Add Ollama provider for self-hosted LLM inference

Adds LLM_PROVIDER=ollama for fully local, zero-cost inference
via Ollama's OpenAI-compatible API. No API key required.
Configurable base URL via OLLAMA_BASE_URL env var.

lib/llm/index.mjs

@@ -6,6 +6,7 @@ import { OpenRouterProvider } from './openrouter.mjs';
 import { GeminiProvider } from './gemini.mjs';
 import { CodexProvider } from './codex.mjs';
 import { MiniMaxProvider } from './minimax.mjs';
+import { OllamaProvider } from './ollama.mjs';
 
 export { LLMProvider } from './provider.mjs';
 export { AnthropicProvider } from './anthropic.mjs';
@@ -14,6 +15,7 @@ export { OpenRouterProvider } from './openrouter.mjs';
 export { GeminiProvider } from './gemini.mjs';
 export { CodexProvider } from './codex.mjs';
 export { MiniMaxProvider } from './minimax.mjs';
+export { OllamaProvider } from './ollama.mjs';
 
 /**
  * Create an LLM provider based on config.
@@ -38,6 +40,8 @@ export function createLLMProvider(llmConfig) {
       return new CodexProvider({ model });
     case 'minimax':
       return new MiniMaxProvider({ apiKey, model });
+    case 'ollama':
+      return new OllamaProvider({ model, baseUrl: llmConfig.baseUrl });
     default:
       console.warn(`[LLM] Unknown provider "${provider}". LLM features disabled.`);
       return null;

lib/llm/ollama.mjs

@@ -0,0 +1,49 @@
+// Ollama Provider — raw fetch, no SDK
+// Uses Ollama's OpenAI-compatible Chat Completions API
+// No API key required — fully local inference
+
+import { LLMProvider } from './provider.mjs';
+
+export class OllamaProvider extends LLMProvider {
+  constructor(config) {
+    super(config);
+    this.name = 'ollama';
+    this.baseUrl = (config.baseUrl || 'http://localhost:11434').replace(/\/+$/, '');
+    this.model = config.model || 'llama3.1:8b';
+  }
+
+  get isConfigured() { return !!this.model; }
+
+  async complete(systemPrompt, userMessage, opts = {}) {
+    const res = await fetch(`${this.baseUrl}/v1/chat/completions`, {
+      method: 'POST',
+      headers: { 'Content-Type': 'application/json' },
+      body: JSON.stringify({
+        model: this.model,
+        max_tokens: opts.maxTokens || 4096,
+        messages: [
+          { role: 'system', content: systemPrompt },
+          { role: 'user', content: userMessage },
+        ],
+      }),
+      signal: AbortSignal.timeout(opts.timeout || 120000),
+    });
+
+    if (!res.ok) {
+      const err = await res.text().catch(() => '');
+      throw new Error(`Ollama API ${res.status}: ${err.substring(0, 200)}`);
+    }
+
+    const data = await res.json();
+    const text = data.choices?.[0]?.message?.content || '';
+
+    return {
+      text,
+      usage: {
+        inputTokens: data.usage?.prompt_tokens || 0,
+        outputTokens: data.usage?.completion_tokens || 0,
+      },
+      model: data.model || this.model,
+    };
+  }
+}