        id TEXT PRIMARY KEY,
+        state_json TEXT
+      )`,
+    )
+    .run();
+
+  session.db
+    .prepare(
+      `INSERT INTO gui_meta (id, state_json)
+       VALUES (?, ?)
+       ON CONFLICT(id) DO UPDATE SET state_json = excluded.state_json`,
+    )
+    .run(session.worldInstanceId, JSON.stringify(state));
+}
+
+// ---------------------------------------------------------------------------
+// Session file management
+// ---------------------------------------------------------------------------
+
+export function deleteSessionFile(id: string): void {
+  const dbPath = path.join(DATA_DIR, `${id}.db`);
+  if (fs.existsSync(dbPath)) {
+    try {
+      fs.unlinkSync(dbPath);
+    } catch (err) {
+      console.error(`Failed to delete session file ${dbPath}:`, err);
+    }
+  }
+}
+
+/**
+ * Lists all saved simulation snapshots by scanning the data directory.
+ * Active in-memory sessions are snapshotted via the provided callback;
+ * inactive ones are read directly from their `.db` files.
+ */
+export function listSavedSessions(
+  activeSessions: Map<string, SimSession>,
+  snapshotFn: (session: SimSession) => SimSnapshot,
+): SimSnapshot[] {
+  if (!fs.existsSync(DATA_DIR)) return [];
+
+  const snapshots: SimSnapshot[] = [];
+  const files = fs
+    .readdirSync(DATA_DIR)
+    .filter((f) => f.startsWith("sim-") && f.endsWith(".db"));
+
+  for (const file of files) {
+    const id = file.replace(".db", "");
+    const dbPath = path.join(DATA_DIR, file);
+
+    const active = activeSessions.get(id);
+    if (active) {
+      snapshots.push(snapshotFn(active));
+      continue;
+    }
+
+    try {
+      const db = new Database(dbPath);
+      const state = loadSessionState(db, id);
+      db.close();
+
+      if (state) {
+        snapshots.push({
+          id,
+          status: state.status,
+          turn: state.turn,
+          maxTurns: state.maxTurns,
+          scenarioName: state.scenarioName,
+          scenarioDescription: state.scenarioDescription,
+          entities: state.entities || [],
+          log: state.log || [],
+          entityIndex: state.entityIndex,
+          waitingEntity: state.waitingEntity,
+          error: state.error,
+        });
+      }
+    } catch {
+      /* skip corrupt / in-use db files */
+    }
+  }
+
+  return snapshots.sort((a, b) => {
+    const tsA = parseInt(a.id.replace("sim-", ""), 10) || 0;
+    const tsB = parseInt(b.id.replace("sim-", ""), 10) || 0;
+    return tsB - tsA;
+  });
+}

apps/gui/src/lib/simulation/simulation-manager.ts

@@ -0,0 +1,471 @@
+import "./env"; // Must be first — loads .env before any code reads process.env
+import Database from "better-sqlite3";
+import path from "path";
+import fs from "fs";
+import { SQLiteRepository } from "@omnia/core";
+import { BufferRepository, LedgerRepository } from "@omnia/memory";
+import { Architect, AliasDeltaGenerator } from "@omnia/architect";
+import { ProviderManager, buildEmbeddingProvider } from "@omnia/llm";
+import type { ModelProviderInstance, IEmbeddingProvider } from "@omnia/llm";
+import { ScenarioLoader } from "@omnia/scenario";
+import type { SimSnapshot } from "../simulation-types";
+import type { SimSession, EntityInfo } from "./types";
+import { resolveProviders } from "./provider-resolver";
+import {
+  DATA_DIR,
+  loadSessionState,
+  saveSession,
+  listSavedSessions,
+  deleteSessionFile,
+} from "./session-store";
+import {
+  preparePlayerTurn,
+  processNpcTurn,
+  executePlayerAction,
+} from "./turn-executor";
+import { runAliasResolution, runHandoffResolution } from "./alias-handoff";
+
+export class SimulationManager {
+  private sessions = new Map<string, SimSession>();
+
+  // ---------------------------------------------------------------------------
+  // Session lifecycle
+  // ---------------------------------------------------------------------------
+
+  async create(
+    scenarioPath: string,
+    playEntityName?: string,
+    providerInstanceId?: string,
+  ): Promise<SimSnapshot> {
+    // Resolve or validate the active generative provider upfront so we can
+    // return a clean error snapshot before touching the filesystem.
+    let activeInstance: ModelProviderInstance | null = providerInstanceId
+      ? ProviderManager.list().find((p) => p.id === providerInstanceId) || null
+      : ProviderManager.getActive("generative");
+
+    if (!activeInstance) {
+      const envKey = process.env.GOOGLE_API_KEY;
+      if (envKey) {
+        activeInstance = ProviderManager.create(
+          "Default (Env)",
+          "google-genai",
+          envKey,
+          undefined,
+          "generative",
+        );
+      }
+    }
+    if (!activeInstance) {
+      return {
+        id: "",
+        status: "error",
+        turn: 0,
+        maxTurns: 20,
+        scenarioName: "",
+        scenarioDescription: "",
+        entities: [],
+        log: [],
+        entityIndex: 0,
+        error:
+          "No active LLM Provider Instance found. Please configure a key in Settings first.",
+      };
+    }
+
+    const scenarioJson = JSON.parse(fs.readFileSync(scenarioPath, "utf-8"));
+    const id = `sim-${Date.now()}`;
+
+    fs.mkdirSync(DATA_DIR, { recursive: true });
+    const dbPath = path.join(DATA_DIR, `${id}.db`);
+    const db = new Database(dbPath);
+    const coreRepo = new SQLiteRepository(db);
+    const bufferRepo = new BufferRepository(db);
+    const ledgerRepo = new LedgerRepository(db);
+    const loader = new ScenarioLoader(coreRepo, bufferRepo);
+
+    const worldInstanceId = id;
+    await loader.initializeWorld(scenarioJson, worldInstanceId);
+
+    const worldState = coreRepo.loadWorldState(worldInstanceId);
+    if (!worldState) {
+      db.close();
+      return {
+        id: "",
+        status: "error",
+        turn: 0,
+        maxTurns: 20,
+        scenarioName: "",
+        scenarioDescription: "",
+        entities: [],
+        log: [],
+        entityIndex: 0,
+        error: "Failed to load world state after initialization.",
+      };
+    }
+
+    // Build entity list
+    const rawEntities = Array.from(worldState.entities.values());
+    const entityInfos: EntityInfo[] = rawEntities.map((e) => ({
+      id: e.id,
+      name: (e.attributes.get("name")?.getValue() as string) || e.id,
+      isPlayer: false,
+      isAgent: e.isAgent,
+    }));
+
+    // Resolve player entity (exact match → name match → fuzzy)
+    let playerEntityId: string | undefined;
+    if (playEntityName) {
+      let matched = worldState.getEntity(playEntityName);
+      if (!matched) {
+        for (const ent of rawEntities) {
+          const nameAttr = ent.attributes.get("name")?.getValue() as
+            string | undefined;
+          if (nameAttr?.toLowerCase() === playEntityName.toLowerCase()) {
+            matched = ent;
+            break;
+          }
+        }
+      }
+      if (!matched) {
+        for (const ent of rawEntities) {
+          const nameAttr = ent.attributes.get("name")?.getValue() as
+            string | undefined;
+          if (
+            nameAttr?.toLowerCase().includes(playEntityName.toLowerCase()) ||
+            ent.id.toLowerCase().includes(playEntityName.toLowerCase())
+          ) {
+            matched = ent;
+            break;
+          }
+        }
+      }
+      if (matched) {
+        playerEntityId = matched.id;
+        const info = entityInfos.find((e) => e.id === matched!.id);
+        if (info) info.isPlayer = true;
+      }
+    }
+
+    const mappings = ProviderManager.getMappings();
+    const {
+      actorProvider,
+      validatorProvider,
+      decoderProvider,
+      timedeltaProvider,
+      handoffProvider,
+      embeddingProvider,
+    } = resolveProviders(mappings, { fallbackInstance: activeInstance });
+
+    const architect = new Architect(
+      { validator: validatorProvider, timedelta: timedeltaProvider },
+      coreRepo,
+    );
+    const aliasGenerator = new AliasDeltaGenerator(actorProvider);
+
+    const session: SimSession = {
+      db,
+      dbPath,
+      coreRepo,
+      bufferRepo,
+      ledgerRepo,
+      worldInstanceId,
+      scenarioName: scenarioJson.name,
+      scenarioDescription: scenarioJson.description || "",
+      turn: 1,
+      maxTurns: 20,
+      entities: entityInfos,
+      playerEntityId,
+      entityIndex: 0,
+      actorProvider,
+      validatorProvider,
+      decoderProvider,
+      timedeltaProvider,
+      handoffProvider,
+      embeddingProvider,
+      architect,
+      aliasGenerator,
+      log: [],
+      status: "running",
+      aliasDoneForTurn: false,
+      providerMappings: mappings,
+    };
+
+    this.sessions.set(id, session);
+    return this.snapshot(session);
+  }
+
+  async load(id: string): Promise<SimSnapshot | null> {
+    const active = this.sessions.get(id);
+    if (active) return this.snapshot(active);
+
+    const dbPath = path.join(DATA_DIR, `${id}.db`);
+    if (!fs.existsSync(dbPath)) return null;
+
+    try {
+      const db = new Database(dbPath);
+      const state = loadSessionState(db, id);
+      if (!state) {
+        db.close();
+        return null;
+      }
+
+      const mappings = state.providerMappings || {};
+      const {
+        actorProvider,
+        validatorProvider,
+        decoderProvider,
+        timedeltaProvider,
+        handoffProvider,
+        embeddingProvider,
+      } = resolveProviders(mappings, { required: true });
+
+      const coreRepo = new SQLiteRepository(db);
+      const bufferRepo = new BufferRepository(db);
+      const ledgerRepo = new LedgerRepository(db);
+      const architect = new Architect(
+        { validator: validatorProvider, timedelta: timedeltaProvider },
+        coreRepo,
+      );
+      const aliasGenerator = new AliasDeltaGenerator(actorProvider);
+
+      const session: SimSession = {
+        db,
+        dbPath,
+        coreRepo,
+        bufferRepo,
+        ledgerRepo,
+        worldInstanceId: id,
+        scenarioName: state.scenarioName,
+        scenarioDescription: state.scenarioDescription,
+        turn: state.turn,
+        maxTurns: state.maxTurns,
+        entities: state.entities || [],
+        playerEntityId: state.playerEntityId,
+        entityIndex: state.entityIndex,
+        actorProvider,
+        validatorProvider,
+        decoderProvider,
+        timedeltaProvider,
+        handoffProvider,
+        embeddingProvider,
+        architect,
+        aliasGenerator,
+        log: state.log || [],
+        status: state.status,
+        error: state.error,
+        waitingEntity: state.waitingEntity,
+        aliasDoneForTurn: state.aliasDoneForTurn || false,
+        providerMappings: mappings,
+      };
+
+      this.sessions.set(id, session);
+      return this.snapshot(session);
+    } catch (err) {
+      console.error(`Failed to load session ${id}:`, err);
+      return null;
+    }
+  }
+
+  close(id: string): void {
+    const session = this.sessions.get(id);
+    if (session) {
+      session.db.close();
+      this.sessions.delete(id);
+    }
+  }
+
+  deleteSession(id: string): void {
+    const session = this.sessions.get(id);
+    if (session) {
+      session.db.close();
+      this.sessions.delete(id);
+    }
+    deleteSessionFile(id);
+  }
+
+  listSavedSessions(): SimSnapshot[] {
+    return listSavedSessions(this.sessions, (s) => this.snapshot(s));
+  }
+
+  getSnapshot(id: string): SimSnapshot | null {
+    const session = this.sessions.get(id);
+    return session ? this.snapshot(session) : null;
+  }
+
+  // ---------------------------------------------------------------------------
+  // Simulation stepping
+  // ---------------------------------------------------------------------------
+
+  async step(id: string): Promise<SimSnapshot | null> {
+    const session = this.sessions.get(id);
+    if (!session) return null;
+    if (session.status !== "running") return this.snapshot(session);
+
+    try {
+      if (session.turn > session.maxTurns) {
+        session.status = "done";
+        saveSession(session);
+        return this.snapshot(session);
+      }
+
+      // Start of turn: alias + handoff resolution before any entity acts
+      if (!session.aliasDoneForTurn && session.entityIndex === 0) {
+        await runAliasResolution(session);
+        await runHandoffResolution(session);
+        session.aliasDoneForTurn = true;
+        saveSession(session);
+        return this.snapshot(session);
+      }
+
+      // End of turn: advance to next turn
+      if (session.entityIndex >= session.entities.length) {
+        session.turn++;
+        session.entityIndex = 0;
+        session.aliasDoneForTurn = false;
+        saveSession(session);
+        return this.snapshot(session);
+      }
+
+      const info = session.entities[session.entityIndex];
+
+      if (!info.isAgent) {
+        session.entityIndex++;
+        saveSession(session);
+        return this.snapshot(session);
+      }
+
+      if (info.isPlayer) {
+        await preparePlayerTurn(session, info);
+        saveSession(session);
+        return this.snapshot(session);
+      }
+
+      await processNpcTurn(session, info);
+      session.entityIndex++;
+    } catch (err) {
+      session.status = "error";
+      session.error = err instanceof Error ? err.message : String(err);
+    }
+
+    saveSession(session);
+    return this.snapshot(session);
+  }
+
+  async submitPlayerAction(
+    id: string,
+    prose: string,
+  ): Promise<SimSnapshot | null> {
+    const session = this.sessions.get(id);
+    if (!session) return null;
+    if (session.status !== "waiting_player") return this.snapshot(session);
+    if (!session.waitingEntity) return this.snapshot(session);
+
+    const ctx = session.waitingEntity;
+    session.waitingEntity = undefined;
+    session.status = "running";
+
+    try {
+      await executePlayerAction(session, ctx, prose);
+      session.entityIndex++;
+    } catch (err) {
+      session.status = "error";
+      session.error = err instanceof Error ? err.message : String(err);
+    }
+
+    saveSession(session);
+    return this.snapshot(session);
+  }
+
+  // ---------------------------------------------------------------------------
+  // Utility
+  // ---------------------------------------------------------------------------
+
+  async regenerateAllEmbeddings(newProviderInstanceId?: string): Promise<void> {
+    if (!fs.existsSync(DATA_DIR)) return;
+
+    const files = fs
+      .readdirSync(DATA_DIR)
+      .filter((f) => f.startsWith("sim-") && f.endsWith(".db"));
+
+    const list = ProviderManager.list();
+    let inst = newProviderInstanceId
+      ? (list.find((p) => p.id === newProviderInstanceId) ?? null)
+      : null;
+    if (!inst || inst.type !== "embedding") {
+      inst = ProviderManager.getActive("embedding");
+    }
+
+    if (!inst) {
+      const envKey = process.env.GOOGLE_API_KEY || "";
+      if (envKey) {
+        inst = {
+          id: "regen-env-fallback",
+          name: "Gemini Embed (Env)",
+          providerName: "google-genai",
+          apiKey: envKey,
+          isActive: true,
+          modelName: "gemini-embedding-001",
+          type: "embedding",
+          maxContext: 0,
+        };
+      } else {
+        inst = {
+          id: "regen-mock-fallback",
+          name: "Mock Embed (Fallback)",
+          providerName: "mock",
+          apiKey: "",
+          isActive: true,
+          modelName: undefined,
+          type: "embedding",
+          maxContext: 0,
+        };
+      }
+    }
+
+    const embeddingProvider: IEmbeddingProvider = buildEmbeddingProvider(inst);
+
+    for (const file of files) {
+      const dbPath = path.join(DATA_DIR, file);
+      const fileId = file.replace(".db", "");
+      const activeSession = this.sessions.get(fileId);
+      const db = activeSession ? activeSession.db : new Database(dbPath);
+
+      try {
+        const rows = db
+          .prepare(`SELECT id, content FROM ledger_entries`)
+          .all() as { id: string; content: string }[];
+
+        for (const row of rows) {
+          const vector = await embeddingProvider.embed(row.content);
+          const buffer = Buffer.from(new Float32Array(vector).buffer);
+          db.prepare(
+            `UPDATE ledger_entries SET embedding = ? WHERE id = ?`,
+          ).run(buffer, row.id);
+        }
+      } catch (err) {
+        console.error(`Failed to regenerate embeddings for ${file}:`, err);
+      } finally {
+        if (!activeSession) db.close();
+      }
+    }
+  }
+
+  // ---------------------------------------------------------------------------
+  // Private
+  // ---------------------------------------------------------------------------
+
+  private snapshot(session: SimSession): SimSnapshot {
+    return {
+      id: session.worldInstanceId,
+      status: session.status,
+      turn: session.turn,
+      maxTurns: session.maxTurns,
+      scenarioName: session.scenarioName,
+      scenarioDescription: session.scenarioDescription,
+      entities: session.entities,
+      log: session.log,
+      entityIndex: session.entityIndex,
+      waitingEntity: session.waitingEntity,
+      error: session.error,
+    };
+  }
+}

apps/gui/src/lib/simulation/turn-executor.ts

@@ -0,0 +1,277 @@
+import {
+  ActorAgent,
+  ActorPromptBuilder,
+  buildBufferEntryForIntent,
+} from "@omnia/actor";
+import type { IActorProseGenerator } from "@omnia/actor";
+import type { SimSession } from "./types";
+import type {
+  EntityInfo,
+  IntentInfo,
+  LogEntry,
+  WaitingContext,
+} from "../simulation-types";
+
+// ---------------------------------------------------------------------------
+// Internal helpers
+// ---------------------------------------------------------------------------
+
+/** Prose generator that returns a fixed player-supplied string verbatim. */
+class FixedProseGenerator implements IActorProseGenerator {
+  constructor(private prose: string) {}
+
+  async generate(
+    entityId: string,
+    systemPrompt: string,
+    userContext: string,
+  ): Promise<string> {
+    void entityId;
+    void systemPrompt;
+    void userContext;
+    return this.prose;
+  }
+}
+
+/**
+ * Processes every intent produced by an actor turn:
+ * - Validates via Architect
+ * - Appends to actor's own buffer
+ * - Fan-outs to co-located observers for dialogue/action intents
+ *
+ * Extracted to eliminate verbatim duplication between NPC and player paths.
+ */
+async function processIntents(
+  // eslint-disable-next-line @typescript-eslint/no-explicit-any
+  intents: any[],
+  actorEntityId: string,
+  // eslint-disable-next-line @typescript-eslint/no-explicit-any
+  entity: any,
+  // eslint-disable-next-line @typescript-eslint/no-explicit-any
+  worldState: any,
+  session: SimSession,
+): Promise<IntentInfo[]> {
+  const intentInfos: IntentInfo[] = [];
+
+  for (const intent of intents) {
+    const outcome = await session.architect.processIntent(worldState, intent);
+    const ts = worldState.clock.get().toISOString();
+
+    intentInfos.push({
+      type: intent.type,
+      description: intent.description,
+      selfDescription: intent.selfDescription,
+      modifiers: intent.modifiers || [],
+      targetIds: intent.targetIds,
+      isValid: outcome.isValid,
+      reason: outcome.reason,
+      minutesToAdvance: outcome.timeDelta?.minutesToAdvance,
+    });
+
+    const actorEntry = buildBufferEntryForIntent(intent, ts, entity.locationId);
+    if (intent.type === "action") {
+      actorEntry.outcome = { isValid: outcome.isValid, reason: outcome.reason };
+    }
+    session.bufferRepo.save(actorEntry);
+
+    // Fan-out observable events to co-located entities
+    if (
+      entity.locationId &&
+      (intent.type === "dialogue" || intent.type === "action")
+    ) {
+      for (const [, other] of worldState.entities) {
+        if (
+          other.id !== actorEntityId &&
+          other.locationId === entity.locationId
+        ) {
+          const observerEntry = buildBufferEntryForIntent(
+            intent,
+            ts,
+            entity.locationId,
+          );
+          if (intent.type === "action") {
+            observerEntry.outcome = {
+              isValid: outcome.isValid,
+              reason: outcome.reason,
+            };
+          }
+          session.bufferRepo.save({ ...observerEntry, ownerId: other.id });
+        }
+      }
+    }
+  }
+
+  return intentInfos;
+}
+
+// ---------------------------------------------------------------------------
+// Exported turn functions
+// ---------------------------------------------------------------------------
+
+/**
+ * Builds the prompt for the player entity and sets the session to
+ * `waiting_player` so the next client call can supply the prose.
+ */
+export async function preparePlayerTurn(
+  session: SimSession,
+  info: EntityInfo,
+): Promise<void> {
+  const worldState = session.coreRepo.loadWorldState(session.worldInstanceId);
+  if (!worldState) throw new Error("World state lost");
+
+  const entity = worldState.getEntity(info.id);
+  if (!entity) throw new Error(`Entity "${info.id}" not found`);
+
+  const promptBuilder = new ActorPromptBuilder(
+    session.bufferRepo,
+    session.ledgerRepo,
+    20,
+  );
+  const { systemPrompt, userContext } = promptBuilder.build(worldState, entity);
+
+  session.waitingEntity = {
+    entityId: info.id,
+    name: info.name,
+    systemPrompt,
+    userContext,
+  };
+  session.status = "waiting_player";
+}
+
+/**
+ * Runs an autonomous NPC turn: generates prose via ActorAgent, validates
+ * and persists all intents, and appends a LogEntry to the session.
+ */
+export async function processNpcTurn(
+  session: SimSession,
+  info: EntityInfo,
+): Promise<void> {
+  const worldState = session.coreRepo.loadWorldState(session.worldInstanceId);
+  if (!worldState) throw new Error("World state lost");
+
+  const entity = worldState.getEntity(info.id);
+  if (!entity) throw new Error(`Entity "${info.id}" not found`);
+
+  const actor = new ActorAgent(
+    { actor: session.actorProvider, decoder: session.decoderProvider },
+    session.bufferRepo,
+    session.ledgerRepo,
+    20,
+  );
+  const result = await actor.act(worldState, entity);
+
+  const entry: LogEntry = {
+    turn: session.turn,
+    entityId: info.id,
+    entityName: info.name,
+    narrativeProse: result.narrativeProse,
+    intents: [],
+    timestamp: worldState.clock.get().toISOString(),
+  };
+
+  if (
+    session.actorProvider.lastCalls &&
+    session.actorProvider.lastCalls.length > 0
+  ) {
+    const actorCall =
+      session.actorProvider.lastCalls[
+        session.actorProvider.lastCalls.length - 1
+      ];
+    entry.rawPrompt = {
+      systemPrompt: actorCall.systemPrompt,
+      userContext: actorCall.userContext,
+    };
+    entry.usage = actorCall.usage;
+  }
+
+  if (
+    session.decoderProvider.lastCalls &&
+    session.decoderProvider.lastCalls.length > 0
+  ) {
+    const decoderCall =
+      session.decoderProvider.lastCalls[
+        session.decoderProvider.lastCalls.length - 1
+      ];
+    entry.decoderPrompt = {
+      systemPrompt: decoderCall.systemPrompt,
+      userContext: decoderCall.userContext,
+    };
+    entry.decoderUsage = decoderCall.usage;
+  }
+
+  entry.intents = await processIntents(
+    result.intents.intents,
+    info.id,
+    entity,
+    worldState,
+    session,
+  );
+
+  session.log.push(entry);
+  session.coreRepo.saveWorldState(worldState);
+}
+
+/**
+ * Executes the player's turn using the prose they supplied.
+ * Uses a `FixedProseGenerator` so the ActorAgent bypasses its LLM call and
+ * returns the player's text directly.
+ */
+export async function executePlayerAction(
+  session: SimSession,
+  ctx: WaitingContext,
+  prose: string,
+): Promise<void> {
+  const worldState = session.coreRepo.loadWorldState(session.worldInstanceId);
+  if (!worldState) throw new Error("World state lost");
+
+  const entity = worldState.getEntity(ctx.entityId);
+  if (!entity) throw new Error(`Player entity "${ctx.entityId}" not found`);
+
+  const playerActor = new ActorAgent(
+    { actor: session.actorProvider, decoder: session.decoderProvider },
+    session.bufferRepo,
+    session.ledgerRepo,
+    20,
+    new FixedProseGenerator(prose),
+  );
+
+  const result = await playerActor.act(worldState, entity);
+
+  const entry: LogEntry = {
+    turn: session.turn,
+    entityId: ctx.entityId,
+    entityName: ctx.name,
+    narrativeProse: result.narrativeProse,
+    intents: [],
+    timestamp: worldState.clock.get().toISOString(),
+    rawPrompt: {
+      systemPrompt: ctx.systemPrompt,
+      userContext: ctx.userContext,
+    },
+  };
+
+  if (
+    session.decoderProvider.lastCalls &&
+    session.decoderProvider.lastCalls.length > 0
+  ) {
+    const call =
+      session.decoderProvider.lastCalls[
+        session.decoderProvider.lastCalls.length - 1
+      ];
+    entry.decoderPrompt = {
+      systemPrompt: call.systemPrompt,
+      userContext: call.userContext,
+    };
+    entry.decoderUsage = call.usage;
+  }
+
+  entry.intents = await processIntents(
+    result.intents.intents,
+    ctx.entityId,
+    entity,
+    worldState,
+    session,
+  );
+
+  session.log.push(entry);
+  session.coreRepo.saveWorldState(worldState);
+}

