feat: Add Venice AI (VVV) support

.env.example

@@ -117,6 +117,10 @@ ANTHROPIC_API_KEY=sk-ant-your_anthropic_api_key_here
 # Get from: https://aistudio.google.com/app/apikey
 GEMINI_API_KEY=your_gemini_api_key_here
 
+# Venice AI - Uncensored models with private inference (VVV token)
+# Get from: https://venice.ai/settings/api
+VENICE_API_KEY=your_venice_api_key_here
+
 # ============================================================================
 # CONTENT & MEDIA
 # ============================================================================

tools/multi-model/README.md

@@ -1,7 +1,7 @@
 # Multi-Model AI Toolkit
 
-> Leverage OpenAI, Anthropic, and Google Gemini together for superior code
-> quality.
+> Leverage OpenAI, Anthropic, Google Gemini, and Venice AI together for superior
+> code quality.
 
 **Part of [Claude Code Sidekick](../../README.md)**
 
@@ -16,7 +16,9 @@
   keywords
 - **💰 Cost Optimization** - Use cheap models for simple tasks, reserve power
   for complex ones
-- **🔧 Flexible Configuration** - Works with just one API key or all three
+- **🔧 Flexible Configuration** - Works with just one API key or all four
+- **🔓 Venice AI Integration** - Access uncensored models with private inference
+  via [VVV token](https://venice.ai/token)
 
 ## File Structure
 
@@ -50,6 +52,7 @@ pnpm install
 OPENAI_API_KEY=sk-...
 ANTHROPIC_API_KEY=sk-ant-...
 GEMINI_API_KEY=...
+VENICE_API_KEY=...  # Optional: Venice AI for uncensored models
 ```
 
 ## Configuration
@@ -58,11 +61,20 @@ GEMINI_API_KEY=...
 
 The toolkit automatically adapts based on your available API keys:
 
-| API Keys Available | Mode         | Behavior                           |
-| ------------------ | ------------ | ---------------------------------- |
-| All three          | Multi-model  | Consensus review across all models |
-| Two keys           | Multi-model  | Consensus with 2 models            |
-| One key            | Single-model | Direct review with one model       |
+| API Keys Available | Mode         | Behavior                       |
+| ------------------ | ------------ | ------------------------------ |
+| 3+ keys            | Multi-model  | Consensus review across models |
+| 2 keys             | Multi-model  | Consensus with 2 models        |
+| 1 key              | Single-model | Direct review with one model   |
+
+### Supported Providers
+
+| Provider  | Env Variable        | Models                               |
+| --------- | ------------------- | ------------------------------------ |
+| OpenAI    | `OPENAI_API_KEY`    | GPT-4o, GPT-4o-mini                  |
+| Anthropic | `ANTHROPIC_API_KEY` | Claude Sonnet, Claude Haiku          |
+| Google    | `GEMINI_API_KEY`    | Gemini Pro, Gemini Flash             |
+| Venice AI | `VENICE_API_KEY`    | Llama 3.3 70B, DeepSeek R1, Qwen 2.5 |
 
 **To explicitly disable multi-model mode** (even if you have multiple keys):
 
@@ -90,11 +102,41 @@ ANTHROPIC_API_KEY=sk-ant-...
 
 # Option 3: Gemini only
 GEMINI_API_KEY=...
+
+# Option 4: Venice AI only (uncensored models)
+VENICE_API_KEY=...
 ```
 
 The toolkit will automatically use the best available model for your
 configuration.
 
+### Venice AI (VVV Token)
+
+[Venice AI](https://venice.ai) provides uncensored AI models with private
+inference. Powered by the [VVV token](https://venice.ai/token).
+
+Available models:
+
+- `llama-3.3-70b` - Llama 3.3 70B (fast, capable)
+- `llama-3.2-3b` - Llama 3.2 3B (lightweight, cheap)
+- `deepseek-r1-671b` - DeepSeek R1 671B (reasoning, most capable)
+- `deepseek-r1-llama-70b` - DeepSeek R1 Llama 70B
+- `dolphin-2.9.2-qwen2-72b` - Dolphin (uncensored)
+- `qwen-2.5-72b` - Qwen 2.5 72B
+
+```javascript
+import { MODELS, complete } from '@claude-sidekick/multi-model';
+
+// Use Venice models directly
+const result = await complete(
+  MODELS.VENICE_LLAMA_70B,
+  'Explain quantum computing'
+);
+
+// Or use venice/ prefix
+const result2 = await complete('venice/llama-3.3-70b', prompt);
+```
+
 ## CLI Tools
 
 ### Multi-Model Code Review

tools/multi-model/index.js

@@ -30,8 +30,10 @@ export {
   getOpenAI,
   getAnthropic,
   getGemini,
+  getVenice,
   MODELS,
   MODEL_COSTS,
+  VENICE_MODELS,
   // Configuration helpers
   isMultiModelEnabled,
   getAvailableProviders,
@@ -40,6 +42,7 @@ export {
   getBestModel,
   hasProvider,
   filterAvailableModels,
+  getProviderFromModel,
 } from './lib/clients.js';
 
 // Code Review

tools/multi-model/lib/clients.js

@@ -16,6 +16,7 @@ config();
 let openaiClient = null;
 let anthropicClient = null;
 let geminiClient = null;
+let veniceClient = null;
 
 /**
  * Check if multi-model mode is enabled
@@ -34,6 +35,7 @@ export function getAvailableProviders() {
   if (process.env.OPENAI_API_KEY) providers.push('openai');
   if (process.env.ANTHROPIC_API_KEY) providers.push('anthropic');
   if (process.env.GEMINI_API_KEY || process.env.GEMENI_API_KEY) providers.push('gemini');
+  if (process.env.VENICE_API_KEY) providers.push('venice');
   return providers;
 }
 
