feat: Add support for separate LLM and embedding model endpoints

- Added LLM_ENDPOINT and EMBEDDING_ENDPOINT configuration options
- Updated GenericLLMProvider to handle custom base URLs
- Enhanced embedding initialization to use separate endpoints
- Improved configuration handling for both LLM and embedding providers

Author: sriramsowmithri9807

gpt_researcher/actions/agent_creator.py

@@ -32,13 +32,13 @@ async def choose_agent(
 
     try:
         response = await create_chat_completion(
-            model=cfg.smart_llm_model,
+            model=cfg.fast_llm_model,
             messages=[
                 {"role": "system", "content": f"{prompt_family.auto_agent_instructions()}"},
                 {"role": "user", "content": f"task: {query}"},
             ],
             temperature=0.15,
-            llm_provider=cfg.smart_llm_provider,
+            llm_provider=cfg.fast_llm_provider,
             llm_kwargs=cfg.llm_kwargs,
             cost_callback=cost_callback,
             **kwargs

gpt_researcher/config/config.py

@@ -62,8 +62,29 @@ def _set_llm_attributes(self) -> None:
         self.smart_llm_provider, self.smart_llm_model = self.parse_llm(self.smart_llm)
         self.strategic_llm_provider, self.strategic_llm_model = self.parse_llm(self.strategic_llm)
         self.reasoning_effort = self.parse_reasoning_effort(os.getenv("REASONING_EFFORT"))
+        
+        # Set base URLs for LLM and embedding endpoints
+        self.llm_base_url = getattr(self, 'llm_endpoint', 'http://localhost:8080/v1')
+        self.embedding_base_url = getattr(self, 'embedding_endpoint', 'http://localhost:8081/v1')
+        
+        # Update LLM kwargs with the appropriate base URL
+        if not self.llm_kwargs.get('base_url'):
+            self.llm_kwargs['base_url'] = self.llm_base_url
+            
+        # Update embedding kwargs with the appropriate base URL
+        if not self.embedding_kwargs.get('base_url'):
+            self.embedding_kwargs['base_url'] = self.embedding_base_url
 
     def _handle_deprecated_attributes(self) -> None:
+        # Handle environment variables for endpoints
+        if os.getenv("LLM_ENDPOINT"):
+            self.llm_base_url = os.environ["LLM_ENDPOINT"]
+            self.llm_kwargs['base_url'] = self.llm_base_url
+            
+        if os.getenv("EMBEDDING_ENDPOINT"):
+            self.embedding_base_url = os.environ["EMBEDDING_ENDPOINT"]
+            self.embedding_kwargs['base_url'] = self.embedding_base_url
+            
         if os.getenv("EMBEDDING_PROVIDER") is not None:
             warnings.warn(
                 "EMBEDDING_PROVIDER is deprecated and will be removed soon. Use EMBEDDING instead.",

gpt_researcher/config/variables/default.py

@@ -42,4 +42,6 @@
     "MCP_ALLOWED_ROOT_PATHS": [],  # List of allowed root paths for local file access
     "MCP_STRATEGY": "fast",  # MCP execution strategy: "fast", "deep", "disabled"
     "REASONING_EFFORT": "medium",
+    "LLM_ENDPOINT": "http://localhost:8080/v1",  # LLM endpoint
+    "EMBEDDING_ENDPOINT": "http://localhost:8081/v1",  # Embedding endpoint
 }

gpt_researcher/llm_provider/generic/base.py

@@ -90,10 +90,22 @@ def __init__(self, llm, chat_log: str | None = None,  verbose: bool = True):
         self.verbose = verbose
     @classmethod
     def from_provider(cls, provider: str, chat_log: str | None = None, verbose: bool=True, **kwargs: Any):
+        # Get the appropriate base URL from kwargs or environment
+        base_url = kwargs.pop('base_url', None) or os.environ.get(f"{provider.upper()}_BASE_URL")
+        
+        if base_url:
+            kwargs['base_url'] = base_url
+            if verbose:
+                print(f"Using {provider} endpoint: {base_url}")
+                
         if provider == "openai":
             _check_pkg("langchain_openai")
             from langchain_openai import ChatOpenAI
-
+            
+            # Handle custom endpoints for OpenAI-compatible APIs
+            if 'base_url' in kwargs and 'api_key' not in kwargs:
+                kwargs['api_key'] = 'dummy'  # Some APIs don't require a key
+                
             llm = ChatOpenAI(**kwargs)
         elif provider == "anthropic":
             _check_pkg("langchain_anthropic")
@@ -133,8 +145,13 @@ def from_provider(cls, provider: str, chat_log: str | None = None, verbose: bool
             _check_pkg("langchain_community")
             _check_pkg("langchain_ollama")
             from langchain_ollama import ChatOllama
-
-            llm = ChatOllama(base_url=os.environ["OLLAMA_BASE_URL"], **kwargs)
+            
+            # Use provided base_url or fall back to environment variable
+            base_url = kwargs.pop('base_url', os.environ.get("OLLAMA_BASE_URL"))
+            if base_url:
+                kwargs['base_url'] = base_url
+                
+            llm = ChatOllama(**kwargs)
         elif provider == "together":
             _check_pkg("langchain_together")
             from langchain_together import ChatTogether