apps/gui/src/lib/simulation/types.ts

@@ -0,0 +1,68 @@
+import type Database from "better-sqlite3";
+import type { SQLiteRepository } from "@omnia/core";
+import type { BufferRepository, LedgerRepository } from "@omnia/memory";
+import type { Architect, AliasDeltaGenerator } from "@omnia/architect";
+import type { ILLMProvider, IEmbeddingProvider } from "@omnia/llm";
+import type { EntityInfo, LogEntry, WaitingContext } from "../simulation-types";
+
+export type {
+  EntityInfo,
+  IntentInfo,
+  LogEntry,
+  SimSnapshot,
+  WaitingContext,
+} from "../simulation-types";
+
+// ---------------------------------------------------------------------------
+// Persisted state (written to sqlite gui_meta table as JSON)
+// ---------------------------------------------------------------------------
+
+export interface SavedState {
+  scenarioName: string;
+  scenarioDescription: string;
+  turn: number;
+  maxTurns: number;
+  entities: EntityInfo[];
+  playerEntityId: string | undefined;
+  entityIndex: number;
+  status: "running" | "waiting_player" | "done" | "error";
+  error?: string;
+  waitingEntity?: WaitingContext;
+  aliasDoneForTurn: boolean;
+  log: LogEntry[];
+  providerMappings: Record<string, string>;
+}
+
+// ---------------------------------------------------------------------------
+// In-memory session (held in SimulationManager.sessions Map)
+// ---------------------------------------------------------------------------
+
+export interface SimSession {
+  db: Database.Database;
+  dbPath: string;
+  coreRepo: SQLiteRepository;
+  bufferRepo: BufferRepository;
+  ledgerRepo: LedgerRepository;
+  worldInstanceId: string;
+  scenarioName: string;
+  scenarioDescription: string;
+  turn: number;
+  maxTurns: number;
+  entities: EntityInfo[];
+  playerEntityId: string | undefined;
+  entityIndex: number;
+  actorProvider: ILLMProvider;
+  validatorProvider: ILLMProvider;
+  decoderProvider: ILLMProvider;
+  timedeltaProvider: ILLMProvider;
+  handoffProvider: ILLMProvider;
+  embeddingProvider: IEmbeddingProvider;
+  architect: Architect;
+  aliasGenerator: AliasDeltaGenerator;
+  log: LogEntry[];
+  status: "running" | "waiting_player" | "done" | "error";
+  error?: string;
+  waitingEntity?: WaitingContext;
+  aliasDoneForTurn: boolean;
+  providerMappings: Record<string, string>;
+}

package.json

@@ -19,7 +19,8 @@
     "watch": "tsc -b --watch",
     "test": "vitest run --project unit",
     "test:watch": "vitest --project unit",
-    "test:evals": "vitest run --project evals"
+    "test:evals": "vitest run --project evals",
+    "setup-provider": "node packages/llm/dist/bin/setup-provider.js"
   },
   "keywords": [],
   "author": "sortedcord",
@@ -47,7 +48,12 @@
     "zod": "^4.4.3"
   },
   "dependencies": {
+    "@langchain/anthropic": "^0.3.11",
+    "@langchain/deepseek": "^1.1.5",
     "@langchain/google-genai": "^2.2.0",
+    "@langchain/groq": "^1.3.1",
+    "@langchain/ollama": "^0.2.3",
+    "@langchain/openai": "^0.3.17",
     "@langchain/openrouter": "^0.4.3",
     "@types/node": "^20.19.43",
     "dotenv": "^17.4.2"

packages/llm/README.md

@@ -0,0 +1,532 @@
+# @omnia/llm
+
+LLM abstraction layer providing pluggable, database-backed provider instances for generative and embedding tasks.
+
+## Architecture Overview
+
+The system is built around four layers:
+
+1. **Registry** — each provider class self-registers its metadata (id, envVar, capabilities, default model, etc.) via `static {}` blocks; `PROVIDER_REGISTRY` is derived from these registrations at runtime — there is no hand-maintained provider list
+2. **Provider Manager** — SQLite-backed CRUD for persisted provider instances, with env-var bootstrap driven by the registry
+3. **Provider Factory** — `buildLLMProvider(inst)` / `buildEmbeddingProvider(inst)` resolve a stored instance to a live provider class via the registry
+4. **Interfaces** — contracts that all providers implement
+
+```mermaid
+graph TD
+    subgraph Self-Registering Providers
+        GP["GeminiProvider"]
+        ORP["OpenRouterProvider"]
+        MP["MockLLMProvider"]
+        GEP["GeminiEmbeddingProvider"]
+        MEP["MockEmbeddingProvider"]
+    end
+
+    subgraph Registry
+        PR["ProviderRegistry\n(derived, not authored)"]
+    end
+
+    subgraph Storage
+        PM["ProviderManager\n(db.ts + bootstrap.ts + row-mapper.ts)"]
+        DB[("settings.db")]
+    end
+
+    subgraph Factory
+        PF["buildLLMProvider()\nbuildEmbeddingProvider()"]
+    end
+
+    GP -->|static block| PR
+    ORP -->|static block| PR
+    MP -->|static block| PR
+    GEP -->|static block| PR
+    MEP -->|static block| PR
+
+    PM -->|reads/writes| DB
+    PM -->|bootstrap from| PR
+
+    PF -->|looks up| PR
+    PF -->|instantiates| GP
+    PF -->|instantiates| ORP
+```
+
+## Core Interfaces
+
+Defined in [`llm.ts`](src/llm.ts):
+
+### `ILLMProvider`
+
+The primary contract for generative (text-to-structured-data) providers.
+
+| Member                                   | Type                      | Description                                                |
+| ---------------------------------------- | ------------------------- | ---------------------------------------------------------- |
+| `providerName`                           | `string`                  | Human-readable provider label                              |
+| `maxContext`                             | `number?`                 | Maximum context window in tokens                           |
+| `generateStructuredResponse<T>(request)` | `Promise<LLMResponse<T>>` | Sends a prompt + Zod schema → returns parsed, typed output |
+| `lastCalls`                              | `LLMCallRecord[]?`        | Audit trail of recent calls (prompts + usage)              |
+
+### `IEmbeddingProvider`
+
+Contract for text-to-vector embedding providers.
+
+| Member         | Type                | Description                                        |
+| -------------- | ------------------- | -------------------------------------------------- |
+| `providerName` | `string`            | Human-readable provider label                      |
+| `embed(text)`  | `Promise<number[]>` | Returns a dense vector embedding of the input text |
+
+### `LLMRequest<T>`
+
+Input to `generateStructuredResponse`:
+
+```typescript
+{
+  systemPrompt: string;    // System-level instructions
+  userContext:   string;    // User/task-specific context
+  schema:        T;         // Zod schema — output is validated against this
+  temperature?:  number;    // Sampling temperature (optional)
+}
+```
+
+### `LLMResponse<T>`
+
+Output from `generateStructuredResponse`:
+
+```typescript
+{
+  success: boolean;
+  data?:   T;               // Parsed, schema-validated output
+  error?:  string;          // Error message on failure
+  usage?:  {
+    inputTokens:          number;
+    outputTokens:         number;
+    totalTokens:          number;
+    modelName?:           string;
+    providerInstanceName?: string;
+    maxContext?:           number;
+  };
+}
+```
+
+### `ModelProviderInstance`
+
+The persisted configuration record for a single provider instance:
+
+```typescript
+{
+  id:           string;                     // Unique ID ("provider-<timestamp>")
+  name:         string;                     // User-facing name ("Gemini (Env)")
+  providerName: string;                     // Provider type key ("google-genai" | "openrouter" | "mock")
+  apiKey:       string;                     // API key
+  isActive:     boolean;                    // Whether this is the active instance for its type
+  modelName?:   string;                     // Specific model to use
+  type:         "generative" | "embedding"; // Instance category
+  maxContext?:  number;                     // Context window limit
+}
+```
+
+### `ModelProviderMeta`
+
+Static metadata for each available provider type (used by the UI's provider picker):
+
+| Member                  | Type     | Description                                                  |
+| ----------------------- | -------- | ------------------------------------------------------------ |
+| `id`                    | `string` | `"google-genai"` \| `"openrouter"` \| `"ollama"` \| `"mock"` |
+| `displayName`           | `string` | Human-readable name                                          |
+| `description`           | `string` | Human-readable description                                   |
+| `defaultModel`          | `string` | Default generative model                                     |
+| `defaultEmbeddingModel` | `string` | Default embedding model                                      |
+
+Provider metadata is **self-declared** by each provider class in a `static {}` block and collected into `PROVIDER_REGISTRY` (derived, not authored). The `getAvailableProviders()` function and `AVAILABLE_PROVIDERS` helper in [`llm.ts`](src/llm.ts) read from the registry at call time.
+
+## Provider Manager
+
+`ProviderManager` is a **static class** that provides full CRUD over provider instances, backed by a SQLite database (`data/settings.db` at the workspace root). Internally split across:
+
+- [`db.ts`](src/db.ts) — memoized DB handle + schema migrations (`PRAGMA user_version`)
+- [`bin/setup-provider.ts`](src/bin/setup-provider.ts) — CLI tool to set up provider instances in the database
+- [`row-mapper.ts`](src/row-mapper.ts) — `mapRow()` (written once, used everywhere)
+- [`provider-manager.ts`](src/provider-manager.ts) — thin CRUD: `list`, `create`, `delete`, `setActive`, `update`, `getActive`, `getMappings`, `setMapping`
+
+### Storage
+
+The database is auto-created on first access. The table schema:
+
+```sql
+CREATE TABLE IF NOT EXISTS provider_instances (
+  id           TEXT PRIMARY KEY,
+  name         TEXT NOT NULL,
+  providerName TEXT NOT NULL,
+  apiKey       TEXT NOT NULL,
+  isActive     INTEGER NOT NULL DEFAULT 0,
+  modelName    TEXT,
+  type         TEXT NOT NULL DEFAULT 'generative',
+  maxContext   INTEGER
+);
+```
+
+A second table stores per-task provider overrides:
+
+```sql
+CREATE TABLE IF NOT EXISTS provider_mappings (
+  task                TEXT PRIMARY KEY,
+  providerInstanceId  TEXT NOT NULL
+);
+```
+
+### API
+
+| Method                                                                    | Signature                         | Description                                                                               |
+| ------------------------------------------------------------------------- | --------------------------------- | ----------------------------------------------------------------------------------------- |
+| `list()`                                                                  | `→ ModelProviderInstance[]`       | Returns all saved instances                                                               |
+| `create(name, providerName, apiKey, modelName?, type?, maxContext?)`      | `→ ModelProviderInstance`         | Creates a new instance. Auto-activates if it's the first of its type                      |
+| `delete(id)`                                                              | `→ void`                          | Removes an instance. If it was active, auto-promotes the next instance of the same type   |
+| `setActive(id)`                                                           | `→ void`                          | Deactivates all instances of the same type, then activates the target                     |
+| `update(id, name, providerName, apiKey?, modelName?, type?, maxContext?)` | `→ void`                          | Updates an existing instance. If `apiKey` is empty/omitted, the existing key is preserved |
+| `getActive(type?)`                                                        | `→ ModelProviderInstance \| null` | Returns the currently active instance for the given type (`"generative"` by default)      |
+| `getMappings()`                                                           | `→ Record<string, string>`        | Returns all task → providerInstanceId mappings                                            |
+| `setMapping(task, providerInstanceId)`                                    | `→ void`                          | Sets or removes (if `providerInstanceId` is empty) a task-specific mapping                |
+
+### Active Instance Invariants
+
+- **Only one active instance per type** — `setActive()` deactivates all sibling instances before activating the target.
+- **Auto-promotion on delete** — if the deleted instance was active, the first remaining instance of the same type is promoted.
+- **Auto-activation on create** — if no active instance exists for the type, the new instance is automatically activated.
+
+### Manual Seeding via CLI
+
+Rather than automatically bootstrapping from environment variables at runtime, which adds runtime complexity, you can quickly seed the database using the CLI setup tool:
+
+#### Seeding All Environment-Variable Providers
+
+```bash
+pnpm setup-provider --all
+```
+
+This command auto-detects and inserts provider instances into `data/settings.db` for any registered providers whose corresponding environment variables (such as `GOOGLE_API_KEY`, `OPENAI_API_KEY`, etc.) are defined.
+
+#### Creating a Specific Provider Instance
+
+```bash
+pnpm setup-provider --provider google-genai --key YOUR_API_KEY [--name "My Gemini"] [--model gemini-2.5-flash] [--type generative] [--max-context 32768] [--endpoint url]
+```
+
+### Credential Resolution Cascade
+
+When initializing a provider (e.g. `new GeminiProvider()`):
+
+1. **Explicit Credentials** — If `apiKey`, `modelName`, etc. are passed directly to the constructor, they are used.
+2. **Active DB Instance** — If not explicitly passed, it looks up the active DB instance via `ProviderManager.getActive()`. If found and matches the provider, its API key and configuration are used.
+3. **Environment Fallback** — If there is no matching active DB instance, it resolves the key directly from the corresponding environment variable (e.g., `GOOGLE_API_KEY`) via `resolveCredentials`.
+
+## Available Providers
+
+### Google Gemini — `GeminiProvider`
+
+| Property                    | Value                                                        |
+| --------------------------- | ------------------------------------------------------------ |
+| **File**                    | [`providers/google-genai.ts`](src/providers/google-genai.ts) |
+| **Provider ID**             | `google-genai`                                               |
+| **SDK**                     | `@langchain/google-genai` (`ChatGoogleGenerativeAI`)         |
+| **Default Model**           | `gemini-2.5-flash`                                           |
+| **Default Embedding Model** | `gemini-embedding-001`                                       |
+| **Default Max Context**     | `32768`                                                      |
+| **Type**                    | Generative                                                   |
+
+**Key resolution** in the constructor follows this cascade:
+
+```
+1. Explicit apiKey argument       → use it
+2. ProviderManager.getActive()    → if providerName matches "google-genai"
+3. GOOGLE_API_KEY env var         → final fallback
+4. None found                     → throw Error
+```
+
+Also exports `GeminiEmbeddingProvider` (implements `IEmbeddingProvider`) using the same key resolution pattern but querying for the `"embedding"` type.
+
+### Anthropic Claude — `AnthropicProvider`
+
+| Property                    | Value                                                  |
+| --------------------------- | ------------------------------------------------------ |
+| **File**                    | [`providers/anthropic.ts`](src/providers/anthropic.ts) |
+| **Provider ID**             | `anthropic`                                            |
+| **SDK**                     | `@langchain/anthropic` (`ChatAnthropic`)               |
+| **Default Model**           | `claude-3-5-sonnet-latest`                             |
+| **Default Embedding Model** | _(none)_                                               |
+| **Default Max Context**     | `200000`                                               |
+| **Type**                    | Generative only (no embedding provider)                |
+
+**Key resolution** in the constructor follows this cascade:
+
+```
+1. Explicit apiKey argument       → use it
+2. ProviderManager.getActive()    → if providerName matches "anthropic"
+3. ANTHROPIC_API_KEY env var      → final fallback
+4. None found                     → throw Error
+```
+
+### Groq — `GroqProvider`
+
+| Property                    | Value                                        |
+| --------------------------- | -------------------------------------------- |
+| **File**                    | [`providers/groq.ts`](src/providers/groq.ts) |
+| **Provider ID**             | `groq`                                       |
+| **SDK**                     | `@langchain/groq` (`ChatGroq`)               |
+| **Default Model**           | `llama-3.3-70b-versatile`                    |
+| **Default Embedding Model** | _(none)_                                     |
+| **Default Max Context**     | `8192`                                       |
+| **Type**                    | Generative only (no embedding provider)      |
+
+**Key resolution** in the constructor follows this cascade:
+
+```
+1. Explicit apiKey argument       → use it
+2. ProviderManager.getActive()    → if providerName matches "groq"
+3. GROQ_API_KEY env var           → final fallback
+4. None found                     → throw Error
+```
+
+### DeepSeek — `DeepSeekProvider`
+
+| Property                    | Value                                                |
+| --------------------------- | ---------------------------------------------------- |
+| **File**                    | [`providers/deepseek.ts`](src/providers/deepseek.ts) |
+| **Provider ID**             | `deepseek`                                           |
+| **SDK**                     | `@langchain/deepseek` (`ChatDeepSeek`)               |
+| **Default Model**           | `deepseek-chat`                                      |
+| **Default Embedding Model** | _(none)_                                             |
+| **Default Max Context**     | `64000`                                              |
+| **Type**                    | Generative only (no embedding provider)              |
+
+**Key resolution** in the constructor follows this cascade:
+
+```
+1. Explicit apiKey argument       → use it
+2. ProviderManager.getActive()    → if providerName matches "deepseek"
+3. DEEPSEEK_API_KEY env var       → final fallback
+4. None found                     → throw Error
+```
+
+### OpenAI — `OpenAIProvider`
+
+| Property                    | Value                                                  |
+| --------------------------- | ------------------------------------------------------ |
+| **File**                    | [`providers/openai.ts`](src/providers/openai.ts)       |
+| **Provider ID**             | `openai`                                               |
+| **SDK**                     | `@langchain/openai` (`ChatOpenAI`, `OpenAIEmbeddings`) |
+| **Default Model**           | `gpt-4o-mini`                                          |
+| **Default Embedding Model** | `text-embedding-3-small`                               |
+| **Default Max Context**     | `128000`                                               |
+| **Type**                    | Generative + Embedding                                 |
+
+**Key resolution** in the constructor follows this cascade:
+
+```
+1. Explicit apiKey argument       → use it
+2. ProviderManager.getActive()    → if providerName matches "openai"
+3. OPENAI_API_KEY env var         → final fallback
+4. None found                     → throw Error
+```
+
+Also exports `OpenAIEmbeddingProvider` (implements `IEmbeddingProvider`) using the same key resolution pattern against the `"embedding"` type instance. The default embedding model is `text-embedding-3-small`.
+
+### OpenRouter — `OpenRouterProvider`
+
+| Property                    | Value                                                    |
+| --------------------------- | -------------------------------------------------------- |
+| **File**                    | [`providers/openrouter.ts`](src/providers/openrouter.ts) |
+| **Provider ID**             | `openrouter`                                             |
+| **SDK**                     | `@langchain/openrouter` (`ChatOpenRouter`)               |
+| **Default Model**           | `google/gemini-2.5-flash`                                |
+| **Default Embedding Model** | `openai/text-embedding-3-small`                          |
+| **Default Max Context**     | `32768`                                                  |
+| **Type**                    | Generative only (no embedding provider)                  |
+
+Same three-step key resolution as Gemini (`explicit → ProviderManager → env var`), using `OPENROUTER_API_KEY`.
+
+### Ollama — `OllamaProvider`
+
+| Property                    | Value                                            |
+| --------------------------- | ------------------------------------------------ |
+| **File**                    | [`providers/ollama.ts`](src/providers/ollama.ts) |
+| **Provider ID**             | `ollama`                                         |
+| **SDK**                     | `@langchain/ollama` (`ChatOllama`)               |
+| **Default Model**           | `llama3.1`                                       |
+| **Default Embedding Model** | `nomic-embed-text`                               |
+| **Default Max Context**     | `32768`                                          |
+| **Type**                    | Generative + Embedding                           |
+
+Ollama runs **locally** — no API key is required. The `endpointUrl` field in `ModelProviderInstance` stores the Ollama server base URL (default: `http://localhost:11434`).
+
+**Key resolution** in the constructor:
+
+```
+1. Explicit baseUrl argument        → use it
+2. ProviderManager.getActive()      → if providerName matches "ollama"
+                                      (endpointUrl field = base URL)
+3. Default                          → http://localhost:11434
+```
+
+Also exports `OllamaEmbeddingProvider` (implements `IEmbeddingProvider`), which uses the same resolution pattern against the `"embedding"` type instance. The default embedding model is `nomic-embed-text`.
+
+> [!TIP]
+> To get started: `ollama pull llama3.1` and `ollama pull nomic-embed-text`. Then create a provider instance with `endpointUrl` = `http://localhost:11434`.
+
+### Mock — `MockLLMProvider`
+
+| Property        | Value                                        |
+| --------------- | -------------------------------------------- |
+| **File**        | [`providers/mock.ts`](src/providers/mock.ts) |
+| **Provider ID** | `mock`                                       |
+| **Type**        | Generative + Embedding                       |
+
+Stateless mock for testing and offline development:
+
+- **Generative** (`MockLLMProvider`): Takes an array of canned responses at construction. Returns them in order, one per call. Returns `{ success: false, error: "Mock responses exhausted" }` when depleted.
+- **Embedding** (`MockEmbeddingProvider`): Returns a deterministic 768-dimensional vector derived from the input text using `Math.sin`.
+
+## Provider Resolution (Runtime)
+
+The [`resolveProviders()`](../../../apps/gui/src/lib/simulation/provider-resolver.ts) function (in `apps/gui`) instantiates all providers needed for a simulation session. It resolves **six** provider slots:
+
+| Slot                | Type       | Task Key           |
+| ------------------- | ---------- | ------------------ |
+| `actorProvider`     | Generative | `"actor-prose"`    |
+| `validatorProvider` | Generative | `"llm-validator"`  |
+| `decoderProvider`   | Generative | `"intent-decoder"` |
+| `timedeltaProvider` | Generative | `"timedelta"`      |
+| `handoffProvider`   | Generative | `"handoff"`        |
+| `embeddingProvider` | Embedding  | `"embeddings"`     |
+
+### Generative Resolution Order
+
+For each generative slot (`resolveGenerative(task)`):
+
+```
+1. Task-specific mapping     → mappings[task] → find instance by ID
+2. Active generative instance → ProviderManager.getActive("generative")
+3. Fallback instance         → options.fallbackInstance (if provided)
+4. GOOGLE_API_KEY env var    → auto-create via ProviderManager.create()
+5. No provider available     → throw Error (if required) or MockLLMProvider
+```
+
+### Embedding Resolution Order
+
+For the embedding slot (`resolveEmbedding()`):
+
+```
+1. Task-specific mapping     → mappings["embeddings"] → find instance by ID
+2. Active embedding instance → ProviderManager.getActive("embedding")
+3. GOOGLE_API_KEY env var    → auto-create via ProviderManager.create()
+4. No provider available     → throw Error (if required) or MockEmbeddingProvider
+```
+
+### Instance → Class Mapping
+
+The `buildLLMProvider()` and `buildEmbeddingProvider()` functions perform the final dispatch:
+
+| `providerName`    | Generative Class     | Embedding Class           |
+| ----------------- | -------------------- | ------------------------- |
+| `"google-genai"`  | `GeminiProvider`     | `GeminiEmbeddingProvider` |
+| `"openai"`        | `OpenAIProvider`     | `OpenAIEmbeddingProvider` |
+| `"openrouter"`    | `OpenRouterProvider` | _(falls through to mock)_ |
+| `"ollama"`        | `OllamaProvider`     | `OllamaEmbeddingProvider` |
+| `"anthropic"`     | `AnthropicProvider`  | _(falls through to mock)_ |
+| `"groq"`          | `GroqProvider`       | _(falls through to mock)_ |
+| `"deepseek"`      | `DeepSeekProvider`   | _(falls through to mock)_ |
+| _(anything else)_ | `MockLLMProvider`    | `MockEmbeddingProvider`   |
+
+## Model Listing and Discovery
+
+The `ModelLister` class provides a unified interface to dynamically query available models from the remote provider APIs.
+
+### Caching and TTL
+
+All list requests are cached in-memory with a **5-minute TTL** (`300,000ms`) to prevent rapid, repetitive remote API requests and avoid rate limit exhaustion.
+
+- **Cache Key**: Generated using `providerName` combined with either the `apiKey` or `endpointUrl`.
+- **Invalidation**: Call `ModelLister.invalidateCache(providerName, apiKey, endpointUrl)` to clear cache for specific instances, or `ModelLister.clearCache()` to wipe all lists.
+
+### Provider Integration Details
+
+| Provider          | Endpoint                     | Auth Header             | Pagination                  |
+| ----------------- | ---------------------------- | ----------------------- | --------------------------- |
+| **Google Gemini** | `GET /v1beta/models?key=KEY` | Query Param             | ✅ Loop via `nextPageToken` |
+| **OpenAI**        | `GET /v1/models`             | `Authorization: Bearer` | ❌                          |
+| **Anthropic**     | `GET /v1/models`             | `x-api-key`             | ✅ Loop via `after_id`      |
+| **Groq**          | `GET /openai/v1/models`      | `Authorization: Bearer` | ❌                          |
+| **DeepSeek**      | `GET /models`                | `Authorization: Bearer` | ❌                          |
+| **Ollama**        | `GET /api/tags`              | None (Local)            | ❌                          |
+| **OpenRouter**    | `GET /api/v1/models`         | Optional Bearer         | ❌                          |
+| **Mock**          | Instant return (no fetch)    | —                       | —                           |
+
+### Methods
+
+- `listModels(providerName: string, apiKey: string, endpointUrl?: string): Promise<ModelInfo[]>`
+- `invalidateCache(providerName: string, apiKey: string, endpointUrl?: string): void`
+- `clearCache(): void`
+
+---
+
+## Structured Output
+
+All real providers use LangChain's `.withStructuredOutput(schema, { includeRaw: true })` pattern:
+
+```typescript
+const structuredModel = this.model.withStructuredOutput(request.schema, {
+  includeRaw: true,
+});
+const result = await structuredModel.invoke([
+  { role: "system", content: request.systemPrompt },
+  { role: "user", content: request.userContext },
+]);
+```
+
+This sends the Zod schema to the model as a structured output constraint. The response includes both `parsed` (schema-validated data) and `raw` (full API response with usage metadata).
+
+## Configuration
+
+[`config.ts`](src/config.ts) parses environment variables using Zod:
+
+| Variable             | Required | Description              |
+| -------------------- | -------- | ------------------------ |
+| `GOOGLE_API_KEY`     | No       | Google Gemini API key    |
+| `OPENAI_API_KEY`     | No       | OpenAI API key           |
+| `OPENROUTER_API_KEY` | No       | OpenRouter API key       |
+| `ANTHROPIC_API_KEY`  | No       | Anthropic Claude API key |
+| `GROQ_API_KEY`       | No       | Groq API key             |
+| `DEEPSEEK_API_KEY`   | No       | DeepSeek API key         |
+
+Env var keys are derived from `PROVIDER_REGISTRY` — each provider's `envVar` field is read by `getLlmConfig()` to build the zod schema lazily. Adding a new provider with `envVar: "NEW_KEY"` automatically adds it to config validation.
+
+## File Map
+
+```
+packages/llm/
+├── src/
+│   ├── index.ts              # Re-exports everything
+│   ├── llm.ts                # Interfaces, types, getAvailableProviders()
+│   ├── registry.ts           # ProviderRegistry (derived), registerProvider/registerGenerative/registerEmbedding
+│   ├── base-provider.ts      # BaseLLMProvider (shared generateStructuredResponse), resolveCredentials
+│   ├── config.ts             # Env var parsing (lazy, registry-derived Zod)
+│   ├── model-lister.ts       # ModelLister (cache + fetchWithTimeout), fetchOpenAICompatibleModels
+│   ├── provider-factory.ts   # buildLLMProvider() / buildEmbeddingProvider() (registry lookup)
+│   ├── provider-manager.ts   # ProviderManager (thin CRUD)
+│   ├── db.ts                 # Memoized DB handle + migrations
+│   ├── row-mapper.ts         # mapRow()
+│   ├── bin/
+│   │   └── setup-provider.ts # CLI tool to set up provider instances in the database
+│   └── providers/
+│       ├── google-genai.ts   # GeminiProvider + GeminiEmbeddingProvider (self-registering)
+│       ├── ollama.ts         # OllamaProvider + OllamaEmbeddingProvider
+│       ├── openrouter.ts     # OpenRouterProvider
+│       ├── anthropic.ts      # AnthropicProvider
+│       ├── openai.ts         # OpenAIProvider + OpenAIEmbeddingProvider
+│       ├── groq.ts           # GroqProvider
+│       ├── deepseek.ts       # DeepSeekProvider
+│       └── mock.ts           # MockLLMProvider + MockEmbeddingProvider
+├── tests/
+│   ├── mock.test.ts
+│   ├── openrouter.test.ts
+│   ├── model-lister.test.ts  # ModelLister cache and fetch logic unit tests
+│   ├── provider-manager.test.ts
+│   └── cli.test.ts           # Integration tests for setup-provider CLI tool
+└── package.json
+```

