Merge pull request #38 from calesthio/feat/ollama-provider

Add Ollama provider for self-hosted LLM inference

.env.example

@@ -31,13 +31,15 @@ REFRESH_INTERVAL_MINUTES=15
 
 # === LLM Layer (optional) ===
 # Enables AI-enhanced trade ideas and breaking news Telegram alerts.
-# Provider options: anthropic | openai | gemini | codex | openrouter | minimax | mistral
+# Provider options: anthropic | openai | gemini | codex | openrouter | minimax | mistral | ollama
 LLM_PROVIDER=
-# Not needed for codex (uses ~/.codex/auth.json)
+# Not needed for codex (uses ~/.codex/auth.json) or ollama (local)
 LLM_API_KEY=
 # Optional override. Each provider has a sensible default:
-# anthropic: claude-sonnet-4-6 | openai: gpt-5.4 | gemini: gemini-3.1-pro | codex: gpt-5.3-codex | openrouter: openrouter/auto | minimax: MiniMax-M2.5
+# anthropic: claude-sonnet-4-6 | openai: gpt-5.4 | gemini: gemini-3.1-pro | codex: gpt-5.3-codex | openrouter: openrouter/auto | minimax: MiniMax-M2.5 | ollama: llama3.1:8b
 LLM_MODEL=
+# Ollama base URL (only needed if not using default http://localhost:11434)
+OLLAMA_BASE_URL=
 
 # === Telegram Alerts (optional, requires LLM) ===
 # Create a bot via @BotFather, get chat ID via @userinfobot

crucix.config.mjs

@@ -1,15 +1,16 @@
 // Crucix Configuration — all settings with env var overrides
 
-import './apis/utils/env.mjs'; // Load .env first
+import "./apis/utils/env.mjs"; // Load .env first
 
 export default {
   port: parseInt(process.env.PORT) || 3117,
   refreshIntervalMinutes: parseInt(process.env.REFRESH_INTERVAL_MINUTES) || 15,
 
   llm: {
-    provider: process.env.LLM_PROVIDER || null, // anthropic | openai | gemini | codex | openrouter | minimax | mistral
+    provider: process.env.LLM_PROVIDER || null, // anthropic | openai | gemini | codex | openrouter | minimax | mistral | ollama
     apiKey: process.env.LLM_API_KEY || null,
     model: process.env.LLM_MODEL || null,
+    baseUrl: process.env.OLLAMA_BASE_URL || null,
   },
 
   telegram: {
@@ -22,7 +23,7 @@ export default {
   discord: {
     botToken: process.env.DISCORD_BOT_TOKEN || null,
     channelId: process.env.DISCORD_CHANNEL_ID || null,
-    guildId: process.env.DISCORD_GUILD_ID || null,   // Server ID (for instant slash command registration)
+    guildId: process.env.DISCORD_GUILD_ID || null, // Server ID (for instant slash command registration)
     webhookUrl: process.env.DISCORD_WEBHOOK_URL || null, // Fallback: webhook-only alerts (no bot needed)
   },
 

lib/llm/index.mjs

@@ -1,21 +1,23 @@
 // LLM Factory — creates the configured provider or returns null
 
-import { AnthropicProvider } from './anthropic.mjs';
-import { OpenAIProvider } from './openai.mjs';
-import { OpenRouterProvider } from './openrouter.mjs';
-import { GeminiProvider } from './gemini.mjs';
-import { CodexProvider } from './codex.mjs';
-import { MiniMaxProvider } from './minimax.mjs';
-import { MistralProvider } from './mistral.mjs';
+import { AnthropicProvider } from "./anthropic.mjs";
+import { OpenAIProvider } from "./openai.mjs";
+import { OpenRouterProvider } from "./openrouter.mjs";
+import { GeminiProvider } from "./gemini.mjs";
+import { CodexProvider } from "./codex.mjs";
+import { MiniMaxProvider } from "./minimax.mjs";
+import { MistralProvider } from "./mistral.mjs";
+import { OllamaProvider } from "./ollama.mjs";
 
-export { LLMProvider } from './provider.mjs';
-export { AnthropicProvider } from './anthropic.mjs';
-export { OpenAIProvider } from './openai.mjs';
-export { OpenRouterProvider } from './openrouter.mjs';
-export { GeminiProvider } from './gemini.mjs';
-export { CodexProvider } from './codex.mjs';
-export { MiniMaxProvider } from './minimax.mjs';
-export { MistralProvider } from './mistral.mjs';
+export { LLMProvider } from "./provider.mjs";
+export { AnthropicProvider } from "./anthropic.mjs";
+export { OpenAIProvider } from "./openai.mjs";
+export { OpenRouterProvider } from "./openrouter.mjs";
+export { GeminiProvider } from "./gemini.mjs";
+export { CodexProvider } from "./codex.mjs";
+export { MiniMaxProvider } from "./minimax.mjs";
+export { MistralProvider } from "./mistral.mjs";
+export { OllamaProvider } from "./ollama.mjs";
 
