diff --git a/README.md b/README.md index d7b516c..f2bc0ba 100644 --- a/README.md +++ b/README.md @@ -290,6 +290,7 @@ nano .env # The file will contain, at least one should be set: # GEMINI_API_KEY=your-gemini-api-key-here # For Gemini models # OPENAI_API_KEY=your-openai-api-key-here # For O3 model +# XAI_API_KEY=your-xai-api-key-here # For Grok models # OPENROUTER_API_KEY=your-openrouter-key # For OpenRouter (see docs/custom_models.md) # DIAL_API_KEY=your-dial-api-key-here # For DIAL platform diff --git a/docs/advanced-usage.md b/docs/advanced-usage.md index 9383354..63856da 100644 --- a/docs/advanced-usage.md +++ b/docs/advanced-usage.md @@ -39,6 +39,9 @@ Regardless of your default configuration, you can specify models per request: | **`o3-mini`** | OpenAI | 200K tokens | Balanced speed/quality | Moderate complexity tasks | | **`o4-mini`** | OpenAI | 200K tokens | Latest reasoning model | Optimized for shorter contexts | | **`gpt4.1`** | OpenAI | 1M tokens | Latest GPT-4 with extended context | Large codebase analysis, comprehensive reviews | +| **`grok-4-latest`** | X.AI | 256K tokens | Latest flagship model with reasoning, vision | Complex analysis, reasoning tasks | +| **`grok-3`** | X.AI | 131K tokens | Advanced reasoning model | Deep analysis, complex problems | +| **`grok-3-fast`** | X.AI | 131K tokens | Higher performance variant | Fast responses with reasoning | | **`llama`** (Llama 3.2) | Custom/Local | 128K tokens | Local inference, privacy | On-device analysis, cost-free processing | | **Any model** | OpenRouter | Varies | Access to GPT-4, Claude, Llama, etc. | User-specified or based on task requirements | @@ -49,6 +52,8 @@ cloud models (expensive/powerful) AND local models (free/private) in the same co - **Gemini Models**: Support thinking modes (minimal to max), web search, 1M context - **O3 Models**: Excellent reasoning, systematic analysis, 200K context - **GPT-4.1**: Extended context window (1M tokens), general capabilities +- **Grok-4**: Extended thinking support, vision capabilities, 256K context +- **Grok-3 Models**: Advanced reasoning, 131K context ## Model Usage Restrictions diff --git a/docs/configuration.md b/docs/configuration.md index 473b6de..12e9d65 100644 --- a/docs/configuration.md +++ b/docs/configuration.md @@ -74,7 +74,8 @@ DEFAULT_MODEL=auto # Claude picks best model for each task (recommended) - **`o3`**: Strong logical reasoning (200K context) - **`o3-mini`**: Balanced speed/quality (200K context) - **`o4-mini`**: Latest reasoning model, optimized for shorter contexts -- **`grok`**: GROK-3 advanced reasoning (131K context) +- **`grok-3`**: GROK-3 advanced reasoning (131K context) +- **`grok-4-latest`**: GROK-4 latest flagship model (256K context) - **Custom models**: via OpenRouter or local APIs ### Thinking Mode Configuration @@ -107,7 +108,7 @@ OPENAI_ALLOWED_MODELS=o3-mini,o4-mini,mini GOOGLE_ALLOWED_MODELS=flash,pro # X.AI GROK model restrictions -XAI_ALLOWED_MODELS=grok-3,grok-3-fast +XAI_ALLOWED_MODELS=grok-3,grok-3-fast,grok-4-latest # OpenRouter model restrictions (affects models via custom provider) OPENROUTER_ALLOWED_MODELS=opus,sonnet,mistral @@ -128,9 +129,11 @@ OPENROUTER_ALLOWED_MODELS=opus,sonnet,mistral - `pro` (shorthand for Pro model) **X.AI GROK Models:** +- `grok-4-latest` (256K context, latest flagship model with reasoning, vision, and structured outputs) - `grok-3` (131K context, advanced reasoning) - `grok-3-fast` (131K context, higher performance) -- `grok` (shorthand for grok-3) +- `grok` (shorthand for grok-4-latest) +- `grok4` (shorthand for grok-4-latest) - `grok3` (shorthand