packages/llm/package.json

@@ -6,6 +6,9 @@
   "exports": {
     ".": "./dist/index.js"
   },
+  "scripts": {
+    "setup-provider": "node ./dist/bin/setup-provider.js"
+  },
   "dependencies": {
     "@types/node": "^26.1.0",
     "better-sqlite3": "^12.11.1"

packages/llm/src/base-provider.ts

@@ -0,0 +1,112 @@
+import { z } from "zod";
+import type {
+  ILLMProvider,
+  LLMRequest,
+  LLMResponse,
+  LLMCallRecord,
+} from "./llm.js";
+import { ProviderManager } from "./provider-manager.js";
+import { getLlmConfig } from "./config.js";
+
+export interface ResolvedCredentials {
+  key: string | undefined;
+  model: string | undefined;
+  providerInstanceName: string | undefined;
+  maxContext: number | undefined;
+}
+
+export function resolveCredentials(opts: {
+  explicitKey?: string;
+  explicitModel?: string;
+  explicitProviderInstanceName?: string;
+  explicitMaxContext?: number;
+  providerId: string;
+  envVarName: string;
+  type: "generative" | "embedding";
+}): ResolvedCredentials {
+  let key = opts.explicitKey;
+  let model = opts.explicitModel;
+  let providerInstanceName = opts.explicitProviderInstanceName;
+  let maxContext = opts.explicitMaxContext;
+
+  if (!key) {
+    const active = ProviderManager.getActive(opts.type);
+    if (active && active.providerName === opts.providerId) {
+      key = active.apiKey;
+      if (!model) model = active.modelName;
+      if (!providerInstanceName) providerInstanceName = active.name;
+      if (maxContext === undefined) maxContext = active.maxContext;
+    }
+  }
+
+  if (!key) {
+    const cfg = getLlmConfig();
+    key = cfg[opts.envVarName];
+    if (!providerInstanceName && key) {
+      providerInstanceName = "Environment Variable";
+    }
+  }
+
+  return { key, model, providerInstanceName, maxContext };
+}
+
+export abstract class BaseLLMProvider implements ILLMProvider {
+  abstract providerName: string;
+  protected abstract readonly model: unknown;
+  protected abstract modelNameUsed: string;
+  protected abstract providerInstanceName?: string;
+  protected abstract maxContextUsed?: number;
+  protected abstract defaultMaxContext: number;
+  lastCalls: LLMCallRecord[] = [];
+
+  async generateStructuredResponse<T extends z.ZodTypeAny>(
+    request: LLMRequest<T>,
+  ): Promise<LLMResponse<z.infer<T>>> {
+    const structuredModel = (
+      this.model as {
+        withStructuredOutput(
+          s: z.ZodTypeAny,
+          o: { includeRaw: true },
+        ): {
+          invoke(m: unknown): Promise<unknown>;
+        };
+      }
+    ).withStructuredOutput(request.schema, { includeRaw: true });
+    const result = (await structuredModel.invoke([
+      { role: "system", content: request.systemPrompt },
+      { role: "user", content: request.userContext },
+    ])) as unknown as {
+      parsed?: z.infer<T>;
+      raw?: {
+        usage_metadata?: {
+          input_tokens?: number;
+          output_tokens?: number;
+          total_tokens?: number;
+        };
+      };
+    };
+
+    const parsed = result?.parsed;
+    const raw = result?.raw;
+
+    const usage = {
+      inputTokens: raw?.usage_metadata?.input_tokens || 0,
+      outputTokens: raw?.usage_metadata?.output_tokens || 0,
+      totalTokens: raw?.usage_metadata?.total_tokens || 0,
+      modelName: this.modelNameUsed,
+      providerInstanceName: this.providerInstanceName || "Default",
+      maxContext:
+        this.maxContextUsed !== undefined
+          ? this.maxContextUsed
+          : this.defaultMaxContext,
+    };
+
+    this.lastCalls.push({
+      systemPrompt: request.systemPrompt,
+      userContext: request.userContext,
+      usage,
+    });
+
+    return { success: true, data: parsed, usage };
+  }
+}

packages/llm/src/bin/setup-provider.ts

@@ -0,0 +1,193 @@
+#!/usr/bin/env node
+import { ProviderRegistry, ProviderManager } from "../index.js";
+import dotenv from "dotenv";
+import path from "path";
+import fs from "fs";
+
+// Load dotenv from workspace root
+function loadEnv() {
+  let current = process.cwd();
+  while (current !== "/" && current !== path.parse(current).root) {
+    if (fs.existsSync(path.join(current, "pnpm-workspace.yaml"))) {
+      dotenv.config({ path: path.join(current, ".env") });
+      return;
+    }
+    current = path.dirname(current);
+  }
+  dotenv.config();
+}
+
+loadEnv();
+
+function printHelp() {
+  console.log(`
+Usage:
+  node packages/llm/dist/bin/setup-provider.js [options]
+
+Options:
+  --provider <id>     ID of the provider (e.g. google-genai, openai, anthropic, groq, etc.)
+  --name <name>       Display name for the instance (default: provider displayName)
+  --key <key>         API key (default: loaded from the provider's env variable, e.g. GOOGLE_API_KEY)
+  --model <model>     Model name (default: provider's default model)
+  --type <type>       "generative" | "embedding" (default: "generative")
+  --max-context <num> Max context window tokens (default: provider's default max context)
+  --endpoint <url>    Custom endpoint URL (optional)
+  --all               Auto-detect and seed all providers whose environment variables are set
+  -h, --help          Show this help message
+
+Registered Providers:
+${ProviderRegistry.all()
+  .map((p) => `  - ${p.id} (${p.displayName}) [Env: ${p.envVar || "None"}]`)
+  .join("\n")}
+`);
+}
+
+async function main() {
+  const args = process.argv.slice(2);
+  if (args.includes("-h") || args.includes("--help") || args.length === 0) {
+    printHelp();
+    process.exit(0);
+  }
+
+  const options: Record<string, string> = {};
+  for (let i = 0; i < args.length; i++) {
+    const arg = args[i];
+    if (arg.startsWith("--")) {
+      const key = arg.slice(2);
+      const nextVal = args[i + 1];
+      if (nextVal && !nextVal.startsWith("--")) {
+        options[key] = nextVal;
+        i++;
+      } else {
+        options[key] = "true";
+      }
+    }
+  }
+
+  if (options.all === "true") {
+    // Seed all providers from environment variables
+    const existing = ProviderManager.list();
+    let seededCount = 0;
+
+    for (const def of ProviderRegistry.all()) {
+      if (!def.envVar) continue;
+      const key = process.env[def.envVar]?.trim();
+      if (!key) continue;
+
+      if (def.capabilities.generative) {
+        const hasGen = existing.some(
+          (p) => p.providerName === def.id && p.type === "generative",
+        );
+        if (!hasGen) {
+          const name = `${def.displayName} (CLI)`;
+          ProviderManager.create(
+            name,
+            def.id,
+            key,
+            def.defaultModel,
+            "generative",
+            def.defaultMaxContext,
+          );
+          console.log(`Created generative instance: ${name}`);
+          seededCount++;
+        }
+      }
+
+      if (def.capabilities.embedding) {
+        const hasEmbed = existing.some(
+          (p) => p.providerName === def.id && p.type === "embedding",
+        );
+        if (!hasEmbed) {
+          const name = `${def.displayName} Embed (CLI)`;
+          ProviderManager.create(
+            name,
+            def.id,
+            key,
+            def.defaultEmbeddingModel || "",
+            "embedding",
+            0,
+          );
+          console.log(`Created embedding instance: ${name}`);
+          seededCount++;
+        }
+      }
+    }
+
+    if (seededCount === 0) {
+      console.log(
+        "No new provider instances seeded. (Either already existed or env vars not set)",
+      );
+    } else {
+      console.log(`Successfully seeded ${seededCount} provider instance(s).`);
+    }
+    process.exit(0);
+  }
+
+  const providerId = options.provider;
+  if (!providerId) {
+    console.error("Error: --provider <id> or --all is required.");
+    printHelp();
+    process.exit(1);
+  }
+
+  const def = ProviderRegistry.get(providerId);
+  if (!def) {
+    console.error(`Error: Provider '${providerId}' is not registered.`);
+    console.error(
+      `Available providers: ${ProviderRegistry.all()
+        .map((p) => p.id)
+        .join(", ")}`,
+    );
+    process.exit(1);
+  }
+
+  const type = (options.type === "embedding" ? "embedding" : "generative") as
+    "generative" | "embedding";
+
+  // Resolve key
+  let apiKey: string | undefined = options.key;
+  if (!apiKey && def.envVar) {
+    apiKey = process.env[def.envVar]?.trim();
+  }
+  if (!apiKey) {
+    console.error(
+      `Error: API Key is required. Please set ${def.envVar || "the environment variable"} or pass --key <apiKey>.`,
+    );
+    process.exit(1);
+  }
+
+  // Resolve model
+  const defaultModel =
+    type === "embedding" ? def.defaultEmbeddingModel || "" : def.defaultModel;
+  const modelName = options.model || defaultModel;
+
+  // Resolve name
+  const name = options.name || `${def.displayName} (CLI)`;
+
+  // Resolve maxContext
+  const maxContext = options["max-context"]
+    ? parseInt(options["max-context"], 10)
+    : type === "embedding"
+      ? 0
+      : def.defaultMaxContext;
+
+  const endpointUrl = options.endpoint;
+
+  const instance = ProviderManager.create(
+    name,
+    def.id,
+    apiKey,
+    modelName,
+    type,
+    maxContext,
+    endpointUrl,
+  );
+
+  console.log(`Successfully created provider instance:`);
+  console.log(JSON.stringify(instance, null, 2));
+}
+
+main().catch((err) => {
+  console.error(err);
+  process.exit(1);
+});

packages/llm/src/config.ts

@@ -1,8 +1,25 @@
 import { z } from "zod";
+import { ProviderRegistry } from "./registry.js";
 
-const LLMConfigSchema = z.object({
-  GOOGLE_API_KEY: z.string().optional(),
-  OPENROUTER_API_KEY: z.string().optional(),
-});
+let _config: Record<string, string | undefined> | null = null;
 
-export const llmConfig = LLMConfigSchema.parse(process.env);
+export function getLlmConfig(): Record<string, string | undefined> {
+  if (!_config) {
+    const envVars: string[] = [];
+    for (const def of ProviderRegistry.all()) {
+      if (def.envVar && !envVars.includes(def.envVar)) {
+        envVars.push(def.envVar);
+      }
+    }
+    const shape: Record<string, z.ZodOptional<z.ZodString>> = {};
+    for (const key of envVars) {
+      shape[key] = z.string().optional();
+    }
+    _config = z.object(shape).parse(process.env);
+  }
+  return _config;
+}
+
+export function resetLlmConfig(): void {
+  _config = null;
+}

packages/llm/src/db.ts

@@ -0,0 +1,82 @@
+import Database from "better-sqlite3";
+import type BetterSqlite3 from "better-sqlite3";
+import path from "path";
+import fs from "fs";
+
+let _db: BetterSqlite3.Database | null = null;
+let _dbPathOverride: string | null = null;
+
+export function setDbPath(p: string | null) {
+  if (_dbPathOverride !== p) {
+    _db?.close();
+    _db = null;
+    _dbPathOverride = p;
+  }
+}
+
+function findDbPath(): string {
+  if (process.env.OMNIA_DB_PATH) {
+    const dir = path.dirname(process.env.OMNIA_DB_PATH);
+    if (!fs.existsSync(dir)) {
+      fs.mkdirSync(dir, { recursive: true });
+    }
+    return process.env.OMNIA_DB_PATH;
+  }
+  let current = process.cwd();
+  while (current !== "/" && current !== path.parse(current).root) {
+    if (fs.existsSync(path.join(current, "pnpm-workspace.yaml"))) {
+      const dbDir = path.resolve(current, "data");
+      if (!fs.existsSync(dbDir)) {
+        fs.mkdirSync(dbDir, { recursive: true });
+      }
+      return path.join(dbDir, "settings.db");
+    }
+    current = path.dirname(current);
+  }
+  const dbDir = path.resolve(process.cwd(), "data");
+  if (!fs.existsSync(dbDir)) {
+    fs.mkdirSync(dbDir, { recursive: true });
+  }
+  return path.join(dbDir, "settings.db");
+}
+
+function runMigrations(db: BetterSqlite3.Database): void {
+  const version = db.pragma("user_version", { simple: true }) as number;
+
+  if (version < 1) {
+    db.prepare(
+      `
+      CREATE TABLE IF NOT EXISTS provider_instances (
+        id TEXT PRIMARY KEY,
+        name TEXT NOT NULL,
+        providerName TEXT NOT NULL,
+        apiKey TEXT NOT NULL,
+        isActive INTEGER NOT NULL DEFAULT 0,
+        modelName TEXT,
+        type TEXT NOT NULL DEFAULT 'generative',
+        maxContext INTEGER,
+        endpointUrl TEXT
+      )
+      `,
+    ).run();
+    db.pragma("user_version = 1");
+  }
+
+  db.prepare(
+    `
+    CREATE TABLE IF NOT EXISTS provider_mappings (
+      task TEXT PRIMARY KEY,
+      providerInstanceId TEXT NOT NULL
+    )
+    `,
+  ).run();
+}
+
+export function getDb(): BetterSqlite3.Database {
+  if (!_db) {
+    const dbPath = _dbPathOverride ?? findDbPath();
+    _db = new Database(dbPath);
+    runMigrations(_db);
+  }
+  return _db;
+}

packages/llm/src/index.ts

@@ -1,6 +1,14 @@
 export * from "./llm.js";
 export * from "./config.js";
+export * from "./registry.js";
+export * from "./provider-factory.js";
+export * from "./model-lister.js";
+export * from "./provider-manager.js";
 export * from "./providers/google-genai.js";
 export * from "./providers/mock.js";
+export * from "./providers/ollama.js";
 export * from "./providers/openrouter.js";
-export * from "./provider-manager.js";
+export * from "./providers/anthropic.js";
+export * from "./providers/openai.js";
+export * from "./providers/groq.js";
+export * from "./providers/deepseek.js";

packages/llm/src/llm.ts

@@ -1,4 +1,5 @@
 import { z } from "zod";
+import { ProviderRegistry } from "./registry.js";
 
 export interface LLMRequest<T extends z.ZodTypeAny> {
   systemPrompt: string;
@@ -57,6 +58,7 @@ export interface ModelProviderInstance {
   modelName?: string;
   type: "generative" | "embedding";
   maxContext?: number;
+  endpointUrl?: string;
 }
 
 export interface ModelProviderMeta {
@@ -67,27 +69,21 @@ export interface ModelProviderMeta {
   defaultEmbeddingModel: string;
 }
 
-export const AVAILABLE_PROVIDERS: ModelProviderMeta[] = [
-  {
-    id: "google-genai",
-    displayName: "Google Gemini",
-    description: "Official Gemini integration using Google Gen AI SDK",
-    defaultModel: "gemini-2.5-flash",
-    defaultEmbeddingModel: "gemini-embedding-001",
+export function getAvailableProviders(): ModelProviderMeta[] {
+  return ProviderRegistry.all().map((def) => ({
+    id: def.id,
+    displayName: def.displayName,
+    description: def.description,
+    defaultModel: def.defaultModel,
+    defaultEmbeddingModel: def.defaultEmbeddingModel || "",
+  }));
+}
+
+export const AVAILABLE_PROVIDERS = {
+  get count(): number {
+    return getAvailableProviders().length;
   },
-  {
-    id: "openrouter",
-    displayName: "OpenRouter",
-    description:
-      "Multi-model router supporting Anthropic, OpenAI, DeepSeek, and local models",
-    defaultModel: "google/gemini-2.5-flash",
-    defaultEmbeddingModel: "openai/text-embedding-3-small",
+  toArray(): ModelProviderMeta[] {
+    return getAvailableProviders();
   },
-  {
-    id: "mock",
-    displayName: "Mock LLM Provider",
-    description: "Stateless mock provider for testing and offline development",
-    defaultModel: "mock",
-    defaultEmbeddingModel: "mock-embeddings",
-  },
-];
+};

packages/llm/src/model-lister.ts

@@ -0,0 +1,105 @@
+/**
+ * ModelLister — fetches available models from each provider's REST API.
+ * Results are cached in-memory with a 5-minute TTL to avoid repeated calls.
+ */
+
+import { ProviderRegistry } from "./registry.js";
+
+export interface ModelInfo {
+  id: string;
+  name: string;
+  ownedBy?: string;
+}
+
+interface CacheEntry {
+  models: ModelInfo[];
+  fetchedAt: number;
+}
+
+const CACHE_TTL_MS = 5 * 60 * 1000;
+const FETCH_TIMEOUT_MS = 10_000;
+
+const modelCache = new Map<string, CacheEntry>();
+
+function cacheKey(
+  providerName: string,
+  apiKey: string,
+  endpointUrl?: string,
+): string {
+  return `${providerName}:${endpointUrl || apiKey}`;
+}
+
+export async function fetchWithTimeout(
+  url: string,
+  init?: RequestInit,
+): Promise<Response> {
+  const controller = new AbortController();
+  const timer = setTimeout(() => controller.abort(), FETCH_TIMEOUT_MS);
+  try {
+    return await fetch(url, { ...init, signal: controller.signal });
+  } finally {
+    clearTimeout(timer);
+  }
+}
+
+export async function fetchOpenAICompatibleModels(
+  baseUrl: string,
+  apiKey: string,
+): Promise<ModelInfo[]> {
+  const res = await fetchWithTimeout(`${baseUrl}/models`, {
+    headers: {
+      Authorization: `Bearer ${apiKey}`,
+      Accept: "application/json",
+    },
+  });
+  if (!res.ok) return [];
+
+  const json = (await res.json()) as {
+    data?: { id: string; owned_by?: string; name?: string }[];
+  };
+
+  return (json.data ?? []).map((m) => ({
+    id: m.id,
+    name: m.name || m.id,
+    ownedBy: m.owned_by,
+  }));
+}
+
+export class ModelLister {
+  static async listModels(
+    providerName: string,
+    apiKey: string,
+    endpointUrl?: string,
+  ): Promise<ModelInfo[]> {
+    const key = cacheKey(providerName, apiKey, endpointUrl);
+    const cached = modelCache.get(key);
+    if (cached && Date.now() - cached.fetchedAt < CACHE_TTL_MS) {
+      return cached.models;
+    }
+
+    const def = ProviderRegistry.get(providerName);
+    let models: ModelInfo[] = [];
+    try {
+      if (def?.listModels) {
+        models = await def.listModels(apiKey, endpointUrl);
+      }
+    } catch {
+      models = [];
+    }
+
+    modelCache.set(key, { models, fetchedAt: Date.now() });
+    return models;
+  }
+
+  static invalidateCache(
+    providerName: string,
+    apiKey: string,
+    endpointUrl?: string,
+  ): void {
+    modelCache.delete(cacheKey(providerName, apiKey, endpointUrl));
+  }
+
+  static clearCache(): void {
+    modelCache.clear();
+  }
+}

packages/llm/src/provider-factory.ts

@@ -0,0 +1,21 @@
+import type {
+  ILLMProvider,
+  IEmbeddingProvider,
+  ModelProviderInstance,
+} from "./llm.js";
+import { MockLLMProvider, MockEmbeddingProvider } from "./providers/mock.js";
+import { ProviderRegistry } from "./registry.js";
+
+export function buildLLMProvider(inst: ModelProviderInstance): ILLMProvider {
+  const def = ProviderRegistry.get(inst.providerName);
+  return def?.generativeCreate?.(inst) ?? new MockLLMProvider([]);
+}
+
+export function buildEmbeddingProvider(
+  inst: ModelProviderInstance,
+): IEmbeddingProvider {
+  const def = ProviderRegistry.get(inst.providerName);
+  return (
+    def?.embeddingCreate?.(inst) ?? new MockEmbeddingProvider(inst.modelName)
+  );
+}

packages/llm/src/provider-manager.ts

@@ -1,196 +1,16 @@
-import Database from "better-sqlite3";
-import path from "path";
-import fs from "fs";
 import type { ModelProviderInstance } from "./llm.js";
+import { getDb } from "./db.js";
+import { mapRow, type DbRow } from "./row-mapper.js";
 
-let dbPathOverride: string | null = null;
-let hasBootstrapped = false;
-
-export function setDbPathOverride(p: string | null) {
-  dbPathOverride = p;
-}
-
-export function resetHasBootstrapped() {
-  hasBootstrapped = false;
-}
-
-function getWorkspaceRoot() {
-  let current = process.cwd();
-  while (current !== "/" && current !== path.parse(current).root) {
-    if (
-      fs.existsSync(path.join(current, "pnpm-workspace.yaml")) ||
-      fs.existsSync(path.join(current, "package.json"))
-    ) {
-      if (fs.existsSync(path.join(current, "pnpm-workspace.yaml"))) {
-        return current;
-      }
-    }
-    current = path.dirname(current);
-  }
-  return process.cwd();
-}
-
-function getSettingsDb() {
-  let dbPath: string;
-  if (dbPathOverride) {
-    dbPath = dbPathOverride;
-  } else {
-    const wsRoot = getWorkspaceRoot();
-    const dbDir = path.resolve(wsRoot, "data");
-    if (!fs.existsSync(dbDir)) {
-      fs.mkdirSync(dbDir, { recursive: true });
-    }
-    dbPath = path.join(dbDir, "settings.db");
-  }
-  const db = new Database(dbPath);
-
-  db.prepare(
-    `
-    CREATE TABLE IF NOT EXISTS provider_instances (
-      id TEXT PRIMARY KEY,
-      name TEXT NOT NULL,
-      providerName TEXT NOT NULL,
-      apiKey TEXT NOT NULL,
-      isActive INTEGER NOT NULL DEFAULT 0,
-      modelName TEXT,
-      type TEXT NOT NULL DEFAULT 'generative'
-    )
-  `,
-  ).run();
-
-  try {
-    db.prepare(
-      `ALTER TABLE provider_instances ADD COLUMN modelName TEXT`,
-    ).run();
-  } catch {
-    // ignore
-  }
-
-  try {
-    db.prepare(
-      `ALTER TABLE provider_instances ADD COLUMN type TEXT NOT NULL DEFAULT 'generative'`,
-    ).run();
-  } catch {
-    // ignore
-  }
-
-  try {
-    db.prepare(
-      `ALTER TABLE provider_instances ADD COLUMN maxContext INTEGER`,
-    ).run();
-  } catch {
-    // ignore
-  }
-
-  // Auto-bootstrap environment variables if DB contains 0 instances
-  try {
-    if (!hasBootstrapped) {
-      const totalCount = db
-        .prepare(`SELECT COUNT(*) as count FROM provider_instances`)
-        .get() as { count: number };
-      if (totalCount.count === 0) {
-        const googleKey = process.env.GOOGLE_API_KEY;
-        const openRouterKey = process.env.OPENROUTER_API_KEY;
-        let hasInsertedGenerative = false;
-
-        if (googleKey && googleKey.trim()) {
-          const id = "provider-default-google";
-          db.prepare(
-            `
-            INSERT INTO provider_instances (id, name, providerName, apiKey, isActive, modelName, type, maxContext)
-            VALUES (?, ?, ?, ?, ?, ?, ?, ?)
-          `,
-          ).run(
-            id,
-            "Gemini (Env)",
-            "google-genai",
-            googleKey.trim(),
-            1,
-            "gemini-2.5-flash",
-            "generative",
-            32768,
-          );
-          hasInsertedGenerative = true;
-
-          const embedId = "provider-default-google-embed";
-          db.prepare(
-            `
-            INSERT INTO provider_instances (id, name, providerName, apiKey, isActive, modelName, type, maxContext)
-            VALUES (?, ?, ?, ?, ?, ?, ?, ?)
-          `,
-          ).run(
-            embedId,
-            "Gemini Embed (Env)",
-            "google-genai",
-            googleKey.trim(),
-            1,
-            "gemini-embedding-001",
-            "embedding",
-            0,
-          );
-        }
-
-        if (openRouterKey && openRouterKey.trim()) {
-          const id = "provider-default-openrouter";
-          const isActive = hasInsertedGenerative ? 0 : 1;
-          db.prepare(
-            `
-            INSERT INTO provider_instances (id, name, providerName, apiKey, isActive, modelName, type, maxContext)
-            VALUES (?, ?, ?, ?, ?, ?, ?, ?)
-          `,
-          ).run(
-            id,
-            "OpenRouter (Env)",
-            "openrouter",
-            openRouterKey.trim(),
-            isActive,
-            "google/gemini-2.5-flash",
-            "generative",
-            32768,
-          );
-        }
-      }
-      hasBootstrapped = true;
-    }
-  } catch {
-    // ignore write lock issues or other DB errors during bootstrap
-  }
-
-  return db;
-}
+export { setDbPath as setDbPathOverride } from "./db.js";
 
 export class ProviderManager {
   static list(): ModelProviderInstance[] {
-    const db = getSettingsDb();
-    try {
-      const rows = db.prepare(`SELECT * FROM provider_instances`).all() as {
-        id: string;
-        name: string;
-        providerName: string;
-        apiKey: string;
-        isActive: number;
-        modelName?: string;
-        type: string;
-        maxContext?: number;
-      }[];
-      return rows.map((r) => ({
-        id: r.id,
-        name: r.name,
-        providerName: r.providerName,
-        apiKey: r.apiKey,
-        isActive: r.isActive === 1,
-        modelName: r.modelName || undefined,
-        type: (r.type as "generative" | "embedding") || "generative",
-        maxContext:
-          r.maxContext !== undefined && r.maxContext !== null
-            ? r.maxContext
-            : r.type === "embedding"
-              ? 0
-              : 32768,
-      }));
-    } finally {
-      db.close();
-    }
+    const db = getDb();
+    const rows = db
+      .prepare("SELECT * FROM provider_instances")
+      .all() as DbRow[];
+    return rows.map(mapRow);
   }
 
   static create(
@@ -200,94 +20,79 @@ export class ProviderManager {
     modelName?: string,
     type: "generative" | "embedding" = "generative",
     maxContext?: number,
+    endpointUrl?: string,
   ): ModelProviderInstance {
-    const db = getSettingsDb();
-    try {
-      const id = "provider-" + Date.now();
-      const activeCount = db
-        .prepare(
-          `SELECT COUNT(*) as count FROM provider_instances WHERE isActive = 1 AND type = ?`,
-        )
-        .get(type) as { count: number };
-      const isActive = activeCount.count === 0 ? 1 : 0;
+    const db = getDb();
+    const id = "provider-" + Date.now();
+    const activeCount = db
+      .prepare(
+        "SELECT COUNT(*) as count FROM provider_instances WHERE isActive = 1 AND type = ?",
+      )
+      .get(type) as { count: number };
+    const isActive = activeCount.count === 0 ? 1 : 0;
 