 /**
  * Create an LLM provider based on config.
@@ -28,22 +30,26 @@ export function createLLMProvider(llmConfig) {
   const { provider, apiKey, model } = llmConfig;
 
   switch (provider.toLowerCase()) {
-    case 'anthropic':
+    case "anthropic":
       return new AnthropicProvider({ apiKey, model });
-    case 'openai':
+    case "openai":
       return new OpenAIProvider({ apiKey, model });
-    case 'openrouter':
+    case "openrouter":
       return new OpenRouterProvider({ apiKey, model });
-    case 'gemini':
+    case "gemini":
       return new GeminiProvider({ apiKey, model });
-    case 'codex':
+    case "codex":
       return new CodexProvider({ model });
-    case 'minimax':
+    case "minimax":
       return new MiniMaxProvider({ apiKey, model });
-    case 'mistral':
+    case "mistral":
       return new MistralProvider({ apiKey, model });
+    case "ollama":
+      return new OllamaProvider({ model, baseUrl: llmConfig.baseUrl });
     default:
-      console.warn(`[LLM] Unknown provider "${provider}". LLM features disabled.`);
+      console.warn(
+        `[LLM] Unknown provider "${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,
+    };
+  }
+}

test/llm-ollama-integration.test.mjs

@@ -0,0 +1,45 @@
+// Ollama provider — integration test (calls real Ollama instance)
+// Requires a running Ollama server with a model pulled
+// Run: OLLAMA_MODEL=llama3.1:8b node --test test/llm-ollama-integration.test.mjs
+
+import { describe, it } from 'node:test';
+import assert from 'node:assert/strict';
+import { OllamaProvider } from '../lib/llm/ollama.mjs';
+
+const BASE_URL = process.env.OLLAMA_BASE_URL || 'http://localhost:11434';
+const MODEL = process.env.OLLAMA_MODEL || 'llama3.1:8b';
+
+// Check if Ollama is reachable and the requested model is installed
+let ollamaAvailable = false;
+let skipReason = 'Ollama not reachable';
+try {
+  const res = await fetch(`${BASE_URL}/api/tags`, { signal: AbortSignal.timeout(3000) });
+  if (res.ok) {
+    const { models = [] } = await res.json();
+    const installed = models.some((m) => m.name === MODEL || m.name.startsWith(`${MODEL}:`));
+    if (installed) {
+      ollamaAvailable = true;
+    } else {
+      skipReason = `Model "${MODEL}" not installed (available: ${models.map((m) => m.name).join(', ') || 'none'})`;
+    }
+  }
+} catch { /* not available */ }
+
+describe('Ollama integration', { skip: !ollamaAvailable && skipReason }, () => {
+  it('should complete a prompt via local Ollama', async () => {
+    const provider = new OllamaProvider({ model: MODEL, baseUrl: BASE_URL });
+    assert.equal(provider.isConfigured, true);
+
+    const result = await provider.complete(
+      'You are a helpful assistant. Respond in exactly one sentence.',
+      'What is 2+2?',
+      { maxTokens: 128, timeout: 60000 }
+    );
+
+    assert.ok(result.text.length > 0, 'Response text should not be empty');
+    assert.ok(result.model, 'Should report model name');
+    console.log(`  Response: ${result.text}`);
+    console.log(`  Tokens: ${result.usage.inputTokens} in / ${result.usage.outputTokens} out`);
+    console.log(`  Model: ${result.model}`);
+  });
+});