for grok-3) - `grokfast` (shorthand for grok-3-fast) diff --git a/tests/test_xai_provider.py b/tests/test_xai_provider.py index 63cef94..7ed62b7 100644 --- a/tests/test_xai_provider.py +++ b/tests/test_xai_provider.py @@ -45,6 +45,9 @@ class TestXAIProvider: provider = XAIModelProvider("test-key") # Test valid models + assert provider.validate_model_name("grok-4") is True + assert provider.validate_model_name("grok-4-latest") is True + assert provider.validate_model_name("grok4") is True assert provider.validate_model_name("grok-3") is True assert provider.validate_model_name("grok-3-fast") is True assert provider.validate_model_name("grok") is True @@ -69,6 +72,7 @@ class TestXAIProvider: assert provider._resolve_model_name("grok3fast") == "grok-3-fast" # Test full name passthrough + assert provider._resolve_model_name("grok-4") == "grok-4" assert provider._resolve_model_name("grok-3") == "grok-3" assert provider._resolve_model_name("grok-3-fast") == "grok-3-fast" @@ -91,6 +95,27 @@ class TestXAIProvider: assert capabilities.temperature_constraint.max_temp == 2.0 assert capabilities.temperature_constraint.default_temp == 0.3 + def test_get_capabilities_grok4(self): + """Test getting model capabilities for GROK-4.""" + provider = XAIModelProvider("test-key") + + capabilities = provider.get_capabilities("grok-4") + assert capabilities.model_name == "grok-4" + assert capabilities.friendly_name == "X.AI (Grok 4)" + assert capabilities.context_window == 256_000 + assert capabilities.provider == ProviderType.XAI + assert capabilities.supports_extended_thinking is True + assert capabilities.supports_system_prompts is True + assert capabilities.supports_streaming is True + assert capabilities.supports_function_calling is True + assert capabilities.supports_json_mode is True + assert capabilities.supports_images is True + + # Test temperature range + assert capabilities.temperature_constraint.min_temp == 0.0 + assert capabilities.temperature_constraint.max_temp == 2.0 + assert capabilities.temperature_constraint.default_temp == 0.7 + def test_get_capabilities_grok3_fast(self): """Test getting model capabilities for GROK-3 Fast.""" provider = XAIModelProvider("test-key") @@ -121,9 +146,14 @@ class TestXAIProvider: provider.get_capabilities("invalid-model") def test_thinking_mode_support(self): - """Test X.AI model thinking mode support - grok-4 supports it, earlier models don't.""" + """Test thinking mode support for X.AI models.""" provider = XAIModelProvider("test-key") + # Grok-4 supports thinking mode + assert provider.supports_thinking_mode("grok-4") is True + assert provider.supports_thinking_mode("grok") is True # Resolves to grok-4 + + # Grok-3 models don't support thinking mode assert not provider.supports_thinking_mode("grok-3") assert not provider.supports_thinking_mode("grok-3-fast") assert provider.supports_thinking_mode("grok-4") # grok-4 supports thinking mode @@ -170,7 +200,10 @@ class TestXAIProvider: # Shorthand "grok" should be allowed (resolves to grok-4) assert provider.validate_model_name("grok") is True - # Full name "grok-3" should NOT be allowed (only shorthand "grok" is in restriction list) + # Full name "grok-4" should NOT be allowed (only shorthand "grok" is in restriction list) + assert provider.validate_model_name("grok-4") is False + + # "grok-3" should NOT be allowed (not in restriction list) assert provider.validate_model_name("grok-3") is False # "grok-3-fast" should be allowed (explicitly listed) @@ -179,7 +212,7 @@ class TestXAIProvider: # Shorthand "grokfast" should be allowed (resolves to