-      const actualMaxContext =
-        maxContext !== undefined
-          ? maxContext
-          : type === "generative"
-            ? 32768
-            : 0;
+    const actualMaxContext =
+      maxContext !== undefined ? maxContext : type === "generative" ? 32768 : 0;
 
-      db.prepare(
-        `
-        INSERT INTO provider_instances (id, name, providerName, apiKey, isActive, modelName, type, maxContext)
-        VALUES (?, ?, ?, ?, ?, ?, ?, ?)
-      `,
-      ).run(
-        id,
-        name,
-        providerName,
-        apiKey,
-        isActive,
-        modelName || null,
-        type,
-        actualMaxContext,
-      );
+    db.prepare(
+      `INSERT INTO provider_instances (id, name, providerName, apiKey, isActive, modelName, type, maxContext, endpointUrl)
+       VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)`,
+    ).run(
+      id,
+      name,
+      providerName,
+      apiKey,
+      isActive,
+      modelName || null,
+      type,
+      actualMaxContext,
+      endpointUrl || null,
+    );
 
-      return {
-        id,
-        name,
-        providerName,
-        apiKey,
-        isActive: isActive === 1,
-        modelName,
-        type,
-        maxContext: actualMaxContext,
-      };
-    } finally {
-      db.close();
-    }
+    return {
+      id,
+      name,
+      providerName,
+      apiKey,
+      isActive: isActive === 1,
+      modelName,
+      type,
+      maxContext: actualMaxContext,
+      endpointUrl,
+    };
   }
 
   static delete(id: string): void {
-    const db = getSettingsDb();
-    try {
-      const provider = db
-        .prepare(`SELECT isActive, type FROM provider_instances WHERE id = ?`)
-        .get(id) as { isActive: number; type: string } | undefined;
-      db.prepare(`DELETE FROM provider_instances WHERE id = ?`).run(id);
+    const db = getDb();
+    const provider = db
+      .prepare("SELECT isActive, type FROM provider_instances WHERE id = ?")
+      .get(id) as { isActive: number; type: string } | undefined;
+    db.prepare("DELETE FROM provider_instances WHERE id = ?").run(id);
 
-      if (provider && provider.isActive === 1) {
-        const next = db
-          .prepare(`SELECT id FROM provider_instances WHERE type = ? LIMIT 1`)
-          .get(provider.type) as { id: string } | undefined;
-        if (next) {
-          db.prepare(
-            `UPDATE provider_instances SET isActive = 1 WHERE id = ?`,
-          ).run(next.id);
-        }
+    if (provider && provider.isActive === 1) {
+      const next = db
+        .prepare("SELECT id FROM provider_instances WHERE type = ? LIMIT 1")
+        .get(provider.type) as { id: string } | undefined;
+      if (next) {
+        db.prepare(
+          "UPDATE provider_instances SET isActive = 1 WHERE id = ?",
+        ).run(next.id);
       }
-    } finally {
-      db.close();
     }
   }
 
   static setActive(id: string): void {
-    const db = getSettingsDb();
-    try {
-      const target = db
-        .prepare(`SELECT type FROM provider_instances WHERE id = ?`)
-        .get(id) as { type: string } | undefined;
-      if (target) {
-        db.prepare(
-          `UPDATE provider_instances SET isActive = 0 WHERE type = ?`,
-        ).run(target.type);
-        db.prepare(
-          `UPDATE provider_instances SET isActive = 1 WHERE id = ?`,
-        ).run(id);
-      }
-    } finally {
-      db.close();
+    const db = getDb();
+    const target = db
+      .prepare("SELECT type FROM provider_instances WHERE id = ?")
+      .get(id) as { type: string } | undefined;
+    if (target) {
+      db.prepare(
+        "UPDATE provider_instances SET isActive = 0 WHERE type = ?",
+      ).run(target.type);
+      db.prepare("UPDATE provider_instances SET isActive = 1 WHERE id = ?").run(
+        id,
+      );
     }
   }
 
@@ -299,329 +104,99 @@ export class ProviderManager {
     modelName?: string,
     type: "generative" | "embedding" = "generative",
     maxContext?: number,
+    endpointUrl?: string,
   ): void {
-    const db = getSettingsDb();
-    try {
-      const actualMaxContext =
-        maxContext !== undefined
-          ? maxContext
-          : type === "generative"
-            ? 32768
-            : 0;
-      if (apiKey && apiKey.trim()) {
-        db.prepare(
-          `
-          UPDATE provider_instances
-          SET name = ?, providerName = ?, apiKey = ?, modelName = ?, type = ?, maxContext = ?
-          WHERE id = ?
-        `,
-        ).run(
-          name,
-          providerName,
-          apiKey,
-          modelName || null,
-          type,
-          actualMaxContext,
-          id,
-        );
-      } else {
-        db.prepare(
-          `
-          UPDATE provider_instances
-          SET name = ?, providerName = ?, modelName = ?, type = ?, maxContext = ?
-          WHERE id = ?
-        `,
-        ).run(
-          name,
-          providerName,
-          modelName || null,
-          type,
-          actualMaxContext,
-          id,
-        );
-      }
-    } finally {
-      db.close();
+    const db = getDb();
+    const actualMaxContext =
+      maxContext !== undefined ? maxContext : type === "generative" ? 32768 : 0;
+
+    if (apiKey && apiKey.trim()) {
+      db.prepare(
+        `UPDATE provider_instances
+         SET name = ?, providerName = ?, apiKey = ?, modelName = ?, type = ?, maxContext = ?, endpointUrl = ?
+         WHERE id = ?`,
+      ).run(
+        name,
+        providerName,
+        apiKey,
+        modelName || null,
+        type,
+        actualMaxContext,
+        endpointUrl || null,
+        id,
+      );
+    } else {
+      db.prepare(
+        `UPDATE provider_instances
+         SET name = ?, providerName = ?, modelName = ?, type = ?, maxContext = ?, endpointUrl = ?
+         WHERE id = ?`,
+      ).run(
+        name,
+        providerName,
+        modelName || null,
+        type,
+        actualMaxContext,
+        endpointUrl || null,
+        id,
+      );
     }
   }
 
   static getActive(
     type: "generative" | "embedding" = "generative",
   ): ModelProviderInstance | null {
-    const db = getSettingsDb();
+    const db = getDb();
     try {
       const row = db
         .prepare(
-          `SELECT * FROM provider_instances WHERE isActive = 1 AND type = ?`,
+          "SELECT * FROM provider_instances WHERE isActive = 1 AND type = ?",
         )
-        .get(type) as
-        | {
-            id: string;
-            name: string;
-            providerName: string;
-            apiKey: string;
-            isActive: number;
-            modelName?: string;
-            type: string;
-            maxContext?: number;
-          }
-        | undefined;
+        .get(type) as DbRow | undefined;
 
-      if (!row) {
-        const totalCount = db
-          .prepare(`SELECT COUNT(*) as count FROM provider_instances`)
-          .get() as { count: number };
-        if (totalCount.count === 0) {
-          const googleKey = process.env.GOOGLE_API_KEY;
-          const openRouterKey = process.env.OPENROUTER_API_KEY;
-          let hasInsertedGenerative = false;
-
-          if (googleKey && googleKey.trim()) {
-            const id = "provider-default-google";
-            db.prepare(
-              `
-              INSERT INTO provider_instances (id, name, providerName, apiKey, isActive, modelName, type, maxContext)
-              VALUES (?, ?, ?, ?, ?, ?, ?, ?)
-            `,
-            ).run(
-              id,
-              "Gemini (Env)",
-              "google-genai",
-              googleKey.trim(),
-              1,
-              "gemini-2.5-flash",
-              "generative",
-              32768,
-            );
-            hasInsertedGenerative = true;
-
-            const embedId = "provider-default-google-embed";
-            db.prepare(
-              `
-              INSERT INTO provider_instances (id, name, providerName, apiKey, isActive, modelName, type, maxContext)
-              VALUES (?, ?, ?, ?, ?, ?, ?, ?)
-            `,
-            ).run(
-              embedId,
-              "Gemini Embed (Env)",
-              "google-genai",
-              googleKey.trim(),
-              1,
-              "gemini-embedding-001",
-              "embedding",
-              0,
-            );
-          }
-
-          if (openRouterKey && openRouterKey.trim()) {
-            const id = "provider-default-openrouter";
-            const isActive = hasInsertedGenerative ? 0 : 1;
-            db.prepare(
-              `
-              INSERT INTO provider_instances (id, name, providerName, apiKey, isActive, modelName, type, maxContext)
-              VALUES (?, ?, ?, ?, ?, ?, ?, ?)
-            `,
-            ).run(
-              id,
-              "OpenRouter (Env)",
-              "openrouter",
-              openRouterKey.trim(),
-              isActive,
-              "google/gemini-2.5-flash",
-              "generative",
-              32768,
-            );
-          }
-
-          const retryRow = db
-            .prepare(
-              `SELECT * FROM provider_instances WHERE isActive = 1 AND type = ?`,
-            )
-            .get(type) as
-            | {
-                id: string;
-                name: string;
-                providerName: string;
-                apiKey: string;
-                isActive: number;
-                modelName?: string;
-                type: string;
-                maxContext?: number;
-              }
-            | undefined;
-
-          if (retryRow) {
-            return {
-              id: retryRow.id,
-              name: retryRow.name,
-              providerName: retryRow.providerName,
-              apiKey: retryRow.apiKey,
-              isActive: true,
-              modelName: retryRow.modelName || undefined,
-              type: retryRow.type as "generative" | "embedding",
-              maxContext:
-                retryRow.maxContext !== undefined &&
-                retryRow.maxContext !== null
-                  ? retryRow.maxContext
-                  : retryRow.type === "embedding"
-                    ? 0
-                    : 32768,
-            };
-          }
-        }
-
-        // If there's no active row but some rows exist, return the first one as active, or update it
-        const firstRow = db
-          .prepare(`SELECT * FROM provider_instances WHERE type = ? LIMIT 1`)
-          .get(type) as
-          | {
-              id: string;
-              name: string;
-              providerName: string;
-              apiKey: string;
-              isActive: number;
-              modelName?: string;
-              type: string;
-              maxContext?: number;
-            }
-          | undefined;
-        if (firstRow) {
-          db.prepare(
-            `UPDATE provider_instances SET isActive = 1 WHERE id = ?`,
-          ).run(firstRow.id);
-          return {
-            id: firstRow.id,
-            name: firstRow.name,
-            providerName: firstRow.providerName,
-            apiKey: firstRow.apiKey,
-            isActive: true,
-            modelName: firstRow.modelName || undefined,
-            type: firstRow.type as "generative" | "embedding",
-            maxContext:
-              firstRow.maxContext !== undefined && firstRow.maxContext !== null
-                ? firstRow.maxContext
-                : firstRow.type === "embedding"
-                  ? 0
-                  : 32768,
-          };
-        }
-        return null;
+      if (row) {
+        return mapRow(row);
       }
 
-      return {
-        id: row.id,
-        name: row.name,
-        providerName: row.providerName,
-        apiKey: row.apiKey,
-        isActive: true,
-        modelName: row.modelName || undefined,
-        type: (row.type as "generative" | "embedding") || "generative",
-        maxContext:
-          row.maxContext !== undefined && row.maxContext !== null
-            ? row.maxContext
-            : row.type === "embedding"
-              ? 0
-              : 32768,
-      };
-    } catch {
-      const googleKey = process.env.GOOGLE_API_KEY;
-      if (type === "embedding") {
-        if (googleKey && googleKey.trim()) {
-          return {
-            id: "provider-default-env-embed-fallback",
-            name: "Gemini Embed (Env Fallback)",
-            providerName: "google-genai",
-            apiKey: googleKey.trim(),
-            isActive: true,
-            modelName: "gemini-embedding-001",
-            type: "embedding",
-            maxContext: 0,
-          };
-        }
-        return null;
+      const firstRow = db
+        .prepare("SELECT * FROM provider_instances WHERE type = ? LIMIT 1")
+        .get(type) as DbRow | undefined;
+
+      if (firstRow) {
+        db.prepare(
+          "UPDATE provider_instances SET isActive = 1 WHERE id = ?",
+        ).run(firstRow.id);
+        return mapRow(firstRow);
       }
 
-      // generative fallback
-      if (googleKey && googleKey.trim()) {
-        return {
-          id: "provider-default-env-fallback",
-          name: "Gemini (Env Fallback)",
-          providerName: "google-genai",
-          apiKey: googleKey.trim(),
-          isActive: true,
-          modelName: "gemini-2.5-flash",
-          type: "generative",
-          maxContext: 32768,
-        };
-      }
-      const openRouterKey = process.env.OPENROUTER_API_KEY;
-      if (openRouterKey && openRouterKey.trim()) {
-        return {
-          id: "provider-default-env-fallback",
-          name: "OpenRouter (Env Fallback)",
-          providerName: "openrouter",
-          apiKey: openRouterKey.trim(),
-          isActive: true,
-          modelName: "google/gemini-2.5-flash",
-          type: "generative",
-          maxContext: 32768,
-        };
-      }
       return null;
-    } finally {
-      db.close();
+    } catch {
+      return null;
     }
   }
 
   static getMappings(): Record<string, string> {
-    const db = getSettingsDb();
-    try {
-      db.prepare(
-        `
-        CREATE TABLE IF NOT EXISTS provider_mappings (
-          task TEXT PRIMARY KEY,
-          providerInstanceId TEXT NOT NULL
-        )
-      `,
-      ).run();
-      const rows = db.prepare(`SELECT * FROM provider_mappings`).all() as {
-        task: string;
-        providerInstanceId: string;
-      }[];
-      const mappings: Record<string, string> = {};
-      for (const row of rows) {
-        mappings[row.task] = row.providerInstanceId;
-      }
-      return mappings;
-    } finally {
-      db.close();
+    const db = getDb();
+    const rows = db.prepare("SELECT * FROM provider_mappings").all() as {
+      task: string;
+      providerInstanceId: string;
+    }[];
+    const mappings: Record<string, string> = {};
+    for (const row of rows) {
+      mappings[row.task] = row.providerInstanceId;
     }
+    return mappings;
   }
 
   static setMapping(task: string, providerInstanceId: string): void {
-    const db = getSettingsDb();
-    try {
+    const db = getDb();
+    if (!providerInstanceId) {
+      db.prepare("DELETE FROM provider_mappings WHERE task = ?").run(task);
+    } else {
       db.prepare(
-        `
-        CREATE TABLE IF NOT EXISTS provider_mappings (
-          task TEXT PRIMARY KEY,
-          providerInstanceId TEXT NOT NULL
-        )
-      `,
-      ).run();
-      if (!providerInstanceId) {
-        db.prepare(`DELETE FROM provider_mappings WHERE task = ?`).run(task);
-      } else {
-        db.prepare(
-          `
-          INSERT INTO provider_mappings (task, providerInstanceId)
-          VALUES (?, ?)
-          ON CONFLICT(task) DO UPDATE SET providerInstanceId = excluded.providerInstanceId
-        `,
-        ).run(task, providerInstanceId);
-      }
-    } finally {
-      db.close();
+        `INSERT INTO provider_mappings (task, providerInstanceId)
+         VALUES (?, ?)
+         ON CONFLICT(task) DO UPDATE SET providerInstanceId = excluded.providerInstanceId`,
+      ).run(task, providerInstanceId);
     }
   }
 }

packages/llm/src/providers/anthropic.ts

@@ -0,0 +1,116 @@
+import { ChatAnthropic } from "@langchain/anthropic";
+import { ILLMProvider } from "../llm.js";
+import type { ModelProviderInstance } from "../llm.js";
+import { BaseLLMProvider, resolveCredentials } from "../base-provider.js";
+import { registerProvider, registerGenerative } from "../registry.js";
+import { fetchWithTimeout, type ModelInfo } from "../model-lister.js";
+
+async function fetchAnthropicModels(apiKey: string): Promise<ModelInfo[]> {
+  const models: ModelInfo[] = [];
+  let afterId: string | undefined;
+
+  do {
+    const url = new URL("https://api.anthropic.com/v1/models");
+    url.searchParams.set("limit", "1000");
+    if (afterId) {
+      url.searchParams.set("after_id", afterId);
+    }
+
+    const res = await fetchWithTimeout(url.toString(), {
+      headers: {
+        "x-api-key": apiKey,
+        "anthropic-version": "2023-06-01",
+        Accept: "application/json",
+      },
+    });
+    if (!res.ok) return models;
+
+    const json = (await res.json()) as {
+      data?: { id: string; display_name?: string }[];
+      has_more?: boolean;
+      last_id?: string;
+    };
+
+    for (const m of json.data ?? []) {
+      models.push({ id: m.id, name: m.display_name || m.id });
+    }
+
+    afterId = json.has_more ? json.last_id : undefined;
+  } while (afterId);
+
+  return models;
+}
+
+export class AnthropicProvider extends BaseLLMProvider {
+  static {
+    registerProvider({
+      id: "anthropic",
+      displayName: "Anthropic Claude",
+      description: "Official Claude integration using @langchain/anthropic SDK",
+      envVar: "ANTHROPIC_API_KEY",
+      capabilities: { generative: true, embedding: false },
+      defaultModel: "claude-3-5-sonnet-latest",
+      defaultMaxContext: 200000,
+      fallbackPriority: 2,
+      listModels: fetchAnthropicModels,
+    });
+    registerGenerative(
+      "anthropic",
+      (inst: ModelProviderInstance) =>
+        new AnthropicProvider(
+          inst.apiKey,
+          inst.modelName,
+          inst.name,
+          inst.maxContext,
+        ),
+    );
+  }
+
+  static create(inst: ModelProviderInstance): ILLMProvider {
+    return new AnthropicProvider(
+      inst.apiKey,
+      inst.modelName,
+      inst.name,
+      inst.maxContext,
+    );
+  }
+
+  providerName = "Anthropic";
+  protected readonly model: ChatAnthropic;
+  protected modelNameUsed: string;
+  protected providerInstanceName?: string;
+  protected maxContextUsed?: number;
+  protected defaultMaxContext = 200000;
+
+  constructor(
+    apiKey?: string,
+    modelName?: string,
+    providerInstanceName?: string,
+    maxContext?: number,
+  ) {
+    super();
+    const {
+      key,
+      model,
+      providerInstanceName: resolvedName,
+      maxContext: resolvedMax,
+    } = resolveCredentials({
+      explicitKey: apiKey,
+      explicitModel: modelName,
+      explicitProviderInstanceName: providerInstanceName,
+      explicitMaxContext: maxContext,
+      providerId: "anthropic",
+      envVarName: "ANTHROPIC_API_KEY",
+      type: "generative",
+    });
+    if (!key) {
+      throw new Error(
+        "ANTHROPIC_API_KEY is required to initialize AnthropicProvider",
+      );
+    }
+    this.providerInstanceName = resolvedName;
+    this.maxContextUsed = resolvedMax;
+    this.modelNameUsed = model || "claude-3-5-sonnet-latest";
+    this.model = new ChatAnthropic({ apiKey: key, model: this.modelNameUsed });
+  }
+}

packages/llm/src/providers/deepseek.ts

@@ -0,0 +1,82 @@
+import { ChatDeepSeek } from "@langchain/deepseek";
+import { ILLMProvider } from "../llm.js";
+import type { ModelProviderInstance } from "../llm.js";
+import { BaseLLMProvider, resolveCredentials } from "../base-provider.js";
+import { registerProvider, registerGenerative } from "../registry.js";
+import { fetchOpenAICompatibleModels } from "../model-lister.js";
+
+export class DeepSeekProvider extends BaseLLMProvider {
+  static {
+    registerProvider({
+      id: "deepseek",
+      displayName: "DeepSeek",
+      description:
+        "Official DeepSeek integration using @langchain/deepseek SDK",
+      envVar: "DEEPSEEK_API_KEY",
+      capabilities: { generative: true, embedding: false },
+      defaultModel: "deepseek-chat",
+      defaultMaxContext: 64000,
+      fallbackPriority: 4,
+      listModels: (apiKey) =>
+        fetchOpenAICompatibleModels("https://api.deepseek.com", apiKey),
+    });
+    registerGenerative(
+      "deepseek",
+      (inst: ModelProviderInstance) =>
+        new DeepSeekProvider(
+          inst.apiKey,
+          inst.modelName,
+          inst.name,
+          inst.maxContext,
+        ),
+    );
+  }
+
+  static create(inst: ModelProviderInstance): ILLMProvider {
+    return new DeepSeekProvider(
+      inst.apiKey,
+      inst.modelName,
+      inst.name,
+      inst.maxContext,
+    );
+  }
+
+  providerName = "DeepSeek";
+  protected readonly model: ChatDeepSeek;
+  protected modelNameUsed: string;
+  protected providerInstanceName?: string;
+  protected maxContextUsed?: number;
+  protected defaultMaxContext = 64000;
+
+  constructor(
+    apiKey?: string,
+    modelName?: string,
+    providerInstanceName?: string,
+    maxContext?: number,
+  ) {
+    super();
+    const {
+      key,
+      model,
+      providerInstanceName: resolvedName,
+      maxContext: resolvedMax,
+    } = resolveCredentials({
+      explicitKey: apiKey,
+      explicitModel: modelName,
+      explicitProviderInstanceName: providerInstanceName,
+      explicitMaxContext: maxContext,
+      providerId: "deepseek",
+      envVarName: "DEEPSEEK_API_KEY",
+      type: "generative",
+    });
+    if (!key) {
+      throw new Error(
+        "DEEPSEEK_API_KEY is required to initialize DeepSeekProvider",
+      );
+    }
+    this.providerInstanceName = resolvedName;
+    this.maxContextUsed = resolvedMax;
+    this.modelNameUsed = model || "deepseek-chat";
+    this.model = new ChatDeepSeek({ apiKey: key, model: this.modelNameUsed });
+  }
+}

packages/llm/src/providers/google-genai.ts

@@ -1,31 +1,96 @@
-import { z } from "zod";
 import {
   ChatGoogleGenerativeAI,
   GoogleGenerativeAIEmbeddings,
 } from "@langchain/google-genai";
-import {
+import type {
   ILLMProvider,
-  LLMRequest,
-  LLMResponse,
-  LLMCallRecord,
   IEmbeddingProvider,
+  ModelProviderInstance,
 } from "../llm.js";
-import { llmConfig } from "../config.js";
+import {
+  registerProvider,
+  registerGenerative,
+  registerEmbedding,
+} from "../registry.js";
+import { fetchWithTimeout, type ModelInfo } from "../model-lister.js";
+import { BaseLLMProvider, resolveCredentials } from "../base-provider.js";
+import { getLlmConfig } from "../config.js";
 import { ProviderManager } from "../provider-manager.js";
 