test/llm-ollama.test.mjs

@@ -0,0 +1,170 @@
+// Ollama provider — unit tests
+// Uses Node.js built-in test runner (node:test) — no extra dependencies
+
+import { describe, it, mock, beforeEach } from 'node:test';
+import assert from 'node:assert/strict';
+import { OllamaProvider } from '../lib/llm/ollama.mjs';
+import { createLLMProvider } from '../lib/llm/index.mjs';
+
+// ─── Unit Tests ───
+
+describe('OllamaProvider', () => {
+  it('should set defaults correctly', () => {
+    const provider = new OllamaProvider({});
+    assert.equal(provider.name, 'ollama');
+    assert.equal(provider.model, 'llama3.1:8b');
+    assert.equal(provider.baseUrl, 'http://localhost:11434');
+    assert.equal(provider.isConfigured, true);
+  });
+
+  it('should accept custom model and base URL', () => {
+    const provider = new OllamaProvider({ model: 'qwen2.5:14b', baseUrl: 'http://192.168.1.10:11434' });
+    assert.equal(provider.model, 'qwen2.5:14b');
+    assert.equal(provider.baseUrl, 'http://192.168.1.10:11434');
+  });
+
+  it('should strip trailing slashes from base URL', () => {
+    const provider = new OllamaProvider({ baseUrl: 'http://localhost:11434/' });
+    assert.equal(provider.baseUrl, 'http://localhost:11434');
+  });
+
+  it('should throw on API error', async () => {
+    const provider = new OllamaProvider({});
+    const originalFetch = globalThis.fetch;
+    globalThis.fetch = mock.fn(() =>
+      Promise.resolve({ ok: false, status: 404, text: () => Promise.resolve('model not found') })
+    );
+    try {
+      await assert.rejects(
+        () => provider.complete('system', 'user'),
+        (err) => {
+          assert.match(err.message, /Ollama API 404/);
+          return true;
+        }
+      );
+    } finally {
+      globalThis.fetch = originalFetch;
+    }
+  });
+
+  it('should parse successful response', async () => {
+    const provider = new OllamaProvider({});
+    const mockResponse = {
+      choices: [{ message: { content: 'Hello from Ollama' } }],
+      usage: { prompt_tokens: 12, completion_tokens: 8 },
+      model: 'llama3.1:8b',
+    };
+    const originalFetch = globalThis.fetch;
+    globalThis.fetch = mock.fn(() =>
+      Promise.resolve({ ok: true, json: () => Promise.resolve(mockResponse) })
+    );
+    try {
+      const result = await provider.complete('You are helpful.', 'Say hello');
+      assert.equal(result.text, 'Hello from Ollama');
+      assert.equal(result.usage.inputTokens, 12);
+      assert.equal(result.usage.outputTokens, 8);
+      assert.equal(result.model, 'llama3.1:8b');
+    } finally {
+      globalThis.fetch = originalFetch;
+    }
+  });
+
+  it('should send correct request format', async () => {
+    const provider = new OllamaProvider({ model: 'qwen2.5:14b', baseUrl: 'http://myhost:11434' });
+    let capturedUrl, capturedOpts;
+    const originalFetch = globalThis.fetch;
+    globalThis.fetch = mock.fn((url, opts) => {
+      capturedUrl = url;
+      capturedOpts = opts;
+      return Promise.resolve({
+        ok: true,
+        json: () => Promise.resolve({
+          choices: [{ message: { content: 'ok' } }],
+          usage: { prompt_tokens: 1, completion_tokens: 1 },
+          model: 'qwen2.5:14b',
+        }),
+      });
+    });
+    try {
+      await provider.complete('system prompt', 'user message', { maxTokens: 2048 });
+      assert.equal(capturedUrl, 'http://myhost:11434/v1/chat/completions');
+      assert.equal(capturedOpts.method, 'POST');
+      const headers = capturedOpts.headers;
+      assert.equal(headers['Content-Type'], 'application/json');
+      assert.equal(headers['Authorization'], undefined);
+      const body = JSON.parse(capturedOpts.body);
+      assert.equal(body.model, 'qwen2.5:14b');
+      assert.equal(body.max_tokens, 2048);
+      assert.equal(body.messages[0].role, 'system');
+      assert.equal(body.messages[0].content, 'system prompt');
+      assert.equal(body.messages[1].role, 'user');
+      assert.equal(body.messages[1].content, 'user message');
+    } finally {
+      globalThis.fetch = originalFetch;
+    }
+  });
+
+  it('should handle empty response gracefully', async () => {
+    const provider = new OllamaProvider({});
+    const originalFetch = globalThis.fetch;
+    globalThis.fetch = mock.fn(() =>
+      Promise.resolve({
+        ok: true,
+        json: () => Promise.resolve({ choices: [], usage: {} }),
+      })
+    );
+    try {
+      const result = await provider.complete('sys', 'user');
+      assert.equal(result.text, '');
+      assert.equal(result.usage.inputTokens, 0);
+      assert.equal(result.usage.outputTokens, 0);
+    } finally {
+      globalThis.fetch = originalFetch;
+    }
+  });
+
+  it('should use longer default timeout than cloud providers', async () => {
+    const provider = new OllamaProvider({});
+    let capturedOpts;
+    const originalFetch = globalThis.fetch;
+    globalThis.fetch = mock.fn((url, opts) => {
+      capturedOpts = opts;
+      return Promise.resolve({
+        ok: true,
+        json: () => Promise.resolve({
+          choices: [{ message: { content: 'ok' } }],
+          usage: { prompt_tokens: 1, completion_tokens: 1 },
+        }),
+      });
+    });
+    try {
+      await provider.complete('sys', 'user');
+      assert.ok(capturedOpts.signal, 'Should have an abort signal');
+    } finally {
+      globalThis.fetch = originalFetch;
+    }
+  });
+});
+
+// ─── Factory Tests ───
+
+describe('createLLMProvider — ollama', () => {
+  it('should create OllamaProvider for provider=ollama', () => {
+    const provider = createLLMProvider({ provider: 'ollama', apiKey: null, model: null });
+    assert.ok(provider instanceof OllamaProvider);
+    assert.equal(provider.name, 'ollama');
+    assert.equal(provider.isConfigured, true);
+  });
+
+  it('should be case-insensitive', () => {
+    const provider = createLLMProvider({ provider: 'Ollama', apiKey: null, model: null });
+    assert.ok(provider instanceof OllamaProvider);
+  });
+
+  it('should pass baseUrl from config', () => {
+    const provider = createLLMProvider({ provider: 'ollama', apiKey: null, model: 'mistral:7b', baseUrl: 'http://gpu-box:11434' });
+    assert.ok(provider instanceof OllamaProvider);
+    assert.equal(provider.baseUrl, 'http://gpu-box:11434');
+    assert.equal(provider.model, 'mistral:7b');
+  });
+});