grok-3-fast) assert provider.validate_model_name("grokfast") is True - @patch.dict(os.environ, {"XAI_ALLOWED_MODELS": "grok,grok-3"}) + @patch.dict(os.environ, {"XAI_ALLOWED_MODELS": "grok,grok-3,grok-4"}) def test_both_shorthand_and_full_name_allowed(self): """Test that both shorthand and full name can be allowed.""" # Clear cached restriction service @@ -190,8 +223,9 @@ class TestXAIProvider: provider = XAIModelProvider("test-key") # Both shorthand and full name should be allowed - assert provider.validate_model_name("grok") is True + assert provider.validate_model_name("grok") is True # Resolves to grok-4 assert provider.validate_model_name("grok-3") is True + assert provider.validate_model_name("grok-4") is True # Other models should not be allowed assert provider.validate_model_name("grok-3-fast") is False @@ -207,10 +241,12 @@ class TestXAIProvider: provider = XAIModelProvider("test-key") + assert provider.validate_model_name("grok-4") is True assert provider.validate_model_name("grok-3") is True assert provider.validate_model_name("grok-3-fast") is True assert provider.validate_model_name("grok") is True assert provider.validate_model_name("grokfast") is True + assert provider.validate_model_name("grok4") is True def test_friendly_name(self): """Test friendly name constant.""" @@ -232,14 +268,23 @@ class TestXAIProvider: # Check model configs have required fields from providers.base import ModelCapabilities + grok4_config = provider.SUPPORTED_MODELS["grok-4"] + assert isinstance(grok4_config, ModelCapabilities) + assert hasattr(grok4_config, "context_window") + assert hasattr(grok4_config, "supports_extended_thinking") + assert hasattr(grok4_config, "aliases") + assert grok4_config.context_window == 256_000 + assert grok4_config.supports_extended_thinking is True + + # Check aliases are correctly structured + assert "grok" in grok4_config.aliases + assert "grok-4" in grok4_config.aliases + assert "grok-4-latest" in grok4_config.aliases + assert "grok4" in grok4_config.aliases + grok3_config = provider.SUPPORTED_MODELS["grok-3"] - assert isinstance(grok3_config, ModelCapabilities) - assert hasattr(grok3_config, "context_window") - assert hasattr(grok3_config, "supports_extended_thinking") - assert hasattr(grok3_config, "aliases") assert grok3_config.context_window == 131_072 assert grok3_config.supports_extended_thinking is False - # Check aliases are correctly structured assert "grok3" in grok3_config.aliases # grok3 resolves to grok-3 @@ -268,7 +313,7 @@ class TestXAIProvider: mock_response.choices = [MagicMock()] mock_response.choices[0].message.content = "Test response" mock_response.choices[0].finish_reason = "stop" - mock_response.model = "grok-3" # API returns the resolved model name + mock_response.model = "grok-4" # API returns the resolved model name mock_response.id = "test-id" mock_response.created = 1234567890 mock_response.usage = MagicMock() @@ -322,6 +367,17 @@ class TestXAIProvider: provider = XAIModelProvider("test-key") + # Test grok4 -> grok-4 + mock_response.model = "grok-4" + provider.generate_content(prompt="Test", model_name="grok4", temperature=0.7) + call_kwargs = mock_client.chat.completions.create.call_args[1] + assert call_kwargs["model"] == "grok-4" + + # Test grok-4 -> grok-4 + provider.generate_content(prompt="Test", model_name="grok-4", temperature=0.7) + call_kwargs = mock_client.chat.completions.create.call_args[1] + assert call_kwargs["model"] == "grok-4" + # Test grok3 -> grok-3 mock_response.model = "grok-3" provider.generate_content(prompt="Test", model_name="grok3", temperature=0.7)