-export class GeminiProvider implements ILLMProvider {
-  static readonly providerId = "google-genai";
-  static readonly displayName = "Google Gemini";
-  static readonly description =
-    "Official Gemini integration using Google Gen AI SDK";
-  static readonly defaultModel = "gemini-2.5-flash";
+async function fetchGeminiModels(apiKey: string): Promise<ModelInfo[]> {
+  const models: ModelInfo[] = [];
+  let pageToken: string | undefined;
+
+  do {
+    const url = new URL(
+      "https://generativelanguage.googleapis.com/v1beta/models",
+    );
+    url.searchParams.set("key", apiKey);
+    url.searchParams.set("pageSize", "100");
+    if (pageToken) {
+      url.searchParams.set("pageToken", pageToken);
+    }
+
+    const res = await fetchWithTimeout(url.toString());
+    if (!res.ok) return models;
+
+    const json = (await res.json()) as {
+      models?: { name: string; displayName?: string }[];
+      nextPageToken?: string;
+    };
+
+    for (const m of json.models ?? []) {
+      const id = m.name.replace(/^models\//, "");
+      models.push({ id, name: m.displayName || id });
+    }
+
+    pageToken = json.nextPageToken;
+  } while (pageToken);
+
+  return models;
+}
+
+export class GeminiProvider extends BaseLLMProvider {
+  static {
+    registerProvider({
+      id: "google-genai",
+      displayName: "Google Gemini",
+      description: "Official Gemini integration using Google Gen AI SDK",
+      envVar: "GOOGLE_API_KEY",
+      capabilities: { generative: true, embedding: true },
+      defaultModel: "gemini-2.5-flash",
+      defaultEmbeddingModel: "gemini-embedding-001",
+      defaultMaxContext: 32768,
+      fallbackPriority: 0,
+      listModels: fetchGeminiModels,
+    });
+    registerGenerative(
+      "google-genai",
+      (inst: ModelProviderInstance) =>
+        new GeminiProvider(
+          inst.apiKey,
+          inst.modelName,
+          inst.name,
+          inst.maxContext,
+        ),
+    );
+  }
 
   providerName = "Gemini";
-  private model: ChatGoogleGenerativeAI;
-  private modelNameUsed: string;
-  private providerInstanceName?: string;
-  private maxContextUsed?: number;
-  lastCalls: LLMCallRecord[] = [];
+  protected readonly model: ChatGoogleGenerativeAI;
+  protected modelNameUsed: string;
+  protected providerInstanceName?: string;
+  protected maxContextUsed?: number;
+  protected readonly defaultMaxContext = 32768;
+
+  static create(inst: ModelProviderInstance): ILLMProvider {
+    return new GeminiProvider(
+      inst.apiKey,
+      inst.modelName,
+      inst.name,
+      inst.maxContext,
+    );
+  }
 
   constructor(
     apiKey?: string,
@@ -33,113 +98,66 @@ export class GeminiProvider implements ILLMProvider {
     providerInstanceName?: string,
     maxContext?: number,
   ) {
-    let key = apiKey;
-    let model = modelName;
-    this.providerInstanceName = providerInstanceName;
-    this.maxContextUsed = maxContext;
-
-    if (!key) {
-      const active = ProviderManager.getActive("generative");
-      if (active && active.providerName === GeminiProvider.providerId) {
-        key = active.apiKey;
-        if (!model) {
-          model = active.modelName;
-        }
-        if (!this.providerInstanceName) {
-          this.providerInstanceName = active.name;
-        }
-        if (this.maxContextUsed === undefined) {
-          this.maxContextUsed = active.maxContext;
-        }
-      }
-    }
-
-    if (!key) {
-      key = llmConfig.GOOGLE_API_KEY;
-      if (!this.providerInstanceName && key) {
-        this.providerInstanceName = "Environment Variable";
-      }
-    }
-
+    super();
+    const {
+      key,
+      model,
+      providerInstanceName: pn,
+      maxContext: mc,
+    } = resolveCredentials({
+      explicitKey: apiKey,
+      explicitModel: modelName,
+      explicitProviderInstanceName: providerInstanceName,
+      explicitMaxContext: maxContext,
+      providerId: "google-genai",
+      envVarName: "GOOGLE_API_KEY",
+      type: "generative",
+    });
     if (!key) {
       throw new Error(
         "GOOGLE_API_KEY is required to initialize GeminiProvider",
       );
     }
-
+    this.providerInstanceName = pn;
+    this.maxContextUsed = mc;
     this.modelNameUsed = model || "gemini-2.5-flash";
     this.model = new ChatGoogleGenerativeAI({
       apiKey: key,
       model: this.modelNameUsed,
     });
   }
-
-  async generateStructuredResponse<T extends z.ZodTypeAny>(
-    request: LLMRequest<T>,
-  ): Promise<LLMResponse<z.infer<T>>> {
-    const structuredModel = this.model.withStructuredOutput(request.schema, {
-      includeRaw: true,
-    });
-    const result = (await structuredModel.invoke([
-      { role: "system", content: request.systemPrompt },
-      { role: "user", content: request.userContext },
-    ])) as unknown as {
-      parsed?: z.infer<T>;
-      raw?: {
-        usage_metadata?: {
-          input_tokens?: number;
-          output_tokens?: number;
-          total_tokens?: number;
-        };
-      };
-    };
-
-    const parsed = result?.parsed;
-    const raw = result?.raw;
-
-    const usage = {
-      inputTokens: raw?.usage_metadata?.input_tokens || 0,
-      outputTokens: raw?.usage_metadata?.output_tokens || 0,
-      totalTokens: raw?.usage_metadata?.total_tokens || 0,
-      modelName: this.modelNameUsed,
-      providerInstanceName: this.providerInstanceName || "Default",
-      maxContext:
-        this.maxContextUsed !== undefined ? this.maxContextUsed : 32768,
-    };
-
-    this.lastCalls.push({
-      systemPrompt: request.systemPrompt,
-      userContext: request.userContext,
-      usage,
-    });
-
-    return { success: true, data: parsed, usage };
-  }
 }
 
 export class GeminiEmbeddingProvider implements IEmbeddingProvider {
-  static readonly providerId = "google-genai";
-  static readonly displayName = "Google Gemini Embeddings";
+  static {
+    registerEmbedding(
+      "google-genai",
+      (inst: ModelProviderInstance) =>
+        new GeminiEmbeddingProvider(inst.apiKey, inst.modelName),
+    );
+  }
 
   providerName = "Gemini";
   private model: GoogleGenerativeAIEmbeddings;
 
+  static create(inst: ModelProviderInstance): IEmbeddingProvider {
+    return new GeminiEmbeddingProvider(inst.apiKey, inst.modelName);
+  }
+
   constructor(apiKey?: string, modelName?: string) {
     let key = apiKey;
     let model = modelName;
 
     if (!key) {
       const active = ProviderManager.getActive("embedding");
-      if (active) {
+      if (active && active.providerName === "google-genai") {
         key = active.apiKey;
-        if (!model) {
-          model = active.modelName;
-        }
+        if (!model) model = active.modelName;
       }
     }
 
     if (!key) {
-      key = llmConfig.GOOGLE_API_KEY;
+      key = getLlmConfig().GOOGLE_API_KEY;
     }
 
     if (!key) {

packages/llm/src/providers/groq.ts

@@ -0,0 +1,79 @@
+import { ChatGroq } from "@langchain/groq";
+import { ILLMProvider } from "../llm.js";
+import type { ModelProviderInstance } from "../llm.js";
+import { BaseLLMProvider, resolveCredentials } from "../base-provider.js";
+import { registerProvider, registerGenerative } from "../registry.js";
+import { fetchOpenAICompatibleModels } from "../model-lister.js";
+
+export class GroqProvider extends BaseLLMProvider {
+  static {
+    registerProvider({
+      id: "groq",
+      displayName: "Groq",
+      description: "Official Groq integration using @langchain/groq SDK",
+      envVar: "GROQ_API_KEY",
+      capabilities: { generative: true, embedding: false },
+      defaultModel: "llama-3.3-70b-versatile",
+      defaultMaxContext: 8192,
+      fallbackPriority: 3,
+      listModels: (apiKey) =>
+        fetchOpenAICompatibleModels("https://api.groq.com/openai/v1", apiKey),
+    });
+    registerGenerative(
+      "groq",
+      (inst: ModelProviderInstance) =>
+        new GroqProvider(
+          inst.apiKey,
+          inst.modelName,
+          inst.name,
+          inst.maxContext,
+        ),
+    );
+  }
+
+  static create(inst: ModelProviderInstance): ILLMProvider {
+    return new GroqProvider(
+      inst.apiKey,
+      inst.modelName,
+      inst.name,
+      inst.maxContext,
+    );
+  }
+
+  providerName = "Groq";
+  protected readonly model: ChatGroq;
+  protected modelNameUsed: string;
+  protected providerInstanceName?: string;
+  protected maxContextUsed?: number;
+  protected defaultMaxContext = 8192;
+
+  constructor(
+    apiKey?: string,
+    modelName?: string,
+    providerInstanceName?: string,
+    maxContext?: number,
+  ) {
+    super();
+    const {
+      key,
+      model,
+      providerInstanceName: resolvedName,
+      maxContext: resolvedMax,
+    } = resolveCredentials({
+      explicitKey: apiKey,
+      explicitModel: modelName,
+      explicitProviderInstanceName: providerInstanceName,
+      explicitMaxContext: maxContext,
+      providerId: "groq",
+      envVarName: "GROQ_API_KEY",
+      type: "generative",
+    });
+    if (!key) {
+      throw new Error("GROQ_API_KEY is required to initialize GroqProvider");
+    }
+    this.providerInstanceName = resolvedName;
+    this.maxContextUsed = resolvedMax;
+    this.modelNameUsed = model || "llama-3.3-70b-versatile";
+    this.model = new ChatGroq({ apiKey: key, model: this.modelNameUsed });
+  }
+}

packages/llm/src/providers/mock.ts

@@ -6,13 +6,33 @@ import {
   LLMCallRecord,
   IEmbeddingProvider,
 } from "../llm.js";
+import type { ModelProviderInstance } from "../llm.js";
+import {
+  registerProvider,
+  registerGenerative,
+  registerEmbedding,
+} from "../registry.js";
 
 export class MockLLMProvider implements ILLMProvider {
-  static readonly providerId = "mock";
-  static readonly displayName = "Mock LLM Provider";
-  static readonly description =
-    "Stateless mock provider for testing and offline development";
-  static readonly defaultModel = "mock";
+  static {
+    registerProvider({
+      id: "mock",
+      displayName: "Mock LLM Provider",
+      description:
+        "Stateless mock provider for testing and offline development",
+      capabilities: { generative: true, embedding: true },
+      defaultModel: "mock",
+      defaultEmbeddingModel: "mock-embeddings",
+      defaultMaxContext: 0,
+      fallbackPriority: 1000,
+      listModels: () => Promise.resolve([{ id: "mock", name: "Mock Model" }]),
+    });
+    registerGenerative("mock", () => new MockLLMProvider([]));
+  }
+
+  static create(inst: ModelProviderInstance): ILLMProvider {
+    return new MockLLMProvider([]);
+  }
 
   providerName = "mock";
   private callCount = 0;
@@ -46,7 +66,17 @@ export class MockLLMProvider implements ILLMProvider {
 }
 
 export class MockEmbeddingProvider implements IEmbeddingProvider {
-  static readonly providerId = "mock";
+  static {
+    registerEmbedding(
+      "mock",
+      (inst: ModelProviderInstance) =>
+        new MockEmbeddingProvider(inst.modelName),
+    );
+  }
+
+  static create(inst: ModelProviderInstance): IEmbeddingProvider {
+    return new MockEmbeddingProvider(inst.modelName);
+  }
 
   providerName = "mock";
 

packages/llm/src/providers/ollama.ts

@@ -0,0 +1,172 @@
+import { ChatOllama, OllamaEmbeddings } from "@langchain/ollama";
+import { ILLMProvider, IEmbeddingProvider } from "../llm.js";
+import type { ModelProviderInstance } from "../llm.js";
+import { ProviderManager } from "../provider-manager.js";
+import { BaseLLMProvider } from "../base-provider.js";
+import {
+  registerProvider,
+  registerGenerative,
+  registerEmbedding,
+} from "../registry.js";
+import { fetchWithTimeout, type ModelInfo } from "../model-lister.js";
+
+async function fetchOllamaModels(endpointUrl: string): Promise<ModelInfo[]> {
+  const base = endpointUrl.replace(/\/$/, "");
+  const res = await fetchWithTimeout(`${base}/api/tags`);
+  if (!res.ok) return [];
+
+  const json = (await res.json()) as {
+    models?: { name: string; model?: string }[];
+  };
+
+  return (json.models ?? []).map((m) => ({
+    id: m.name,
+    name: m.name,
+  }));
+}
+
+export class OllamaProvider extends BaseLLMProvider {
+  static {
+    registerProvider({
+      id: "ollama",
+      displayName: "Ollama",
+      description:
+        "Local model runner supporting open-source LLMs via the Ollama server",
+      capabilities: { generative: true, embedding: true },
+      defaultModel: "llama3.1",
+      defaultEmbeddingModel: "nomic-embed-text",
+      defaultMaxContext: 32768,
+      fallbackPriority: 100,
+      listModels: (_apiKey, endpointUrl) =>
+        fetchOllamaModels(endpointUrl || "http://localhost:11434"),
+    });
+    registerGenerative(
+      "ollama",
+      (inst: ModelProviderInstance) =>
+        new OllamaProvider(
+          inst.endpointUrl,
+          inst.modelName,
+          inst.name,
+          inst.maxContext,
+        ),
+    );
+  }
+
+  static create(inst: ModelProviderInstance): ILLMProvider {
+    return new OllamaProvider(
+      inst.endpointUrl,
+      inst.modelName,
+      inst.name,
+      inst.maxContext,
+    );
+  }
+
+  providerName = "Ollama";
+  protected readonly model: ChatOllama;
+  protected modelNameUsed: string;
+  protected providerInstanceName?: string;
+  protected maxContextUsed?: number;
+  protected defaultMaxContext = 32768;
+
+  /**
+   * Creates an OllamaProvider.
+   *
+   * Resolution order for configuration:
+   *  1. Explicit constructor arguments
+   *  2. Active "generative" instance in ProviderManager whose providerName === "ollama"
+   *  3. Defaults (baseUrl: http://localhost:11434, model: llama3.1)
+   *
+   * No API key is required for Ollama. The `endpointUrl` in
+   * ModelProviderInstance stores the Ollama server base URL
+   * (e.g. "http://localhost:11434").
+   */
+  constructor(
+    baseUrl?: string,
+    modelName?: string,
+    providerInstanceName?: string,
+    maxContext?: number,
+  ) {
+    super();
+    let url = baseUrl;
+    let model = modelName;
+    this.providerInstanceName = providerInstanceName;
+    this.maxContextUsed = maxContext;
+
+    if (!url || !model) {
+      const active = ProviderManager.getActive("generative");
+      if (active && active.providerName === "ollama") {
+        if (!url) {
+          url = active.endpointUrl;
+        }
+        if (!model) {
+          model = active.modelName;
+        }
+        if (!this.providerInstanceName) {
+          this.providerInstanceName = active.name;
+        }
+        if (this.maxContextUsed === undefined) {
+          this.maxContextUsed = active.maxContext;
+        }
+      }
+    }
+
+    this.modelNameUsed = model || "llama3.1";
+    this.model = new ChatOllama({
+      baseUrl: url || "http://localhost:11434",
+      model: this.modelNameUsed,
+    });
+  }
+}
+
+export class OllamaEmbeddingProvider implements IEmbeddingProvider {
+  static {
+    registerEmbedding(
+      "ollama",
+      (inst: ModelProviderInstance) =>
+        new OllamaEmbeddingProvider(inst.endpointUrl, inst.modelName),
+    );
+  }
+
+  static create(inst: ModelProviderInstance): IEmbeddingProvider {
+    return new OllamaEmbeddingProvider(inst.endpointUrl, inst.modelName);
+  }
+
+  providerName = "Ollama";
+  private model: OllamaEmbeddings;
+
+  /**
+   * Creates an OllamaEmbeddingProvider.
+   *
+   * Resolution order:
+   *  1. Explicit constructor arguments
+   *  2. Active "embedding" instance in ProviderManager
+   *  3. Defaults (baseUrl: http://localhost:11434, model: nomic-embed-text)
+   *
+   * The `endpointUrl` field in ModelProviderInstance stores the base URL.
+   */
+  constructor(baseUrl?: string, modelName?: string) {
+    let url = baseUrl;
+    let model = modelName;
+
+    if (!url || !model) {
+      const active = ProviderManager.getActive("embedding");
+      if (active && active.providerName === "ollama") {
+        if (!url) {
+          url = active.endpointUrl;
+        }
+        if (!model) {
+          model = active.modelName;
+        }
+      }
+    }
+
+    this.model = new OllamaEmbeddings({
+      baseUrl: url || "http://localhost:11434",
+      model: model || "nomic-embed-text",
+    });
+  }
+
+  async embed(text: string): Promise<number[]> {
+    return this.model.embedQuery(text);
+  }
+}

packages/llm/src/providers/openai.ts

@@ -0,0 +1,139 @@
+import { ChatOpenAI, OpenAIEmbeddings } from "@langchain/openai";
+import { ILLMProvider, IEmbeddingProvider } from "../llm.js";
+import type { ModelProviderInstance } from "../llm.js";
+import { getLlmConfig } from "../config.js";
+import { ProviderManager } from "../provider-manager.js";
+import { BaseLLMProvider, resolveCredentials } from "../base-provider.js";
+import {
+  registerProvider,
+  registerGenerative,
+  registerEmbedding,
+} from "../registry.js";
+import { fetchOpenAICompatibleModels } from "../model-lister.js";
+
+export class OpenAIProvider extends BaseLLMProvider {
+  static {
+    registerProvider({
+      id: "openai",
+      displayName: "OpenAI",
+      description: "Official OpenAI integration using @langchain/openai SDK",
+      envVar: "OPENAI_API_KEY",
+      capabilities: { generative: true, embedding: true },
+      defaultModel: "gpt-4o-mini",
+      defaultEmbeddingModel: "text-embedding-3-small",
+      defaultMaxContext: 128000,
+      fallbackPriority: 1,
+      listModels: (apiKey) =>
+        fetchOpenAICompatibleModels("https://api.openai.com/v1", apiKey),
+    });
+    registerGenerative(
+      "openai",
+      (inst: ModelProviderInstance) =>
+        new OpenAIProvider(
+          inst.apiKey,
+          inst.modelName,
+          inst.name,
+          inst.maxContext,
+        ),
+    );
+  }
+
+  static create(inst: ModelProviderInstance): ILLMProvider {
+    return new OpenAIProvider(
+      inst.apiKey,
+      inst.modelName,
+      inst.name,
+      inst.maxContext,
+    );
+  }
+
+  providerName = "OpenAI";
+  protected readonly model: ChatOpenAI;
+  protected modelNameUsed: string;
+  protected providerInstanceName?: string;
+  protected maxContextUsed?: number;
+  protected defaultMaxContext = 128000;
+
+  constructor(
+    apiKey?: string,
+    modelName?: string,
+    providerInstanceName?: string,
+    maxContext?: number,
+  ) {
+    super();
+    const {
+      key,
+      model,
+      providerInstanceName: resolvedName,
+      maxContext: resolvedMax,
+    } = resolveCredentials({
+      explicitKey: apiKey,
+      explicitModel: modelName,
+      explicitProviderInstanceName: providerInstanceName,
+      explicitMaxContext: maxContext,
+      providerId: "openai",
+      envVarName: "OPENAI_API_KEY",
+      type: "generative",
+    });
+    if (!key) {
+      throw new Error(
+        "OPENAI_API_KEY is required to initialize OpenAIProvider",
+      );
+    }
+    this.providerInstanceName = resolvedName;
+    this.maxContextUsed = resolvedMax;
+    this.modelNameUsed = model || "gpt-4o-mini";
+    this.model = new ChatOpenAI({ apiKey: key, model: this.modelNameUsed });
+  }
+}
+
+export class OpenAIEmbeddingProvider implements IEmbeddingProvider {
+  static {
+    registerEmbedding(
+      "openai",
+      (inst: ModelProviderInstance) =>
+        new OpenAIEmbeddingProvider(inst.apiKey, inst.modelName),
+    );
+  }
+
+  static create(inst: ModelProviderInstance): IEmbeddingProvider {
+    return new OpenAIEmbeddingProvider(inst.apiKey, inst.modelName);
+  }
+
+  providerName = "OpenAI";
+  private model: OpenAIEmbeddings;
+
+  constructor(apiKey?: string, modelName?: string) {
+    let key = apiKey;
+    let model = modelName;
+
+    if (!key) {
+      const active = ProviderManager.getActive("embedding");
+      if (active && active.providerName === "openai") {
+        key = active.apiKey;
+        if (!model) {
+          model = active.modelName;
+        }
+      }
+    }
+
+    if (!key) {
+      key = getLlmConfig().OPENAI_API_KEY;
+    }
+
+    if (!key) {
+      throw new Error(
+        "OPENAI_API_KEY is required to initialize OpenAIEmbeddingProvider",
+      );
+    }
+
+    this.model = new OpenAIEmbeddings({
+      apiKey: key,
+      model: model || "text-embedding-3-small",
+    });
+  }
+
+  async embed(text: string): Promise<number[]> {
+    return this.model.embedQuery(text);
+  }
+}

packages/llm/src/providers/openrouter.ts

@@ -1,27 +1,77 @@
-import { z } from "zod";
 import { ChatOpenRouter } from "@langchain/openrouter";
-import {
-  ILLMProvider,
-  LLMRequest,
-  LLMResponse,
-  LLMCallRecord,
-} from "../llm.js";
-import { llmConfig } from "../config.js";
-import { ProviderManager } from "../provider-manager.js";
+import { ILLMProvider } from "../llm.js";
+import type { ModelProviderInstance } from "../llm.js";
+import { BaseLLMProvider, resolveCredentials } from "../base-provider.js";
+import { registerProvider, registerGenerative } from "../registry.js";
+import { fetchWithTimeout, type ModelInfo } from "../model-lister.js";
 
-export class OpenRouterProvider implements ILLMProvider {
-  static readonly providerId = "openrouter";
-  static readonly displayName = "OpenRouter";
-  static readonly description =
-    "Multi-model router supporting Anthropic, OpenAI, DeepSeek, and local models";
-  static readonly defaultModel = "google/gemini-2.5-flash";
+async function fetchOpenRouterModels(apiKey: string): Promise<ModelInfo[]> {
+  const res = await fetchWithTimeout(
+    "https://openrouter.ai/api/v1/models",
+    apiKey
+      ? {
+          headers: {
+            Authorization: `Bearer ${apiKey}`,
+            Accept: "application/json",
+          },
+        }
+      : { headers: { Accept: "application/json" } },
+  );
+  if (!res.ok) return [];
+
+  const json = (await res.json()) as {
+    data?: { id: string; name?: string; owned_by?: string }[];
+  };
+
+  return (json.data ?? []).map((m) => ({
+    id: m.id,
+    name: m.name || m.id,
+    ownedBy: m.owned_by,
+  }));
+}
+
+export class OpenRouterProvider extends BaseLLMProvider {
+  static {
+    registerProvider({
+      id: "openrouter",
+      displayName: "OpenRouter",
+      description:
+        "Multi-model router supporting Anthropic, OpenAI, DeepSeek, and local models",
+      envVar: "OPENROUTER_API_KEY",
+      capabilities: { generative: true, embedding: false },
+      defaultModel: "google/gemini-2.5-flash",
+      defaultEmbeddingModel: "openai/text-embedding-3-small",
+      defaultMaxContext: 32768,
+      fallbackPriority: 5,
+      listModels: fetchOpenRouterModels,
+    });
+    registerGenerative(
+      "openrouter",
+      (inst: ModelProviderInstance) =>
+        new OpenRouterProvider(
+          inst.apiKey,
+          inst.modelName,
+          inst.name,
+          inst.maxContext,
+        ),
+    );
+  }
+
+  static create(inst: ModelProviderInstance): ILLMProvider {
+    return new OpenRouterProvider(
+      inst.apiKey,
+      inst.modelName,
+      inst.name,
+      inst.maxContext,
+    );
+  }
 
   providerName = "OpenRouter";
-  private model: ChatOpenRouter;
-  private modelNameUsed: string;
-  private providerInstanceName?: string;
-  private maxContextUsed?: number;
-  lastCalls: LLMCallRecord[] = [];
+  protected readonly model: ChatOpenRouter;
+  protected modelNameUsed: string;
+  protected providerInstanceName?: string;
+  protected maxContextUsed?: number;
+  protected defaultMaxContext = 32768;
 
   constructor(
     apiKey?: string,
@@ -29,86 +79,29 @@ export class OpenRouterProvider implements ILLMProvider {
     providerInstanceName?: string,
     maxContext?: number,
   ) {
-    let key = apiKey;
-    let model = modelName;
-    this.providerInstanceName = providerInstanceName;
-    this.maxContextUsed = maxContext;
-
-    if (!key) {
-      const active = ProviderManager.getActive("generative");
-      if (active && active.providerName === OpenRouterProvider.providerId) {
-        key = active.apiKey;
-        if (!model) {
-          model = active.modelName;
-        }
-        if (!this.providerInstanceName) {
-          this.providerInstanceName = active.name;
-        }
-        if (this.maxContextUsed === undefined) {
-          this.maxContextUsed = active.maxContext;
-        }
-      }
-    }
-
-    if (!key) {
-      key = llmConfig.OPENROUTER_API_KEY;
-      if (!this.providerInstanceName && key) {
-        this.providerInstanceName = "Environment Variable";
-      }
-    }
-
+    super();
+    const {
+      key,
+      model,
+      providerInstanceName: resolvedName,
+      maxContext: resolvedMax,
+    } = resolveCredentials({
+      explicitKey: apiKey,
+      explicitModel: modelName,
+      explicitProviderInstanceName: providerInstanceName,
+      explicitMaxContext: maxContext,
+      providerId: "openrouter",
+      envVarName: "OPENROUTER_API_KEY",
+      type: "generative",
+    });
     if (!key) {
       throw new Error(
         "OPENROUTER_API_KEY is required to initialize OpenRouterProvider",
       );
     }
-
+    this.providerInstanceName = resolvedName;
+    this.maxContextUsed = resolvedMax;
     this.modelNameUsed = model || "google/gemini-2.5-flash";
-    this.model = new ChatOpenRouter({
-      apiKey: key,
-      model: this.modelNameUsed,
-    });
-  }
-
-  async generateStructuredResponse<T extends z.ZodTypeAny>(
-    request: LLMRequest<T>,
-  ): Promise<LLMResponse<z.infer<T>>> {
-    const structuredModel = this.model.withStructuredOutput(request.schema, {
-      includeRaw: true,
-    });
-    const result = (await structuredModel.invoke([
-      { role: "system", content: request.systemPrompt },
-      { role: "user", content: request.userContext },
-    ])) as unknown as {
-      parsed?: z.infer<T>;
-      raw?: {
-        usage_metadata?: {
-          input_tokens?: number;
-          output_tokens?: number;
-          total_tokens?: number;
-        };
-      };
-    };
-
-    const parsed = result?.parsed;
-    const raw = result?.raw;
-
-    const usage = {
-      inputTokens: raw?.usage_metadata?.input_tokens || 0,
-      outputTokens: raw?.usage_metadata?.output_tokens || 0,
-      totalTokens: raw?.usage_metadata?.total_tokens || 0,
-      modelName: this.modelNameUsed,
-      providerInstanceName: this.providerInstanceName || "Default",
-      maxContext:
-        this.maxContextUsed !== undefined ? this.maxContextUsed : 32768,
-    };
-
-    this.lastCalls.push({
-      systemPrompt: request.systemPrompt,
-      userContext: request.userContext,
-      usage,
-    });
-
-    return { success: true, data: parsed, usage };
+    this.model = new ChatOpenRouter({ apiKey: key, model: this.modelNameUsed });
   }
 }