@@ -88,6 +90,24 @@ export function getGemini() {
   return geminiClient;
 }
 
+/**
+ * Get Venice AI client (OpenAI-compatible)
+ * Venice provides uncensored models with private inference
+ * https://venice.ai
+ */
+export function getVenice() {
+  if (!veniceClient) {
+    if (!process.env.VENICE_API_KEY) {
+      throw new Error('VENICE_API_KEY not set in environment');
+    }
+    veniceClient = new OpenAI({
+      apiKey: process.env.VENICE_API_KEY,
+      baseURL: 'https://api.venice.ai/api/v1',
+    });
+  }
+  return veniceClient;
+}
+
 /**
  * Unified completion interface
  */
@@ -147,6 +167,29 @@ export async function complete(model, prompt, options = {}) {
     };
   }
 
+  // Venice AI models (OpenAI-compatible API)
+  if (model.startsWith('venice/') || VENICE_MODELS.includes(model)) {
+    const client = getVenice();
+    const messages = [];
+    if (systemPrompt) messages.push({ role: 'system', content: systemPrompt });
+    messages.push({ role: 'user', content: prompt });
+
+    // Strip venice/ prefix if present
+    const veniceModel = model.replace('venice/', '');
+
+    const response = await client.chat.completions.create({
+      model: veniceModel,
+      messages,
+      max_tokens: maxTokens,
+      temperature,
+    });
+    return {
+      content: response.choices[0].message.content,
+      model,
+      usage: response.usage,
+    };
+  }
+
   throw new Error(`Unknown model: ${model}`);
 }
 
@@ -174,6 +217,16 @@ export async function embedBatch(texts, model = 'text-embedding-3-small') {
   return response.data.map((d) => d.embedding);
 }
 
+// Venice AI model identifiers
+export const VENICE_MODELS = [
+  'llama-3.3-70b',
+  'llama-3.2-3b',
+  'deepseek-r1-671b',
+  'deepseek-r1-llama-70b',
+  'dolphin-2.9.2-qwen2-72b',
+  'qwen-2.5-72b',
+];
+
 export const MODELS = {
   // OpenAI
   GPT4: 'gpt-4-turbo-preview',
@@ -188,6 +241,14 @@ export const MODELS = {
   // Google
   GEMINI_PRO: 'gemini-1.5-pro',
   GEMINI_FLASH: 'gemini-2.0-flash',
+
+  // Venice AI (uncensored, private inference)
+  VENICE_LLAMA_70B: 'llama-3.3-70b',
+  VENICE_LLAMA_3B: 'llama-3.2-3b',
+  VENICE_DEEPSEEK_R1: 'deepseek-r1-671b',
+  VENICE_DEEPSEEK_LLAMA: 'deepseek-r1-llama-70b',
+  VENICE_DOLPHIN: 'dolphin-2.9.2-qwen2-72b',
+  VENICE_QWEN: 'qwen-2.5-72b',
 };
 
 export const MODEL_COSTS = {
@@ -199,6 +260,13 @@ export const MODEL_COSTS = {
   'claude-3-5-haiku-20241022': [0.8, 4],
   'gemini-1.5-pro': [3.5, 10.5],
   'gemini-2.0-flash': [0.075, 0.3],
+  // Venice AI (VVV token) - prices approximate
+  'llama-3.3-70b': [0.5, 0.75],
+  'llama-3.2-3b': [0.1, 0.15],
+  'deepseek-r1-671b': [2.0, 4.0],
+  'deepseek-r1-llama-70b': [0.5, 0.75],
+  'dolphin-2.9.2-qwen2-72b': [0.5, 0.75],
+  'qwen-2.5-72b': [0.5, 0.75],
 };
 
 /**
@@ -208,6 +276,7 @@ export function getProviderFromModel(model) {
   if (model.startsWith('gpt-')) return 'openai';
   if (model.startsWith('claude-')) return 'anthropic';
   if (model.startsWith('gemini-')) return 'gemini';
+  if (model.startsWith('venice/') || VENICE_MODELS.includes(model)) return 'venice';
   return null;
 }
 
@@ -225,6 +294,9 @@ export function getAvailableModels() {
   if (hasProvider('gemini')) {
     models.push(MODELS.GEMINI_PRO, MODELS.GEMINI_FLASH);
   }
+  if (hasProvider('venice')) {
+    models.push(MODELS.VENICE_LLAMA_70B, MODELS.VENICE_LLAMA_3B, MODELS.VENICE_DEEPSEEK_R1);
+  }
   return models;
 }
 
@@ -234,6 +306,7 @@ export function getAvailableModels() {
 export function getDefaultModel() {
   // Prefer cheaper/faster models as default
   if (hasProvider('gemini')) return MODELS.GEMINI_FLASH;
+  if (hasProvider('venice')) return MODELS.VENICE_LLAMA_3B;
   if (hasProvider('openai')) return MODELS.GPT4O_MINI;
   if (hasProvider('anthropic')) return MODELS.CLAUDE_HAIKU;
   throw new Error('No API keys configured');
@@ -245,6 +318,7 @@ export function getDefaultModel() {
 export function getBestModel() {
   if (hasProvider('anthropic')) return MODELS.CLAUDE_SONNET;
   if (hasProvider('openai')) return MODELS.GPT4O;
+  if (hasProvider('venice')) return MODELS.VENICE_DEEPSEEK_R1;
   if (hasProvider('gemini')) return MODELS.GEMINI_PRO;
   throw new Error('No API keys configured');
 }