Add support for Llama API with LiteLLM

Co-Authored-By: Joe Moura <joao@crewai.com>

src/crewai/cli/constants.py

@@ -103,6 +103,17 @@ ENV_VARS = {
             "key_name": "SAMBANOVA_API_KEY",
         }
     ],
+    "meta-llama": [
+        {
+            "prompt": "Enter your LLAMA API key (press Enter to skip)",
+            "key_name": "LLAMA_API_KEY",
+        },
+        {
+            "prompt": "Enter your LLAMA API base URL (press Enter to skip)",
+            "key_name": "LLAMA_API_BASE",
+            "default": "https://api.llama.com/compat/v1",
+        },
+    ],
 }
 
 
@@ -119,6 +130,7 @@ PROVIDERS = [
     "azure",
     "cerebras",
     "sambanova",
+    "meta-llama",
 ]
 
 MODELS = {
@@ -314,6 +326,12 @@ MODELS = {
         "sambanova/Meta-Llama-3.2-3B-Instruct",
         "sambanova/Meta-Llama-3.2-1B-Instruct",
     ],
+    "meta-llama": [
+        "meta-llama/llama-4-scout-17b-16e-instruct-fp8",
+        "meta-llama/llama-4-maverick-17b-128e-instruct-fp8",
+        "meta-llama/llama-3.3-70b-instruct",
+        "meta-llama/llama-3.3-8b-instruct",
+    ],
 }
 
 DEFAULT_LLM_MODEL = "gpt-4o-mini"
@@ -321,4 +339,4 @@ DEFAULT_LLM_MODEL = "gpt-4o-mini"
 JSON_URL = "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json"
 
 
-LITELLM_PARAMS = ["api_key", "api_base", "api_version"]
+LITELLM_PARAMS = ["api_key", "api_base", "api_version", "llama_api_key", "llama_api_base"]

src/crewai/llm.py

@@ -120,6 +120,9 @@ LLM_CONTEXT_WINDOW_SIZES = {
     "mixtral-8x7b-32768": 32768,
     "llama-3.3-70b-versatile": 128000,
     "llama-3.3-70b-instruct": 128000,
+    "llama-4-scout-17b-16e-instruct-fp8": 128000,
+    "llama-4-maverick-17b-128e-instruct-fp8": 128000,
+    "llama-3.3-8b-instruct": 128000,
     # sambanova
     "Meta-Llama-3.3-70B-Instruct": 131072,
     "QwQ-32B-Preview": 8192,

tests/llm_test.py

@@ -618,3 +618,26 @@ def test_handle_streaming_tool_calls_no_tools(mock_emit):
         expected_completed_llm_call=1,
         expected_final_chunk_result=response,
     )
+
+
+def test_llama_api_support():
+    """Test that Llama API models are correctly configured."""
+    from crewai.cli.constants import MODELS, PROVIDERS, ENV_VARS, LITELLM_PARAMS
+    from crewai.llm import LLM_CONTEXT_WINDOW_SIZES
+    
+    assert "meta-llama" in PROVIDERS
+    
+    assert "meta-llama" in MODELS
+    assert len(MODELS["meta-llama"]) > 0
+    
+    assert "meta-llama" in ENV_VARS
+    assert any(env_var["key_name"] == "LLAMA_API_KEY" for env_var in ENV_VARS["meta-llama"])
+    assert any(env_var["key_name"] == "LLAMA_API_BASE" for env_var in ENV_VARS["meta-llama"])
+    
+    assert "llama_api_key" in LITELLM_PARAMS
+    assert "llama_api_base" in LITELLM_PARAMS
+    
+    for model in MODELS["meta-llama"]:
+        model_name = model.split("/")[-1]
+        assert any(model_name in key or key in model_name 
+                  for key in LLM_CONTEXT_WINDOW_SIZES.keys()), f"Context window size for {model} not found"