packages/llm/src/registry.ts

@@ -0,0 +1,79 @@
+import type {
+  ILLMProvider,
+  IEmbeddingProvider,
+  ModelProviderInstance,
+  ModelProviderMeta,
+} from "./llm.js";
+import type { ModelInfo } from "./model-lister.js";
+
+export interface ProviderDefinition {
+  id: string;
+  displayName: string;
+  description: string;
+  envVar?: string;
+  capabilities: { generative: boolean; embedding: boolean };
+  defaultModel: string;
+  defaultEmbeddingModel?: string;
+  defaultMaxContext: number;
+  fallbackPriority: number;
+  listModels?: (apiKey: string, endpointUrl?: string) => Promise<ModelInfo[]>;
+  generativeCreate?: (inst: ModelProviderInstance) => ILLMProvider;
+  embeddingCreate?: (inst: ModelProviderInstance) => IEmbeddingProvider;
+}
+
+const _entries = new Map<string, ProviderDefinition>();
+
+type ProviderMeta = Omit<
+  ProviderDefinition,
+  "generativeCreate" | "embeddingCreate"
+>;
+
+export function registerProvider(meta: ProviderMeta) {
+  const existing = _entries.get(meta.id);
+  _entries.set(meta.id, {
+    ...existing,
+    ...meta,
+    generativeCreate: existing?.generativeCreate,
+    embeddingCreate: existing?.embeddingCreate,
+  });
+}
+
+export function registerGenerative(
+  id: string,
+  createFn: (inst: ModelProviderInstance) => ILLMProvider,
+) {
+  const existing = _entries.get(id);
+  if (existing) {
+    existing.generativeCreate = createFn;
+  } else {
+    _entries.set(id, { id, generativeCreate: createFn } as ProviderDefinition);
+  }
+}
+
+export function registerEmbedding(
+  id: string,
+  createFn: (inst: ModelProviderInstance) => IEmbeddingProvider,
+) {
+  const existing = _entries.get(id);
+  if (existing) {
+    existing.embeddingCreate = createFn;
+  } else {
+    _entries.set(id, { id, embeddingCreate: createFn } as ProviderDefinition);
+  }
+}
+
+export const ProviderRegistry = {
+  all: (): ProviderDefinition[] => [..._entries.values()],
+  get: (id: string): ProviderDefinition | undefined => _entries.get(id),
+  has: (id: string): boolean => _entries.has(id),
+} as const;
+
+export function toProviderMeta(def: ProviderDefinition): ModelProviderMeta {
+  return {
+    id: def.id,
+    displayName: def.displayName,
+    description: def.description,
+    defaultModel: def.defaultModel,
+    defaultEmbeddingModel: def.defaultEmbeddingModel || "",
+  };
+}

packages/llm/src/row-mapper.ts

@@ -0,0 +1,32 @@
+import type { ModelProviderInstance } from "./llm.js";
+
+export type DbRow = {
+  id: string;
+  name: string;
+  providerName: string;
+  apiKey: string;
+  isActive: number;
+  modelName?: string;
+  type: string;
+  maxContext?: number;
+  endpointUrl?: string;
+};
+
+export function mapRow(r: DbRow): ModelProviderInstance {
+  return {
+    id: r.id,
+    name: r.name,
+    providerName: r.providerName,
+    apiKey: r.apiKey,
+    isActive: r.isActive === 1,
+    modelName: r.modelName || undefined,
+    type: (r.type as "generative" | "embedding") || "generative",
+    maxContext:
+      r.maxContext !== undefined && r.maxContext !== null
+        ? r.maxContext
+        : r.type === "embedding"
+          ? 0
+          : 32768,
+    endpointUrl: r.endpointUrl || undefined,
+  };
+}

packages/llm/tests/cli.test.ts

@@ -0,0 +1,126 @@
+import { describe, test, expect, beforeEach, afterEach } from "vitest";
+import { execSync } from "child_process";
+import fs from "fs";
+import path from "path";
+import Database from "better-sqlite3";
+
+describe("setup-provider CLI Tool Tests", () => {
+  let tempDbPath: string;
+  let scriptPath: string;
+
+  beforeEach(() => {
+    // Generate a unique temp database path
+    tempDbPath = path.resolve(
+      process.cwd(),
+      `test-cli-${Date.now()}-${Math.random().toString(36).substring(2)}.db`,
+    );
+    scriptPath = path.resolve(
+      process.cwd(),
+      "packages/llm/dist/bin/setup-provider.js",
+    );
+  });
+
+  afterEach(() => {
+    if (fs.existsSync(tempDbPath)) {
+      try {
+        fs.unlinkSync(tempDbPath);
+      } catch {
+        // ignore
+      }
+    }
+  });
+
+  test("prints help message when --help or -h is passed", () => {
+    const stdout = execSync(`node ${scriptPath} --help`).toString();
+    expect(stdout).toContain("Usage:");
+    expect(stdout).toContain("Options:");
+    expect(stdout).toContain("Registered Providers:");
+  });
+
+  test("creates a provider instance successfully via CLI flags", () => {
+    const cmd = `node ${scriptPath} --provider google-genai --key mock-key-abc --name "Test Gemini" --model "gemini-2.5-flash"`;
+    const stdout = execSync(cmd, {
+      env: { ...process.env, OMNIA_DB_PATH: tempDbPath },
+    }).toString();
+
+    expect(stdout).toContain("Successfully created provider instance:");
+    expect(stdout).toContain("Test Gemini");
+    expect(stdout).toContain("google-genai");
+    expect(stdout).toContain("mock-key-abc");
+
+    // Read the SQLite db directly to verify
+    const db = new Database(tempDbPath);
+    const rows = db.prepare("SELECT * FROM provider_instances").all() as {
+      name: string;
+      providerName: string;
+      apiKey: string;
+      modelName: string;
+      isActive: number;
+    }[];
+    expect(rows.length).toBe(1);
+    expect(rows[0].name).toBe("Test Gemini");
+    expect(rows[0].providerName).toBe("google-genai");
+    expect(rows[0].apiKey).toBe("mock-key-abc");
+    expect(rows[0].modelName).toBe("gemini-2.5-flash");
+    expect(rows[0].isActive).toBe(1);
+    db.close();
+  });
+
+  test("fails when required key is missing and env var is not set", () => {
+    let error: { status?: number; stderr?: Buffer } | undefined;
+    try {
+      execSync(`node ${scriptPath} --provider google-genai`, {
+        env: { ...process.env, OMNIA_DB_PATH: tempDbPath, GOOGLE_API_KEY: "" },
+        stdio: "pipe",
+      });
+    } catch (e) {
+      error = e as { status?: number; stderr?: Buffer };
+    }
+    expect(error).toBeDefined();
+    expect(error?.status).toBe(1);
+    expect(error?.stderr?.toString()).toContain("Error: API Key is required");
+  });
+
+  test("seeds from environment variables when using --all", () => {
+    const cmd = `node ${scriptPath} --all`;
+    const stdout = execSync(cmd, {
+      env: {
+        ...process.env,
+        OMNIA_DB_PATH: tempDbPath,
+        GOOGLE_API_KEY: "mock-google-key-all",
+        OPENAI_API_KEY: "",
+        ANTHROPIC_API_KEY: "",
+        GROQ_API_KEY: "",
+        DEEPSEEK_API_KEY: "",
+        OPENROUTER_API_KEY: "",
+      },
+    }).toString();
+
+    expect(stdout).toContain(
+      "Created generative instance: Google Gemini (CLI)",
+    );
+    expect(stdout).toContain(
+      "Created embedding instance: Google Gemini Embed (CLI)",
+    );
+
+    const db = new Database(tempDbPath);
+    const rows = db.prepare("SELECT * FROM provider_instances").all() as {
+      type: string;
+      apiKey: string;
+      modelName: string;
+    }[];
+    // Should have both generative and embedding instances
+    expect(rows.length).toBe(2);
+    const gen = rows.find((r) => r.type === "generative");
+    const embed = rows.find((r) => r.type === "embedding");
+
+    expect(gen).toBeDefined();
+    expect(gen?.apiKey).toBe("mock-google-key-all");
+    expect(gen?.modelName).toBe("gemini-2.5-flash");
+
+    expect(embed).toBeDefined();
+    expect(embed?.apiKey).toBe("mock-google-key-all");
+    expect(embed?.modelName).toBe("gemini-embedding-001");
+    db.close();
+  });
+});

packages/llm/tests/deepseek.test.ts

@@ -0,0 +1,128 @@
+import { describe, test, expect, vi } from "vitest";
+import { z } from "zod";
+
+const mockConfig: Record<string, string | undefined> = {};
+
+vi.mock("../src/config.js", () => ({
+  getLlmConfig: () => mockConfig,
+  resetLlmConfig: () => {
+    for (const key of Object.keys(mockConfig)) {
+      delete mockConfig[key];
+    }
+  },
+}));
+
+const { getActiveMock } = vi.hoisted(() => ({
+  getActiveMock: vi.fn().mockReturnValue(null),
+}));
+
+vi.mock("../src/provider-manager.js", async (importOriginal) => {
+  const actual =
+    await importOriginal<typeof import("../src/provider-manager.js")>();
+  return {
+    ...actual,
+    ProviderManager: {
+      ...actual.ProviderManager,
+      getActive: getActiveMock,
+    },
+  };
+});
+
+import { DeepSeekProvider } from "../src/providers/deepseek.js";
+
+// Mock the ChatDeepSeek class
+vi.mock("@langchain/deepseek", () => {
+  return {
+    ChatDeepSeek: class {
+      config: unknown;
+      constructor(config: unknown) {
+        this.config = config;
+      }
+      withStructuredOutput = vi.fn().mockImplementation(() => {
+        return {
+          invoke: vi.fn().mockImplementation(async () => {
+            return {
+              parsed: {
+                name: "mocked response",
+                success: true,
+              },
+              raw: {
+                usage_metadata: {
+                  input_tokens: 10,
+                  output_tokens: 5,
+                  total_tokens: 15,
+                },
+              },
+            };
+          }),
+        };
+      });
+    },
+  };
+});
+
+describe("DeepSeekProvider Unit Tests (Tier 1)", () => {
+  test("initializes successfully with a provided apiKey", () => {
+    const provider = new DeepSeekProvider("dummy-key");
+    expect(provider.providerName).toBe("DeepSeek");
+  });
+
+  test("initializes successfully with apiKey from config", () => {
+    const originalKey = process.env.DEEPSEEK_API_KEY;
+    process.env.DEEPSEEK_API_KEY = "env-dummy-key";
+    mockConfig.DEEPSEEK_API_KEY = "env-dummy-key";
+
+    try {
+      const provider = new DeepSeekProvider();
+      expect(provider.providerName).toBe("DeepSeek");
+    } finally {
+      process.env.DEEPSEEK_API_KEY = originalKey;
+      delete mockConfig.DEEPSEEK_API_KEY;
+    }
+  });
+
+  test("throws error if no API key is provided or in config", () => {
+    const originalKey = process.env.DEEPSEEK_API_KEY;
+    process.env.DEEPSEEK_API_KEY = undefined;
+    mockConfig.DEEPSEEK_API_KEY = undefined;
+
+    try {
+      expect(() => new DeepSeekProvider()).toThrow(
+        "DEEPSEEK_API_KEY is required to initialize DeepSeekProvider",
+      );
+    } finally {
+      process.env.DEEPSEEK_API_KEY = originalKey;
+    }
+  });
+
+  test("generateStructuredResponse invokes the model with structured output, records usage and updates lastCalls", async () => {
+    const provider = new DeepSeekProvider("dummy-key");
+    const TestSchema = z.object({
+      name: z.string(),
+      success: z.boolean(),
+    });
+
+    const response = await provider.generateStructuredResponse({
+      systemPrompt: "system prompt",
+      userContext: "user context",
+      schema: TestSchema,
+    });
+
+    expect(response.success).toBe(true);
+    expect(response.data).toEqual({
+      name: "mocked response",
+      success: true,
+    });
+
+    expect(response.usage).toEqual({
+      inputTokens: 10,
+      outputTokens: 5,
+      totalTokens: 15,
+      modelName: "deepseek-chat",
+      providerInstanceName: "Default",
+      maxContext: 64000,
+    });
+
+    expect(provider.lastCalls.length).toBe(1);
+  });
+});

packages/llm/tests/groq.test.ts

@@ -0,0 +1,128 @@
+import { describe, test, expect, vi } from "vitest";
+import { z } from "zod";
+
+const mockConfig: Record<string, string | undefined> = {};
+
+vi.mock("../src/config.js", () => ({
+  getLlmConfig: () => mockConfig,
+  resetLlmConfig: () => {
+    for (const key of Object.keys(mockConfig)) {
+      delete mockConfig[key];
+    }
+  },
+}));
+
+const { getActiveMock } = vi.hoisted(() => ({
+  getActiveMock: vi.fn().mockReturnValue(null),
+}));
+
+vi.mock("../src/provider-manager.js", async (importOriginal) => {
+  const actual =
+    await importOriginal<typeof import("../src/provider-manager.js")>();
+  return {
+    ...actual,
+    ProviderManager: {
+      ...actual.ProviderManager,
+      getActive: getActiveMock,
+    },
+  };
+});
+
+import { GroqProvider } from "../src/providers/groq.js";
+
+// Mock the ChatGroq class
+vi.mock("@langchain/groq", () => {
+  return {
+    ChatGroq: class {
+      config: unknown;
+      constructor(config: unknown) {
+        this.config = config;
+      }
+      withStructuredOutput = vi.fn().mockImplementation(() => {
+        return {
+          invoke: vi.fn().mockImplementation(async () => {
+            return {
+              parsed: {
+                name: "mocked response",
+                success: true,
+              },
+              raw: {
+                usage_metadata: {
+                  input_tokens: 10,
+                  output_tokens: 5,
+                  total_tokens: 15,
+                },
+              },
+            };
+          }),
+        };
+      });
+    },
+  };
+});
+
+describe("GroqProvider Unit Tests (Tier 1)", () => {
+  test("initializes successfully with a provided apiKey", () => {
+    const provider = new GroqProvider("dummy-key");
+    expect(provider.providerName).toBe("Groq");
+  });
+
+  test("initializes successfully with apiKey from config", () => {
+    const originalKey = process.env.GROQ_API_KEY;
+    process.env.GROQ_API_KEY = "env-dummy-key";
+    mockConfig.GROQ_API_KEY = "env-dummy-key";
+
+    try {
+      const provider = new GroqProvider();
+      expect(provider.providerName).toBe("Groq");
+    } finally {
+      process.env.GROQ_API_KEY = originalKey;
+      delete mockConfig.GROQ_API_KEY;
+    }
+  });
+
+  test("throws error if no API key is provided or in config", () => {
+    const originalKey = process.env.GROQ_API_KEY;
+    process.env.GROQ_API_KEY = undefined;
+    mockConfig.GROQ_API_KEY = undefined;
+
+    try {
+      expect(() => new GroqProvider()).toThrow(
+        "GROQ_API_KEY is required to initialize GroqProvider",
+      );
+    } finally {
+      process.env.GROQ_API_KEY = originalKey;
+    }
+  });
+
+  test("generateStructuredResponse invokes the model with structured output, records usage and updates lastCalls", async () => {
+    const provider = new GroqProvider("dummy-key");
+    const TestSchema = z.object({
+      name: z.string(),
+      success: z.boolean(),
+    });
+
+    const response = await provider.generateStructuredResponse({
+      systemPrompt: "system prompt",
+      userContext: "user context",
+      schema: TestSchema,
+    });
+
+    expect(response.success).toBe(true);
+    expect(response.data).toEqual({
+      name: "mocked response",
+      success: true,
+    });
+
+    expect(response.usage).toEqual({
+      inputTokens: 10,
+      outputTokens: 5,
+      totalTokens: 15,
+      modelName: "llama-3.3-70b-versatile",
+      providerInstanceName: "Default",
+      maxContext: 8192,
+    });
+
+    expect(provider.lastCalls.length).toBe(1);
+  });
+});

packages/llm/tests/model-lister.test.ts

@@ -0,0 +1,120 @@
+import { describe, test, expect, vi, beforeEach, afterEach } from "vitest";
+import { ModelLister } from "@omnia/llm";
+
+describe("ModelLister Unit Tests (Tier 1)", () => {
+  beforeEach(() => {
+    vi.stubGlobal("fetch", vi.fn());
+    ModelLister.clearCache();
+  });
+
+  afterEach(() => {
+    vi.restoreAllMocks();
+  });
+
+  test("returns mock provider list instantly without fetch", async () => {
+    const models = await ModelLister.listModels("mock", "none");
+    expect(models).toEqual([{ id: "mock", name: "Mock Model" }]);
+    expect(fetch).not.toHaveBeenCalled();
+  });
+
+  test("fetches and caches OpenAI-compatible models", async () => {
+    const mockResponse = {
+      data: [
+        { id: "gpt-4o", owned_by: "openai" },
+        { id: "gpt-4o-mini", owned_by: "openai" },
+      ],
+    };
+
+    const mockFetch = vi.fn().mockResolvedValue({
+      ok: true,
+      json: async () => mockResponse,
+    });
+    vi.stubGlobal("fetch", mockFetch);
+
+    // First call: Should fetch
+    const models = await ModelLister.listModels("openai", "test-key");
+    expect(models).toEqual([
+      { id: "gpt-4o", name: "gpt-4o", ownedBy: "openai" },
+      { id: "gpt-4o-mini", name: "gpt-4o-mini", ownedBy: "openai" },
+    ]);
+    expect(mockFetch).toHaveBeenCalledTimes(1);
+    expect(mockFetch).toHaveBeenCalledWith(
+      "https://api.openai.com/v1/models",
+      expect.objectContaining({
+        headers: {
+          Authorization: "Bearer test-key",
+          Accept: "application/json",
+        },
+      }),
+    );
+
+    // Second call: Should read from cache
+    const cachedModels = await ModelLister.listModels("openai", "test-key");
+    expect(cachedModels).toEqual(models);
+    expect(mockFetch).toHaveBeenCalledTimes(1);
+  });
+
+  test("respects cache invalidation", async () => {
+    const mockFetch = vi.fn().mockResolvedValue({
+      ok: true,
+      json: async () => ({ data: [{ id: "model-1" }] }),
+    });
+    vi.stubGlobal("fetch", mockFetch);
+
+    await ModelLister.listModels("openai", "test-key");
+    expect(mockFetch).toHaveBeenCalledTimes(1);
+
+    // Invalidate
+    ModelLister.invalidateCache("openai", "test-key");
+
+    // Second call: Should fetch again
+    await ModelLister.listModels("openai", "test-key");
+    expect(mockFetch).toHaveBeenCalledTimes(2);
+  });
+
+  test("gracefully returns empty array on fetch failure", async () => {
+    const mockFetch = vi.fn().mockResolvedValue({
+      ok: false,
+      status: 500,
+    });
+    vi.stubGlobal("fetch", mockFetch);
+
+    const models = await ModelLister.listModels("openai", "bad-key");
+    expect(models).toEqual([]);
+    expect(mockFetch).toHaveBeenCalledTimes(1);
+  });
+
+  test("handles Gemini pagination correctly", async () => {
+    const mockFetch = vi
+      .fn()
+      .mockResolvedValueOnce({
+        ok: true,
+        json: async () => ({
+          models: [
+            {
+              name: "models/gemini-2.5-flash",
+              displayName: "Gemini 2.5 Flash",
+            },
+          ],
+          nextPageToken: "token-1",
+        }),
+      })
+      .mockResolvedValueOnce({
+        ok: true,
+        json: async () => ({
+          models: [
+            { name: "models/gemini-2.5-pro", displayName: "Gemini 2.5 Pro" },
+          ],
+        }),
+      });
+
+    vi.stubGlobal("fetch", mockFetch);
+
+    const models = await ModelLister.listModels("google-genai", "gemini-key");
+    expect(models).toEqual([
+      { id: "gemini-2.5-flash", name: "Gemini 2.5 Flash" },
+      { id: "gemini-2.5-pro", name: "Gemini 2.5 Pro" },
+    ]);
+    expect(mockFetch).toHaveBeenCalledTimes(2);
+  });
+});

packages/llm/tests/openai.test.ts

@@ -0,0 +1,167 @@
+import { describe, test, expect, vi } from "vitest";
+import { z } from "zod";
+
+const mockConfig: Record<string, string | undefined> = {};
+
+vi.mock("../src/config.js", () => ({
+  getLlmConfig: () => mockConfig,
+  resetLlmConfig: () => {
+    for (const key of Object.keys(mockConfig)) {
+      delete mockConfig[key];
+    }
+  },
+}));
+
+const { getActiveMock } = vi.hoisted(() => ({
+  getActiveMock: vi.fn().mockReturnValue(null),
+}));
+
+vi.mock("../src/provider-manager.js", async (importOriginal) => {
+  const actual =
+    await importOriginal<typeof import("../src/provider-manager.js")>();
+  return {
+    ...actual,
+    ProviderManager: {
+      ...actual.ProviderManager,
+      getActive: getActiveMock,
+    },
+  };
+});
+
+import {
+  OpenAIProvider,
+  OpenAIEmbeddingProvider,
+} from "../src/providers/openai.js";
+
+// Mock the ChatOpenAI and OpenAIEmbeddings classes
+vi.mock("@langchain/openai", () => {
+  return {
+    ChatOpenAI: class {
+      config: unknown;
+      constructor(config: unknown) {
+        this.config = config;
+      }
+      withStructuredOutput = vi.fn().mockImplementation(() => {
+        return {
+          invoke: vi.fn().mockImplementation(async () => {
+            return {
+              parsed: {
+                name: "mocked response",
+                success: true,
+              },
+              raw: {
+                usage_metadata: {
+                  input_tokens: 10,
+                  output_tokens: 5,
+                  total_tokens: 15,
+                },
+              },
+            };
+          }),
+        };
+      });
+    },
+    OpenAIEmbeddings: class {
+      config: unknown;
+      constructor(config: unknown) {
+        this.config = config;
+      }
+      embedQuery = vi.fn().mockImplementation(async (text: string) => {
+        return [0.1, 0.2, 0.3];
+      });
+    },
+  };
+});
+
+describe("OpenAIProvider Unit Tests (Tier 1)", () => {
+  test("initializes successfully with a provided apiKey", () => {
+    const provider = new OpenAIProvider("dummy-key");
+    expect(provider.providerName).toBe("OpenAI");
+  });
+
+  test("initializes successfully with apiKey from config", () => {
+    const originalKey = process.env.OPENAI_API_KEY;
+    process.env.OPENAI_API_KEY = "env-dummy-key";
+    mockConfig.OPENAI_API_KEY = "env-dummy-key";
+
+    try {
+      const provider = new OpenAIProvider();
+      expect(provider.providerName).toBe("OpenAI");
+    } finally {
+      process.env.OPENAI_API_KEY = originalKey;
+      delete mockConfig.OPENAI_API_KEY;
+    }
+  });
+
+  test("throws error if no API key is provided or in config", () => {
+    const originalKey = process.env.OPENAI_API_KEY;
+    process.env.OPENAI_API_KEY = undefined;
+    mockConfig.OPENAI_API_KEY = undefined;
+
+    try {
+      expect(() => new OpenAIProvider()).toThrow(
+        "OPENAI_API_KEY is required to initialize OpenAIProvider",
+      );
+    } finally {
+      process.env.OPENAI_API_KEY = originalKey;
+    }
+  });
+
+  test("generateStructuredResponse invokes the model with structured output, records usage and updates lastCalls", async () => {
+    const provider = new OpenAIProvider("dummy-key");
+    const TestSchema = z.object({
+      name: z.string(),
+      success: z.boolean(),
+    });
+
+    const response = await provider.generateStructuredResponse({
+      systemPrompt: "system prompt",
+      userContext: "user context",
+      schema: TestSchema,
+    });
+
+    expect(response.success).toBe(true);
+    expect(response.data).toEqual({
+      name: "mocked response",
+      success: true,
+    });
+
+    expect(response.usage).toEqual({
+      inputTokens: 10,
+      outputTokens: 5,
+      totalTokens: 15,
+      modelName: "gpt-4o-mini",
+      providerInstanceName: "Default",
+      maxContext: 128000,
+    });
+
+    expect(provider.lastCalls.length).toBe(1);
+  });
+});
+
+describe("OpenAIEmbeddingProvider Unit Tests (Tier 1)", () => {
+  test("initializes successfully with a provided apiKey", () => {
+    const provider = new OpenAIEmbeddingProvider("dummy-key");
+    expect(provider.providerName).toBe("OpenAI");
+  });
+
+  test("initializes successfully with apiKey from config", () => {
+    const originalKey = process.env.OPENAI_API_KEY;
+    process.env.OPENAI_API_KEY = "env-dummy-key";
+    mockConfig.OPENAI_API_KEY = "env-dummy-key";
+
+    try {
+      const provider = new OpenAIEmbeddingProvider();
+      expect(provider.providerName).toBe("OpenAI");
+    } finally {
+      process.env.OPENAI_API_KEY = originalKey;
+      delete mockConfig.OPENAI_API_KEY;
+    }
+  });
+
+  test("embed returns dummy array successfully", async () => {
+    const provider = new OpenAIEmbeddingProvider("dummy-key");
+    const result = await provider.embed("hello");
+    expect(result).toEqual([0.1, 0.2, 0.3]);
+  });
+});

packages/llm/tests/openrouter.test.ts

@@ -1,7 +1,34 @@
 import { describe, test, expect, vi } from "vitest";
 import { z } from "zod";
+
+const mockConfig: Record<string, string | undefined> = {};
+
+vi.mock("../src/config.js", () => ({
+  getLlmConfig: () => mockConfig,
+  resetLlmConfig: () => {
+    for (const key of Object.keys(mockConfig)) {
+      delete mockConfig[key];
+    }
+  },
+}));
+
+const { getActiveMock } = vi.hoisted(() => ({
+  getActiveMock: vi.fn().mockReturnValue(null),
+}));
+
+vi.mock("../src/provider-manager.js", async (importOriginal) => {
+  const actual =
+    await importOriginal<typeof import("../src/provider-manager.js")>();
+  return {
+    ...actual,
+    ProviderManager: {
+      ...actual.ProviderManager,
+      getActive: getActiveMock,
+    },
+  };
+});
+
 import { OpenRouterProvider } from "../src/providers/openrouter.js";
-import { llmConfig } from "../src/config.js";
 
