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;