 // Mock the ChatOpenRouter class
 vi.mock("@langchain/openrouter", () => {
@@ -14,7 +41,6 @@ vi.mock("@langchain/openrouter", () => {
       withStructuredOutput = vi.fn().mockImplementation(() => {
         return {
           invoke: vi.fn().mockImplementation(async () => {
-            // Return a mock output that matches the includeRaw: true structure
             return {
               parsed: {
                 name: "mocked response",
@@ -42,29 +68,30 @@ describe("OpenRouterProvider Unit Tests (Tier 1)", () => {
   });
 
   test("initializes successfully with apiKey from config", () => {
-    // Save current config
-    const originalKey = llmConfig.OPENROUTER_API_KEY;
-    llmConfig.OPENROUTER_API_KEY = "env-dummy-key";
+    const originalKey = process.env.OPENROUTER_API_KEY;
+    process.env.OPENROUTER_API_KEY = "env-dummy-key";
+    mockConfig.OPENROUTER_API_KEY = "env-dummy-key";
 
     try {
       const provider = new OpenRouterProvider();
       expect(provider.providerName).toBe("OpenRouter");
     } finally {
-      llmConfig.OPENROUTER_API_KEY = originalKey;
+      process.env.OPENROUTER_API_KEY = originalKey;
+      delete mockConfig.OPENROUTER_API_KEY;
     }
   });
 
   test("throws error if no API key is provided or in config", () => {
-    // Save current config
-    const originalKey = llmConfig.OPENROUTER_API_KEY;
-    llmConfig.OPENROUTER_API_KEY = undefined;
+    const originalKey = process.env.OPENROUTER_API_KEY;
+    process.env.OPENROUTER_API_KEY = undefined;
+    mockConfig.OPENROUTER_API_KEY = undefined;
 
     try {
       expect(() => new OpenRouterProvider()).toThrow(
         "OPENROUTER_API_KEY is required to initialize OpenRouterProvider",
       );
     } finally {
-      llmConfig.OPENROUTER_API_KEY = originalKey;
+      process.env.OPENROUTER_API_KEY = originalKey;
     }
   });
 

packages/llm/tests/provider-manager.test.ts

@@ -1,24 +1,27 @@
 import { describe, test, expect, beforeEach, afterEach } from "vitest";
 import fs from "fs";
 import path from "path";
-import {
-  ProviderManager,
-  setDbPathOverride,
-  resetHasBootstrapped,
-} from "../src/index.js";
+import { ProviderManager, setDbPathOverride } from "../src/index.js";
 
 describe("ProviderManager Bootstrapping & CRUD Unit Tests", () => {
   let tempDbPath: string;
-  let originalGoogle: string | undefined;
-  let originalOpenRouter: string | undefined;
+  let savedEnv: Record<string, string | undefined>;
 
   beforeEach(() => {
-    originalGoogle = process.env.GOOGLE_API_KEY;
-    originalOpenRouter = process.env.OPENROUTER_API_KEY;
+    savedEnv = {
+      GOOGLE_API_KEY: process.env.GOOGLE_API_KEY,
+      OPENROUTER_API_KEY: process.env.OPENROUTER_API_KEY,
+      ANTHROPIC_API_KEY: process.env.ANTHROPIC_API_KEY,
+      OPENAI_API_KEY: process.env.OPENAI_API_KEY,
+      GROQ_API_KEY: process.env.GROQ_API_KEY,
+      DEEPSEEK_API_KEY: process.env.DEEPSEEK_API_KEY,
+    };
     delete process.env.GOOGLE_API_KEY;
     delete process.env.OPENROUTER_API_KEY;
-
-    resetHasBootstrapped();
+    delete process.env.ANTHROPIC_API_KEY;
+    delete process.env.OPENAI_API_KEY;
+    delete process.env.GROQ_API_KEY;
+    delete process.env.DEEPSEEK_API_KEY;
 
     // Generate a unique temp database path for this test run
     tempDbPath = path.resolve(
@@ -37,54 +40,189 @@ describe("ProviderManager Bootstrapping & CRUD Unit Tests", () => {
         // ignore
       }
     }
-    if (originalGoogle !== undefined) {
-      process.env.GOOGLE_API_KEY = originalGoogle;
-    } else {
-      delete process.env.GOOGLE_API_KEY;
-    }
-    if (originalOpenRouter !== undefined) {
-      process.env.OPENROUTER_API_KEY = originalOpenRouter;
-    } else {
-      delete process.env.OPENROUTER_API_KEY;
+    for (const [key, value] of Object.entries(savedEnv)) {
+      if (value !== undefined) {
+        process.env[key] = value;
+      } else {
+        delete process.env[key];
+      }
     }
   });
 
-  test("auto-bootstraps Gemini and OpenRouter when database is empty and environment variables are present", () => {
-    process.env.GOOGLE_API_KEY = "mock-google-key-123";
-    process.env.OPENROUTER_API_KEY = "mock-openrouter-key-456";
-
+  test("returns empty list when database is empty and no auto-bootstraps", () => {
+    process.env.GOOGLE_API_KEY = "mock-google-key";
     const list = ProviderManager.list();
-    expect(list.length).toBe(3);
-
-    const gemini = list.find((p) => p.providerName === "google-genai");
-    expect(gemini).toBeDefined();
-    expect(gemini?.name).toBe("Gemini (Env)");
-    expect(gemini?.apiKey).toBe("mock-google-key-123");
-    expect(gemini?.modelName).toBe("gemini-2.5-flash");
-    expect(gemini?.isActive).toBe(true); // first inserted is active
-
-    const openrouter = list.find((p) => p.providerName === "openrouter");
-    expect(openrouter).toBeDefined();
-    expect(openrouter?.name).toBe("OpenRouter (Env)");
-    expect(openrouter?.apiKey).toBe("mock-openrouter-key-456");
-    expect(openrouter?.modelName).toBe("google/gemini-2.5-flash");
-    expect(openrouter?.isActive).toBe(false); // second inserted is inactive
+    expect(list.length).toBe(0);
   });
 
-  test("treats bootstrapped instances as normal provider instances (editable and deletable)", () => {
-    process.env.GOOGLE_API_KEY = "mock-google-key-123";
+  test("getActive returns null when no providers exist and no env vars", () => {
+    const active = ProviderManager.getActive("generative");
+    expect(active).toBeNull();
+    const activeEmbed = ProviderManager.getActive("embedding");
+    expect(activeEmbed).toBeNull();
+  });
+
+  test("getActive returns null when DB is empty", () => {
+    process.env.GOOGLE_API_KEY = "mock-google-key-123";
+    const active = ProviderManager.getActive("generative");
+    expect(active).toBeNull();
+  });
+
+  test("getActive returns first instance of type when none is active", () => {
+    // Manually create instances without any env var bootstrap
+    const inst1 = ProviderManager.create("Test Gemini", "google-genai", "key1");
+    const inst2 = ProviderManager.create(
+      "Test OpenAI",
+      "openai",
+      "key2",
+      "gpt-4o",
+      "generative",
+      128000,
+    );
+    expect(inst1.isActive).toBe(true); // first created auto-activates
+    expect(inst2.isActive).toBe(false);
+
+    // Deactivate both
+    ProviderManager.setActive("__nonexistent__"); // no-op for nonexistent
+
+    // Deactivate inst1 by setting another as active, then delete that
+    ProviderManager.setActive(inst2.id);
+    expect(
+      ProviderManager.list().find((p) => p.id === inst2.id)?.isActive,
+    ).toBe(true);
+    expect(
+      ProviderManager.list().find((p) => p.id === inst1.id)?.isActive,
+    ).toBe(false);
+
+    // Delete the active one → auto-promotes inst1
+    ProviderManager.delete(inst2.id);
+    const promoted = ProviderManager.list().find((p) => p.id === inst1.id);
+    expect(promoted?.isActive).toBe(true);
+  });
+
+  test("setActive correctly deactivates siblings and activates target", () => {
+    const inst1 = ProviderManager.create(
+      "First Gemini",
+      "google-genai",
+      "key1",
+      undefined,
+      "generative",
+    );
+    const inst2 = ProviderManager.create(
+      "Second Gemini",
+      "google-genai",
+      "key2",
+      undefined,
+      "generative",
+    );
+
+    expect(inst1.isActive).toBe(true);
+    expect(inst2.isActive).toBe(false);
+
+    ProviderManager.setActive(inst2.id);
 
-    // Trigger bootstrap
     const list = ProviderManager.list();
-    expect(list.length).toBe(2);
-    const bootstrapped = list.find((p) => p.name === "Gemini (Env)");
-    expect(bootstrapped).toBeDefined();
-    if (!bootstrapped) return;
-    expect(bootstrapped.isActive).toBe(true);
+    const updated1 = list.find((p) => p.id === inst1.id);
+    const updated2 = list.find((p) => p.id === inst2.id);
+    expect(updated1?.isActive).toBe(false);
+    expect(updated2?.isActive).toBe(true);
+  });
+
+  test("getMappings returns empty object initially, setMapping persists mappings", () => {
+    const mappings = ProviderManager.getMappings();
+    expect(mappings).toEqual({});
+
+    const inst = ProviderManager.create(
+      "Test Provider",
+      "google-genai",
+      "key1",
+    );
+
+    ProviderManager.setMapping("actor-prose", inst.id);
+    ProviderManager.setMapping("embeddings", inst.id);
+
+    const updated = ProviderManager.getMappings();
+    expect(updated["actor-prose"]).toBe(inst.id);
+    expect(updated["embeddings"]).toBe(inst.id);
+  });
+
+  test("setMapping with empty providerInstanceId deletes the mapping", () => {
+    const inst = ProviderManager.create(
+      "Test Provider",
+      "google-genai",
+      "key1",
+    );
+
+    ProviderManager.setMapping("test-task", inst.id);
+    expect(ProviderManager.getMappings()["test-task"]).toBe(inst.id);
+
+    ProviderManager.setMapping("test-task", "");
+    expect(ProviderManager.getMappings()["test-task"]).toBeUndefined();
+  });
+
+  test("create returns instance with correct fields and endpointUrl support", () => {
+    const inst = ProviderManager.create(
+      "Ollama Local",
+      "ollama",
+      "",
+      "llama3.1",
+      "generative",
+      32768,
+      "http://localhost:11434",
+    );
+
+    expect(inst.id).toMatch(/^provider-/);
+    expect(inst.name).toBe("Ollama Local");
+    expect(inst.providerName).toBe("ollama");
+    expect(inst.modelName).toBe("llama3.1");
+    expect(inst.endpointUrl).toBe("http://localhost:11434");
+  });
+
+  test("update preserves apiKey when not provided", () => {
+    const inst = ProviderManager.create(
+      "Original",
+      "openai",
+      "original-key",
+      "gpt-4o",
+      "generative",
+      128000,
+    );
+
+    ProviderManager.update(
+      inst.id,
+      "Renamed",
+      "openai",
+      undefined, // no apiKey → preserve existing
+      "gpt-4o-mini",
+      "generative",
+      64000,
+    );
+
+    const updated = ProviderManager.list().find((p) => p.id === inst.id);
+    expect(updated?.name).toBe("Renamed");
+    expect(updated?.apiKey).toBe("original-key"); // preserved
+    expect(updated?.modelName).toBe("gpt-4o-mini");
+    expect(updated?.maxContext).toBe(64000);
+  });
+  test("treats created instances as normal provider instances (editable and deletable)", () => {
+    const inst = ProviderManager.create(
+      "Google Gemini (Env)",
+      "google-genai",
+      "mock-google-key-123",
+      "gemini-2.5-flash",
+      "generative",
+    );
+
+    const list = ProviderManager.list();
+    expect(list.length).toBe(1);
+    const created = list.find((p) => p.id === inst.id);
+    expect(created).toBeDefined();
+    if (!created) return;
+    expect(created.isActive).toBe(true);
 
     // Edit name and key
     ProviderManager.update(
-      bootstrapped.id,
+      created.id,
       "My Gemini Key",
       "google-genai",
       "new-secret-key",
@@ -92,8 +230,8 @@ describe("ProviderManager Bootstrapping & CRUD Unit Tests", () => {
     );
 
     const listAfterUpdate = ProviderManager.list();
-    expect(listAfterUpdate.length).toBe(2);
-    const updated = listAfterUpdate.find((p) => p.id === bootstrapped.id);
+    expect(listAfterUpdate.length).toBe(1);
+    const updated = listAfterUpdate.find((p) => p.id === created.id);
     expect(updated).toBeDefined();
     if (!updated) return;
     expect(updated.name).toBe("My Gemini Key");
@@ -101,8 +239,8 @@ describe("ProviderManager Bootstrapping & CRUD Unit Tests", () => {
     expect(updated.modelName).toBe("gemini-2.5-pro");
 
     // Delete instance
-    ProviderManager.delete(bootstrapped.id);
+    ProviderManager.delete(created.id);
     const listAfterDelete = ProviderManager.list();
-    expect(listAfterDelete.length).toBe(1);
+    expect(listAfterDelete.length).toBe(0);
   });
 });

packages/memory/src/handoff.ts

@@ -259,11 +259,7 @@ ${candidatesList}
     }
 
     const result = response.data;
-    const db = (
-      this.bufferRepo as unknown as {
-        db: { transaction: (fn: () => void) => void };
-      }
-    ).db;
+    const db = (this.bufferRepo as any).db;
 
     const ledgerEntries: LedgerEntry[] = [];
     for (const chunk of result.chunks) {

pnpm-lock.yaml

@@ -260,12 +260,27 @@ settings:
 importers:
   .:
     dependencies:
+      "@langchain/anthropic":
+        specifier: ^0.3.11
+        version: 0.3.34(zod@4.4.3)
+      "@langchain/deepseek":
+        specifier: ^1.1.5
+        version: 1.1.5(ws@8.21.0)
       "@langchain/google-genai":
         specifier: ^2.2.0
         version: 2.2.0(@langchain/core@1.2.1(openai@6.45.0(ws@8.21.0)(zod@4.4.3))(ws@8.21.0))
+      "@langchain/groq":
+        specifier: ^1.3.1
+        version: 1.3.1
+      "@langchain/ollama":
+        specifier: ^0.2.3
+        version: 0.2.4
+      "@langchain/openai":
+        specifier: ^0.3.17
+        version: 0.3.17(ws@8.21.0)
       "@langchain/openrouter":
         specifier: ^0.4.3
-        version: 0.4.3(@langchain/core@1.2.1(openai@6.45.0(ws@8.21.0)(zod@4.4.3))(ws@8.21.0))(ws@8.21.0)(zod@4.4.3)
+        version: 0.4.3(ws@8.21.0)(zod@4.4.3)
       "@types/node":
         specifier: ^20.19.43
         version: 20.19.43
@@ -561,6 +576,18 @@ packages:
         integrity: sha512-MGQsmw10ZyI+EJo45CdSER4zEb+p31LpDAFp2Z3gkSd1yqVZGi0Ebx++YTEMonJy4oChEMLsxZ64j8FH6sSqtQ==,
       }
 
+  "@anthropic-ai/sdk@0.65.0":
+    resolution:
+      {
+        integrity: sha512-zIdPOcrCVEI8t3Di40nH4z9EoeyGZfXbYSvWdDLsB/KkaSYMnEgC7gmcgWu83g2NTn1ZTpbMvpdttWDGGIk6zw==,
+      }
+    hasBin: true
+    peerDependencies:
+      zod: ^3.25.0 || ^4.0.0
+    peerDependenciesMeta:
+      zod:
+        optional: true
+
   "@astrojs/compiler-binding-darwin-arm64@0.3.0":
     resolution:
       {
@@ -2049,6 +2076,15 @@ packages:
         integrity: sha512-3Belt6tdc8bPgAtbcmdtNJlirVoTmEb5e2gC94PnkwEW9jI6CAHUeoG85tjWP5WquqfavoMtMwiG4P926ZKKuQ==,
       }
 
+  "@langchain/anthropic@0.3.34":
+    resolution:
+      {
+        integrity: sha512-8bOW1A2VHRCjbzdYElrjxutKNs9NSIxYRGtR+OJWVzluMqoKKh2NmmFrpPizEyqCUEG2tTq5xt6XA1lwfqMJRA==,
+      }
+    engines: { node: ">=18" }
+    peerDependencies:
+      "@langchain/core": ">=0.3.58 <0.4.0"
+
   "@langchain/core@1.2.1":
     resolution:
       {
@@ -2056,6 +2092,15 @@ packages:
       }
     engines: { node: ">=20" }
 
+  "@langchain/deepseek@1.1.5":
+    resolution:
+      {
+        integrity: sha512-5IRoEUaHAgIF8TyIncNVhhjavCqsjWTjakWsnus1yJN2X3W15Bw8Qmf+vJzCnFo7yndICsGdOGfHJoIN5xNxoQ==,
+      }
+    engines: { node: ">=20" }
+    peerDependencies:
+      "@langchain/core": ^1.0.0
+
   "@langchain/google-genai@2.2.0":
     resolution:
       {
@@ -2065,6 +2110,33 @@ packages:
     peerDependencies:
       "@langchain/core": ^1.2.0
 
+  "@langchain/groq@1.3.1":
+    resolution:
+      {
+        integrity: sha512-ImxfGBis4FEHpZdeT6dot6V6l09uBLYm/1BHQ6x3XEQmJFF7aKwbOeSOV5h/h5IeRx+2gaInR+LfyYoqT8satQ==,
+      }
+    engines: { node: ">=20" }
+    peerDependencies:
+      "@langchain/core": ^1.1.30
+
+  "@langchain/ollama@0.2.4":
+    resolution:
+      {
+        integrity: sha512-XThDrZurNPcUO6sasN13rkes1aGgu5gWAtDkkyIGT3ZeMOvrYgPKGft+bbhvsigTIH9C01TfPzrSp8LAmvHIjA==,
+      }
+    engines: { node: ">=18" }
+    peerDependencies:
+      "@langchain/core": ">=0.3.58 <0.4.0"
+
+  "@langchain/openai@0.3.17":
+    resolution:
+      {
+        integrity: sha512-uw4po32OKptVjq+CYHrumgbfh4NuD7LqyE+ZgqY9I/LrLc6bHLMc+sisHmI17vgek0K/yqtarI0alPJbzrwyag==,
+      }
+    engines: { node: ">=18" }
+    peerDependencies:
+      "@langchain/core": ">=0.3.29 <0.4.0"
+
   "@langchain/openai@1.5.3":
     resolution:
       {
@@ -2074,6 +2146,15 @@ packages:
     peerDependencies:
       "@langchain/core": ^1.2.1
 
+  "@langchain/openai@1.5.5":
+    resolution:
+      {
+        integrity: sha512-wX7dwb9z4nf5FHXlIl/X2mk08pzonvRHCt1D4+s1zXLP0duYDC95j7dulPIQJ6fmhbyYQc9Ki8mEhY/D1lB8kw==,
+      }
+    engines: { node: ">=20" }
+    peerDependencies:
+      "@langchain/core": ^1.2.2
+
   "@langchain/openrouter@0.4.3":
     resolution:
       {
@@ -3858,6 +3939,18 @@ packages:
         integrity: sha512-vSYNSDe6Ix3q+6Z7ri9lyWqgGhJTmzRjZRqyq15N0Z/1/UnVsno9G/N40NBijoYx2seFDIl0+B2mgAb9mezUCA==,
       }
 
+  "@types/node-fetch@2.6.13":
+    resolution:
+      {
+        integrity: sha512-QGpRVpzSaUs30JBSGPjOg4Uveu384erbHBoT1zeONvyCfwQxIkUshLAOqN/k9EjGviPRmWTTe6aH2qySWKTVSw==,
+      }
+
+  "@types/node@18.19.130":
+    resolution:
+      {
+        integrity: sha512-GRaXQx6jGfL8sKfaIDD6OupbIHBr9jv7Jnaml9tB7l4v068PAOXqfcujMMo5PhbIs6ggR1XODELqahT2R8v0fg==,
+      }
+
   "@types/node@20.19.43":
     resolution:
       {
@@ -4071,6 +4164,13 @@ packages:
         integrity: sha512-A51o8ymO5PpqlWNnBP9ZHPXDIpuMtTLlGSjN7la4US+LJzoUMyhwjA5QXlm39JexgwHKW4Xjs8Z2d3dLCXOeuA==,
       }
 
+  abort-controller@3.0.0:
+    resolution:
+      {
+        integrity: sha512-h8lQ8tacZYnR3vNQTgibj+tODHI5/+l06Au2Pcriv/Gmet0eaj4TwWH41sO9wnHDiQsEj19q0drzdWdeAHtweg==,
+      }
+    engines: { node: ">=6.5" }
+
   accepts@2.0.0:
     resolution:
       {
@@ -4094,6 +4194,13 @@ packages:
     engines: { node: ">=0.4.0" }
     hasBin: true
 
+  agentkeepalive@4.6.0:
+    resolution:
+      {
+        integrity: sha512-kja8j7PjmncONqaTsB8fQ+wE2mSU2DJ9D4XKoJ5PFWIdRMa6SLSN1ff4mOr4jCbfRSsxR4keIiySJU0N9T5hIQ==,
+      }
+    engines: { node: ">= 8.0.0" }
+
   ajv-formats@2.1.1:
     resolution:
       {
@@ -4250,6 +4357,12 @@ packages:
       "@astrojs/markdown-remark":
         optional: true
 
+  asynckit@0.4.0:
+    resolution:
+      {
+        integrity: sha512-Oei9OH4tRh0YqU3GxhX79dM/mwVgvbZJaSNaRk+bshkj0S5cfHcgYakreBjrHwatXKbz+IoIdYLxrKim2MjW0Q==,
+      }
+
   atomically@1.7.0:
     resolution:
       {
@@ -4555,6 +4668,13 @@ packages:
       }
     engines: { node: ">=12.5.0" }
 
+  combined-stream@1.0.8:
+    resolution:
+      {
+        integrity: sha512-FQN4MRfuJeHf7cBbBMJFXhKSDq+2kAArBlmRBvcvFE5BB1HZKXtSFASDhdlz9zOYwxh8lDdnvmMOe/+5cdoEdg==,
+      }
+    engines: { node: ">= 0.8" }
+
   comma-separated-tokens@2.0.3:
     resolution:
       {
@@ -5137,6 +5257,13 @@ packages:
         integrity: sha512-AGrQ4QSgssa1NGmWmLPqN5NY2KajF5MqxetNEO+o0n3ZwZZeTmt7bBnvzHWrmkZFxGgr4HdyFgelzgi06otLuQ==,
       }
 
+  delayed-stream@1.0.0:
+    resolution:
+      {
+        integrity: sha512-ZySD7Nf91aLB0RxL4KGrKHBXl7Eds1DAmEdcoVawXnLD7SDhpNgtuII2aAkg7a7QS41jxPSZ17p4VdGnMHk3MQ==,
+      }
+    engines: { node: ">=0.4.0" }
+
   depd@2.0.0:
     resolution:
       {
@@ -5360,6 +5487,13 @@ packages:
       }
     engines: { node: ">= 0.4" }
 
+  es-set-tostringtag@2.1.0:
+    resolution:
+      {
+        integrity: sha512-j6vWzfrGVfyXxge+O0x5sh6cvxAog0a/4Rdd2K36zCMV5eJ+/+tOAngRO8cODMNWbVRdVlmGZQL2YS3yR8bIUA==,
+      }
+    engines: { node: ">= 0.4" }
+
   es-toolkit@1.49.0:
     resolution:
       {
@@ -5554,6 +5688,13 @@ packages:
       }
     engines: { node: ">= 0.6" }
 
+  event-target-shim@5.0.1:
+    resolution:
+      {
+        integrity: sha512-i/2XbnSz/uxRCU6+NdVJgKWDTM427+MqYbkQzD321DuCQJUqOuJKIA0IM2+W2xtYHdKOmZ4dR6fExsd4SXL+WQ==,
+      }
+    engines: { node: ">=6" }
+
   eventemitter3@4.0.7:
     resolution:
       {
@@ -5685,6 +5826,13 @@ packages:
         integrity: sha512-7F2Fl+TjRSenLqlU3UjSH0iyqopqoZIu7eZVpEirP2g1GtWa2G/ecEmBdgz31+Mxr+ELclgg6sokpSFIQiZ02Q==,
       }
 
+  fast-xml-parser@4.5.7:
+    resolution:
+      {
+        integrity: sha512-a6Qh1RMCNbSrU1+sAyAAZH3rTe+OaWJbNZIq0S+ifZciUUOQtlVxBJwoTUE2bYhysmG/RYyI5WJFIKdBahJdrQ==,
+      }
+    hasBin: true
+
   fastq@1.20.1:
     resolution:
       {
@@ -5784,6 +5932,26 @@ packages:
       }
     engines: { node: ">=20" }
 
+  form-data-encoder@1.7.2:
+    resolution:
+      {
+        integrity: sha512-qfqtYan3rxrnCk1VYaA4H+Ms9xdpPqvLZa6xmMgFvhO32x7/3J/ExcTd6qpxM0vH2GdMI+poehyBZvqfMTto8A==,
+      }
+
+  form-data@4.0.6:
+    resolution:
+      {
+        integrity: sha512-vKatAh4SlVfgbv+YtmhiRjhEMJsYpsG1Y2rMQtR+SVSbytsSD1YGzDIcrAJmdFec88u/+VoGmxnl+80gL1tRCQ==,
+      }
+    engines: { node: ">= 6" }
+
+  formdata-node@4.4.1:
+    resolution:
+      {
+        integrity: sha512-0iirZp3uVDjVGt9p49aTaqjk84TrglENEDuqfdlZQ1roC9CWlPk6Avf8EEnZNcAqPonwkG35x4n3ww/1THYAeQ==,
+      }
+    engines: { node: ">= 12.20" }
+
   forwarded@0.2.0:
     resolution:
       {
@@ -5946,6 +6114,13 @@ packages:
         integrity: sha512-RbJ5/jmFcNNCcDV5o9eTnBLJ/HszWV0P73bc+Ff4nS/rJj+YaS6IGyiOL0VoBYX+l1Wrl3k63h/KrH+nhJ0XvQ==,
       }
 
+  groq-sdk@1.3.0:
+    resolution:
+      {
+        integrity: sha512-mvgUIpAxlk/VxWIoliHx4R+Ha78Bd/g0t24OFjCXtdLbNiY1rW4h9AcznKBwFho7K/Nq732ZSjxMYkNb1xeCFg==,
+      }
+    hasBin: true
+
   h3@1.15.11:
     resolution:
       {
@@ -5965,6 +6140,13 @@ packages:
       }
     engines: { node: ">= 0.4" }
 
+  has-tostringtag@1.0.2:
+    resolution:
+      {
+        integrity: sha512-NqADB8VjPFLM2V0VvHUewwwsw0ZWBaIdgo+ieHtK3hasLz4qeCRjYcqfB6AQrBggRKppKF8L52/VqdVsO47Dlw==,
+      }
+    engines: { node: ">= 0.4" }
+
   hasown@2.0.4:
     resolution:
       {
@@ -6144,6 +6326,12 @@ packages:
       }
     engines: { node: ">=18.18.0" }
 
+  humanize-ms@1.2.1:
+    resolution:
+      {
+        integrity: sha512-Fl70vYtsAFb/C06PTS9dZBo7ihau+Tu/DNCk/OyHhea07S+aeMWpFFkUaXRa8fI+ScZbEI8dfSxwY7gxZ9SAVQ==,
+      }
+
   i18next@26.3.4:
     resolution:
       {
@@ -6504,6 +6692,13 @@ packages:
         integrity: sha512-xyFwyhro/JEof6Ghe2iz2NcXoj2sloNsWr/XsERDK/oiPCfaNhl5ONfp+jQdAZRQQ0IJWNzH9zIZF7li91kh2w==,
       }
 
+  json-schema-to-ts@3.1.1:
+    resolution:
+      {
+        integrity: sha512-+DWg8jCJG2TEnpy7kOm/7/AxaYoaRbjVB4LFZLySZlWn8exGs3A4OLJR966cVvU26N7X9TWxl+Jsw7dzAqKT6g==,
+      }
+    engines: { node: ">=16" }
+
   json-schema-traverse@0.4.1:
     resolution:
       {
@@ -7222,6 +7417,13 @@ packages:
       }
     engines: { node: ">=8.6" }
 
+  mime-db@1.52.0:
+    resolution:
+      {
+        integrity: sha512-sPU4uV7dYlvtWJxwwxHD0PuihVNiE7TyAbQ5SWxDCB9mUYvOgroQOwYQQOKPJ8CIbE+1ETVlOoK1UC2nU3gYvg==,
+      }
+    engines: { node: ">= 0.6" }
+
   mime-db@1.54.0:
     resolution:
       {
@@ -7229,6 +7431,13 @@ packages:
       }
     engines: { node: ">= 0.6" }
 
+  mime-types@2.1.35:
+    resolution:
+      {
+        integrity: sha512-ZDY+bPm5zTTF+YpCrAU9nK0UgICYPT0QtT1NZWFv4s++TNkcgVaT0g6+4R2uI4MjQjzysHB1zxuWL50hzaeXiw==,
+      }
+    engines: { node: ">= 0.6" }
+
   mime-types@3.0.2:
     resolution:
       {
@@ -7382,12 +7591,32 @@ packages:
       }
     engines: { node: ">=10" }
 
+  node-domexception@1.0.0:
+    resolution:
+      {
+        integrity: sha512-/jKZoMpw0F8GRwl4/eLROPA3cfcXtLApP0QzLmUT/HuPCZWyB7IY9ZrMeKw2O/nFIqPQB3PVM9aYm0F312AXDQ==,
+      }
+    engines: { node: ">=10.5.0" }
+    deprecated: Use your platform's native DOMException instead
+
   node-fetch-native@1.6.7:
     resolution:
       {
         integrity: sha512-g9yhqoedzIUm0nTnTqAQvueMPVOuIY16bqgAJJC8XOOubYFNwz6IER9qs0Gq2Xd0+CecCKFjtdDTMA4u4xG06Q==,
       }
 
+  node-fetch@2.7.0:
+    resolution:
+      {
+        integrity: sha512-c4FRfUm/dbcWZ7U+1Wq0AwCyFL+3nt2bEw05wfxSz+DWpWsitgmSgYmy2dQdWyKC1694ELPqMs/YzUSNozLt8A==,
+      }
+    engines: { node: 4.x || >=6.0.0 }
+    peerDependencies:
+      encoding: ^0.1.0
+    peerDependenciesMeta:
+      encoding:
+        optional: true
+
   node-mock-http@1.0.4:
     resolution:
       {
@@ -7468,6 +7697,12 @@ packages:
         integrity: sha512-RdR9FQrFwNBNXAr4GixM8YaRZRJ5PUWbKYbE5eOsrwAjJW0q2REGcf79oYPsLyskQCZG1PLN+S/K1V00joZAoQ==,
       }
 
+  ollama@0.5.18:
+    resolution:
+      {
+        integrity: sha512-lTFqTf9bo7Cd3hpF6CviBe/DEhewjoZYd9N/uCe7O20qYTvGqrNOFOBDj3lbZgFWHUgDv5EeyusYxsZSLS8nvg==,
+      }
+
   on-finished@2.4.1:
     resolution:
       {
@@ -7521,6 +7756,21 @@ packages:
       }
     engines: { node: ">=12" }
 
+  openai@4.104.0:
+    resolution:
+      {
+        integrity: sha512-p99EFNsA/yX6UhVO93f5kJsDRLAg+CTA2RBqdHK4RtK8u5IJw32Hyb2dTGKbnnFmnuoBv5r7Z2CURI9sGZpSuA==,
+      }
+    hasBin: true
+    peerDependencies:
+      ws: ^8.18.0
+      zod: ^3.23.8
+    peerDependenciesMeta:
+      ws:
+        optional: true
+      zod:
+        optional: true
+
   openai@6.45.0:
     resolution:
       {
@@ -8624,6 +8874,12 @@ packages:
       }
     engines: { node: ">=0.10.0" }
 
+  strnum@1.1.2:
+    resolution:
+      {
+        integrity: sha512-vrN+B7DBIoTTZjnPNewwhx6cBA/H+IS7rfW68n7XxC1y7uoiGQBxaKzqucGUgavX15dJgiGztLJ8vxuEzwqBdA==,
+      }
+
   style-to-js@1.1.21:
     resolution:
       {
@@ -8782,6 +9038,12 @@ packages:
       }
     engines: { node: ">=0.6" }
 
+  tr46@0.0.3:
+    resolution:
+      {
+        integrity: sha512-N3WMsuqV66lT30CrXNbEjx4GEwlow3v6rr4mCcv6prnfwhS01rkgyFdjPNBYd9br7LpXV1+Emh01fHnq2Gdgrw==,
+      }
+
   trim-lines@3.0.1:
     resolution:
       {
@@ -8794,6 +9056,12 @@ packages:
         integrity: sha512-tmMpK00BjZiUyVyvrBK7knerNgmgvcV/KLVyuma/SC+TQN167GrMRciANTz09+k3zW8L8t60jWO1GpfkZdjTaw==,
       }
 
+  ts-algebra@2.0.0:
+    resolution:
+      {
+        integrity: sha512-FPAhNPFMrkwz76P7cdjdmiShwMynZYN6SgOujD1urY4oNm80Ou9oMdmbR45LotcKOXoy7wSmHkRFE6Mxbrhefw==,
+      }
+
   ts-api-utils@2.5.0:
     resolution:
       {
@@ -8885,6 +9153,12 @@ packages:
         integrity: sha512-Ql87qFHB3s/De2ClA9e0gsnS6zXG27SkTiSJwjCc9MebbfapQfuPzumMIUMi38ezPZVNFcHI9sUIepeQfw8J8Q==,
       }
 
+  undici-types@5.26.5:
+    resolution:
+      {
+        integrity: sha512-JlCMO+ehdEIKqlFxk6IfVoAUVmgz7cU7zD/h9XZ0qzeosSHmUJVOzSQvvYSYWXkFXC+IfLKSIffhv0sVZup6pA==,
+      }
+
   undici-types@6.21.0:
     resolution:
       {
@@ -9136,6 +9410,14 @@ packages:
         integrity: sha512-EPD5q1uXyFxJpCrLnCc1nHnq3gOa6DZBocAIiI2TaSCA7VCJ1UJDMagCzIkXNsUYfD1daK//LTEQ8xiIbrHtcw==,
       }
 
+  uuid@10.0.0:
+    resolution:
+      {
+        integrity: sha512-8XkAphELsDnEGrDxUOHB3RGvXz6TeuYSGEZBOjtTtPm2lwhGBjLgOzLHB63IUWfBpNucQjND6d3AOudO+H3RWQ==,
+      }
+    deprecated: uuid@10 and below is no longer supported.  For ESM codebases, update to uuid@latest.  For CommonJS codebases, use uuid@11 (but be aware this version will likely be deprecated in 2028).
+    hasBin: true
+
   uuid@14.0.1:
     resolution:
       {
@@ -9282,6 +9564,31 @@ packages:
         integrity: sha512-bKr1DkiNa2krS7qxNtdrtHAmzuYGFQLiQ13TsorsdT6ULTkPLKuu5+GsFpDlg6JFjUTwX2DyhMPG2be8uPrqsQ==,
       }
 
+  web-streams-polyfill@4.0.0-beta.3:
+    resolution:
+      {
+        integrity: sha512-QW95TCTaHmsYfHDybGMwO5IJIM93I/6vTRk+daHTWFPhwh+C8Cg7j7XyKrwrj8Ib6vYXe0ocYNrmzY4xAAN6ug==,
+      }
+    engines: { node: ">= 14" }
+
+  webidl-conversions@3.0.1:
+    resolution:
+      {
+        integrity: sha512-2JAn3z8AR6rjK8Sm8orRC0h/bcl/DqL7tRPdGZ4I1CjdF+EaMLmYxBHyXuKL849eucPFhvBoxMsflfOb8kxaeQ==,
+      }
+
+  whatwg-fetch@3.6.20:
+    resolution:
+      {
+        integrity: sha512-EqhiFU6daOA8kpjOWTL0olhVOF3i7OrFzSYiGsEMB8GcXS+RrzauAERX65xMeNWVqxA6HXH2m69Z9LaKKdisfg==,
+      }
+
+  whatwg-url@5.0.0:
+    resolution:
+      {
+        integrity: sha512-saE57nupxk6v3HY35+jzBwYa0rKSy0XR8JSxZPwgLr7ys0IBzhGviA1/TUGJLmSVqs8pb9AnvICXEuOHLprYTw==,
+      }
+
   which@2.0.2:
     resolution:
       {
@@ -9455,6 +9762,12 @@ snapshots:
       package-manager-detector: 1.7.0
       tinyexec: 1.2.4
 
+  "@anthropic-ai/sdk@0.65.0(zod@4.4.3)":
+    dependencies:
+      json-schema-to-ts: 3.1.1
+    optionalDependencies:
+      zod: 4.4.3
+
   "@astrojs/compiler-binding-darwin-arm64@0.3.0":
     optional: true
 
@@ -10335,6 +10648,13 @@ snapshots:
       "@jridgewell/resolve-uri": 3.1.2
       "@jridgewell/sourcemap-codec": 1.5.5
 
+  "@langchain/anthropic@0.3.34(zod@4.4.3)":
+    dependencies:
+      "@anthropic-ai/sdk": 0.65.0(zod@4.4.3)
+      fast-xml-parser: 4.5.7
+    transitivePeerDependencies:
+      - zod
+
   "@langchain/core@1.2.1(openai@6.45.0(ws@8.21.0)(zod@4.4.3))(ws@8.21.0)":
     dependencies:
       "@cfworker/json-schema": 4.1.1
@@ -10351,14 +10671,41 @@ snapshots:
       - openai
       - ws
 
+  "@langchain/deepseek@1.1.5(ws@8.21.0)":
+    dependencies:
+      "@langchain/openai": 1.5.5(ws@8.21.0)
+    transitivePeerDependencies:
+      - "@aws-sdk/credential-provider-node"
+      - "@smithy/hash-node"
+      - "@smithy/signature-v4"
+      - ws
+
   "@langchain/google-genai@2.2.0(@langchain/core@1.2.1(openai@6.45.0(ws@8.21.0)(zod@4.4.3))(ws@8.21.0))":
     dependencies:
       "@google/generative-ai": 0.24.1
       "@langchain/core": 1.2.1(openai@6.45.0(ws@8.21.0)(zod@4.4.3))(ws@8.21.0)
 
-  "@langchain/openai@1.5.3(@langchain/core@1.2.1(openai@6.45.0(ws@8.21.0)(zod@4.4.3))(ws@8.21.0))(ws@8.21.0)":
+  "@langchain/groq@1.3.1":
+    dependencies:
+      groq-sdk: 1.3.0
+
+  "@langchain/ollama@0.2.4":
+    dependencies:
+      ollama: 0.5.18
+      uuid: 10.0.0
+
+  "@langchain/openai@0.3.17(ws@8.21.0)":
+    dependencies:
+      js-tiktoken: 1.0.21
+      openai: 4.104.0(ws@8.21.0)(zod@3.25.76)
+      zod: 3.25.76
+      zod-to-json-schema: 3.25.2(zod@3.25.76)
+    transitivePeerDependencies:
+      - encoding
+      - ws
+
+  "@langchain/openai@1.5.3(ws@8.21.0)":
     dependencies:
-      "@langchain/core": 1.2.1(openai@6.45.0(ws@8.21.0)(zod@4.4.3))(ws@8.21.0)
       js-tiktoken: 1.0.21
       openai: 6.45.0(ws@8.21.0)(zod@4.4.3)
       zod: 4.4.3
@@ -10368,10 +10715,20 @@ snapshots:
       - "@smithy/signature-v4"
       - ws
 
-  "@langchain/openrouter@0.4.3(@langchain/core@1.2.1(openai@6.45.0(ws@8.21.0)(zod@4.4.3))(ws@8.21.0))(ws@8.21.0)(zod@4.4.3)":
+  "@langchain/openai@1.5.5(ws@8.21.0)":
     dependencies:
-      "@langchain/core": 1.2.1(openai@6.45.0(ws@8.21.0)(zod@4.4.3))(ws@8.21.0)
-      "@langchain/openai": 1.5.3(@langchain/core@1.2.1(openai@6.45.0(ws@8.21.0)(zod@4.4.3))(ws@8.21.0))(ws@8.21.0)
+      js-tiktoken: 1.0.21
+      openai: 6.45.0(ws@8.21.0)(zod@4.4.3)
+      zod: 4.4.3
+    transitivePeerDependencies:
+      - "@aws-sdk/credential-provider-node"
+      - "@smithy/hash-node"
+      - "@smithy/signature-v4"
+      - ws
+
+  "@langchain/openrouter@0.4.3(ws@8.21.0)(zod@4.4.3)":
+    dependencies:
+      "@langchain/openai": 1.5.3(ws@8.21.0)
       eventsource-parser: 3.1.0
       openai: 6.45.0(ws@8.21.0)(zod@4.4.3)
     transitivePeerDependencies:
@@ -11622,6 +11979,15 @@ snapshots:
     dependencies:
       "@types/unist": 3.0.3
 
+  "@types/node-fetch@2.6.13":
+    dependencies:
+      "@types/node": 26.1.0
+      form-data: 4.0.6
+
+  "@types/node@18.19.130":
+    dependencies:
+      undici-types: 5.26.5
+
   "@types/node@20.19.43":
     dependencies:
       undici-types: 6.21.0
@@ -11794,6 +12160,10 @@ snapshots:
       convert-source-map: 2.0.0
       tinyrainbow: 3.1.0
 
+  abort-controller@3.0.0:
+    dependencies:
+      event-target-shim: 5.0.1
+
   accepts@2.0.0:
     dependencies:
       mime-types: 3.0.2
@@ -11805,6 +12175,10 @@ snapshots:
 
   acorn@8.17.0: {}
 
+  agentkeepalive@4.6.0:
+    dependencies:
+      humanize-ms: 1.2.1
+
   ajv-formats@2.1.1(ajv@8.20.0):
     optionalDependencies:
       ajv: 8.20.0
@@ -11969,6 +12343,8 @@ snapshots:
       - uploadthing
       - yaml
 
+  asynckit@0.4.0: {}
+
   atomically@1.7.0: {}
 
   autoprefixer@10.5.2(postcss@8.5.16):
@@ -12128,6 +12504,10 @@ snapshots:
       color-convert: 2.0.1
       color-string: 1.9.1
 
+  combined-stream@1.0.8:
+    dependencies:
+      delayed-stream: 1.0.0
+
   comma-separated-tokens@2.0.3: {}
 
   commander@11.1.0: {}
@@ -12458,6 +12838,8 @@ snapshots:
     dependencies:
       robust-predicates: 3.0.3
 
+  delayed-stream@1.0.0: {}
+
   depd@2.0.0: {}
 
   dequal@2.0.3: {}
@@ -12558,6 +12940,13 @@ snapshots:
     dependencies:
       es-errors: 1.3.0
 
+  es-set-tostringtag@2.1.0:
+    dependencies:
+      es-errors: 1.3.0
+      get-intrinsic: 1.3.0
+      has-tostringtag: 1.0.2
+      hasown: 2.0.4
+
   es-toolkit@1.49.0: {}
 
   esast-util-from-estree@2.0.0:
@@ -12720,6 +13109,8 @@ snapshots:
 
   etag@1.8.1: {}
 
+  event-target-shim@5.0.1: {}
+
   eventemitter3@4.0.7: {}
 
   eventemitter3@5.0.4: {}
@@ -12834,6 +13225,10 @@ snapshots:
     dependencies:
       fast-string-width: 3.0.2
 
+  fast-xml-parser@4.5.7:
+    dependencies:
+      strnum: 1.1.2
+
   fastq@1.20.1:
     dependencies:
       reusify: 1.1.0
@@ -12893,6 +13288,21 @@ snapshots:
     dependencies:
       tiny-inflate: 1.0.3
 
+  form-data-encoder@1.7.2: {}
+
+  form-data@4.0.6:
+    dependencies:
+      asynckit: 0.4.0
+      combined-stream: 1.0.8
+      es-set-tostringtag: 2.1.0
+      hasown: 2.0.4
+      mime-types: 2.1.35
+
+  formdata-node@4.4.1:
+    dependencies:
+      node-domexception: 1.0.0
+      web-streams-polyfill: 4.0.0-beta.3
+
   forwarded@0.2.0: {}
 
   fraction.js@5.3.4: {}
@@ -12969,6 +13379,8 @@ snapshots:
 
   graceful-fs@4.2.11: {}
 
+  groq-sdk@1.3.0: {}
+
   h3@1.15.11:
     dependencies:
       cookie-es: 1.2.3
@@ -12985,6 +13397,10 @@ snapshots:
 
   has-symbols@1.1.0: {}
 
+  has-tostringtag@1.0.2:
+    dependencies:
+      has-symbols: 1.1.0
+
   hasown@2.0.4:
     dependencies:
       function-bind: 1.1.2
@@ -13200,6 +13616,10 @@ snapshots:
 
   human-signals@8.0.1: {}
 
+  humanize-ms@1.2.1:
+    dependencies:
+      ms: 2.1.3
+
   i18next@26.3.4(typescript@6.0.3):
     optionalDependencies:
       typescript: 6.0.3
@@ -13330,6 +13750,11 @@ snapshots:
 
   json-parse-even-better-errors@2.3.1: {}
 
+  json-schema-to-ts@3.1.1:
+    dependencies:
+      "@babel/runtime": 7.29.7
+      ts-algebra: 2.0.0
+
   json-schema-traverse@0.4.1: {}
 
   json-schema-traverse@1.0.0: {}
@@ -13977,8 +14402,14 @@ snapshots:
       braces: 3.0.3
       picomatch: 2.3.2
 
+  mime-db@1.52.0: {}
+
   mime-db@1.54.0: {}
 
+  mime-types@2.1.35:
+    dependencies:
+      mime-db: 1.52.0
+
   mime-types@3.0.2:
     dependencies:
       mime-db: 1.54.0
@@ -14059,8 +14490,14 @@ snapshots:
     dependencies:
       semver: 7.8.5
 
+  node-domexception@1.0.0: {}
+
   node-fetch-native@1.6.7: {}
 
+  node-fetch@2.7.0:
+    dependencies:
+      whatwg-url: 5.0.0
+
   node-mock-http@1.0.4: {}
 
   node-releases@2.0.51: {}
@@ -14096,6 +14533,10 @@ snapshots:
 
   ohash@2.0.11: {}
 
+  ollama@0.5.18:
+    dependencies:
+      whatwg-fetch: 3.6.20
+
   on-finished@2.4.1:
     dependencies:
       ee-first: 1.1.1
@@ -14135,6 +14576,21 @@ snapshots:
       is-docker: 2.2.1
       is-wsl: 2.2.0
 
+  openai@4.104.0(ws@8.21.0)(zod@3.25.76):
+    dependencies:
+      "@types/node": 18.19.130
+      "@types/node-fetch": 2.6.13
+      abort-controller: 3.0.0
+      agentkeepalive: 4.6.0
+      form-data-encoder: 1.7.2
+      formdata-node: 4.4.1
+      node-fetch: 2.7.0
+    optionalDependencies:
+      ws: 8.21.0
+      zod: 3.25.76
+    transitivePeerDependencies:
+      - encoding
+
   openai@6.45.0(ws@8.21.0)(zod@4.4.3):
     optionalDependencies:
       ws: 8.21.0
@@ -15013,6 +15469,8 @@ snapshots:
 
   strip-json-comments@2.0.1: {}
 
+  strnum@1.1.2: {}
+
   style-to-js@1.1.21:
     dependencies:
       style-to-object: 1.0.14
@@ -15092,10 +15550,14 @@ snapshots:
 
   toidentifier@1.0.1: {}
 
+  tr46@0.0.3: {}
+
   trim-lines@3.0.1: {}
 
   trough@2.2.0: {}
 
+  ts-algebra@2.0.0: {}
+
   ts-api-utils@2.5.0(typescript@6.0.3):
     dependencies:
       typescript: 6.0.3
@@ -15148,6 +15610,8 @@ snapshots:
 
   uncrypto@0.1.3: {}
 
+  undici-types@5.26.5: {}
+
   undici-types@6.21.0: {}
 
   undici-types@7.18.2: {}
@@ -15272,6 +15736,8 @@ snapshots:
 
   util-deprecate@1.0.2: {}
 
+  uuid@10.0.0: {}
+
   uuid@14.0.1: {}
 
   validate-npm-package-name@7.0.2: {}
@@ -15352,6 +15818,17 @@ snapshots:
 
   web-namespaces@2.0.1: {}
 
+  web-streams-polyfill@4.0.0-beta.3: {}
+
+  webidl-conversions@3.0.1: {}
+
+  whatwg-fetch@3.6.20: {}
+
+  whatwg-url@5.0.0:
+    dependencies:
+      tr46: 0.0.3
+      webidl-conversions: 3.0.1
+
   which@2.0.2:
     dependencies:
       isexe: 2.0.0

web/docs/src/content/docs/architecture/llm-providers.md

@@ -71,14 +71,27 @@ If no specific provider instance is mapped to a task, the task automatically rou
 
 ---
 
-## Automatic Bootstrapping (Environment Key Fallback)
+## CLI Setup & Seeding
 
-To maintain backwards-compatibility and support headless runs, live evaluation suites, and automated unit tests without requiring database pre-configuration, the config manager supports **self-bootstrapping**:
+Rather than automatically bootstrapping from environment variables at runtime, which adds runtime complexity, you can quickly seed the database using the CLI setup tool:
 
-1. When any database connection is initialized via the provider manager, if `data/settings.db` contains **0 registered keys**, it checks the process environment for `GOOGLE_API_KEY` and `OPENROUTER_API_KEY`.
-2. If `process.env.GOOGLE_API_KEY` is present, it automatically creates, saves, and activates a default provider instance (`Gemini (Env)`) in `settings.db`.
-3. If `process.env.OPENROUTER_API_KEY` is present, it automatically creates and saves a default provider instance (`OpenRouter (Env)`) in `settings.db`.
-4. If database write locks occur (e.g., during high-concurrency Vitest test suites), the system seamlessly returns a temporary in-memory `LLMProviderInstance` (`Gemini (Env Fallback)` or `OpenRouter (Env Fallback)`) to keep execution fluent and error-free.
+### Seeding All Environment-Variable Providers
+
+```bash
+pnpm setup-provider --all
+```
+
+This command auto-detects and inserts provider instances into `data/settings.db` for any registered providers whose corresponding environment variables (such as `GOOGLE_API_KEY`, `OPENAI_API_KEY`, etc.) are defined.
+
+### Creating a Specific Provider Instance
+
+```bash
+pnpm setup-provider --provider google-genai --key YOUR_API_KEY [--name "My Gemini"] [--model gemini-2.5-flash] [--type generative] [--max-context 32768] [--endpoint url]
+```
+
+### Environment Variable Fallback
+
+If the database contains no active provider instances, the LLM providers (e.g. `GeminiProvider`, `OpenAIProvider`, etc.) will fall back directly to reading their keys from environment variables (e.g. `GOOGLE_API_KEY`, `OPENAI_API_KEY`) via `resolveCredentials`.
 
 ---