gpt_researcher/memory/embeddings.py

@@ -27,105 +27,111 @@
 
 
 class Memory:
-    def __init__(self, embedding_provider: str, model: str, **embdding_kwargs: Any):
+    def __init__(self, embedding_provider: str, model: str, **embedding_kwargs: Any):
         _embeddings = None
+        
+        # Get base URL from kwargs or environment
+        base_url = embedding_kwargs.pop('base_url', None) or os.environ.get('EMBEDDING_ENDPOINT')
+        
         match embedding_provider:
-            case "custom":
+            case "custom" | "openai":
                 from langchain_openai import OpenAIEmbeddings
-
+                
+                # For custom endpoints, use a dummy key if none provided
+                api_key = os.getenv("OPENAI_API_KEY", "dummy")
+                if embedding_provider == "custom" and not base_url:
+                    base_url = os.getenv("OPENAI_BASE_URL", "http://localhost:1234/v1")
+                
                 _embeddings = OpenAIEmbeddings(
                     model=model,
-                    openai_api_key=os.getenv("OPENAI_API_KEY", "custom"),
-                    openai_api_base=os.getenv(
-                        "OPENAI_BASE_URL", "http://localhost:1234/v1"
-                    ),  # default for lmstudio
+                    openai_api_key=api_key,
+                    openai_api_base=base_url,
                     check_embedding_ctx_length=False,
-                    **embdding_kwargs,
-                )  # quick fix for lmstudio
-            case "openai":
-                from langchain_openai import OpenAIEmbeddings
-
-                _embeddings = OpenAIEmbeddings(model=model, **embdding_kwargs)
+                    **embedding_kwargs,
+                )
             case "azure_openai":
                 from langchain_openai import AzureOpenAIEmbeddings
 
                 _embeddings = AzureOpenAIEmbeddings(
                     model=model,
-                    azure_endpoint=os.environ["AZURE_OPENAI_ENDPOINT"],
-                    openai_api_key=os.environ["AZURE_OPENAI_API_KEY"],
-                    openai_api_version=os.environ["AZURE_OPENAI_API_VERSION"],
-                    **embdding_kwargs,
+                    openai_api_version=os.getenv("AZURE_OPENAI_API_VERSION", "2023-05-15"),
+                    azure_deployment=model,
+                    **embedding_kwargs,
                 )
             case "cohere":
                 from langchain_cohere import CohereEmbeddings
-
-                _embeddings = CohereEmbeddings(model=model, **embdding_kwargs)
+                _embeddings = CohereEmbeddings(model=model, **embedding_kwargs)
+                
             case "google_vertexai":
                 from langchain_google_vertexai import VertexAIEmbeddings
-
-                _embeddings = VertexAIEmbeddings(model=model, **embdding_kwargs)
+                _embeddings = VertexAIEmbeddings(model=model, **embedding_kwargs)
+                
             case "google_genai":
                 from langchain_google_genai import GoogleGenerativeAIEmbeddings
-
                 _embeddings = GoogleGenerativeAIEmbeddings(
-                    model=model, **embdding_kwargs
+                    model=model,
+                    **embedding_kwargs,
                 )
+                
             case "fireworks":
                 from langchain_fireworks import FireworksEmbeddings
-
-                _embeddings = FireworksEmbeddings(model=model, **embdding_kwargs)
+                _embeddings = FireworksEmbeddings(model=model, **embedding_kwargs)
+                
             case "gigachat":
                 from langchain_gigachat import GigaChatEmbeddings
-
-                _embeddings = GigaChatEmbeddings(model=model, **embdding_kwargs)
+                _embeddings = GigaChatEmbeddings(model=model, **embedding_kwargs)
+                
             case "ollama":
                 from langchain_ollama import OllamaEmbeddings
-
+                # Use provided base_url or fall back to environment variable
+                ollama_base = base_url or os.getenv("OLLAMA_BASE_URL", "http://localhost:11434")
                 _embeddings = OllamaEmbeddings(
                     model=model,
-                    base_url=os.environ["OLLAMA_BASE_URL"],
-                    **embdding_kwargs,
+                    base_url=ollama_base,
+                    **embedding_kwargs,
                 )
+                
             case "together":
                 from langchain_together import TogetherEmbeddings
-
-                _embeddings = TogetherEmbeddings(model=model, **embdding_kwargs)
+                _embeddings = TogetherEmbeddings(model=model, **embedding_kwargs)
+                
             case "mistralai":
                 from langchain_mistralai import MistralAIEmbeddings
-
-                _embeddings = MistralAIEmbeddings(model=model, **embdding_kwargs)
+                _embeddings = MistralAIEmbeddings(model=model, **embedding_kwargs)
+                
             case "huggingface":
                 from langchain_huggingface import HuggingFaceEmbeddings
-
-                _embeddings = HuggingFaceEmbeddings(model_name=model, **embdding_kwargs)
+                _embeddings = HuggingFaceEmbeddings(
+                    model_name=model, **embedding_kwargs
+                )
+                
             case "nomic":
                 from langchain_nomic import NomicEmbeddings
-
-                _embeddings = NomicEmbeddings(model=model, **embdding_kwargs)
+                _embeddings = NomicEmbeddings(model=model, **embedding_kwargs)
+                
             case "voyageai":
                 from langchain_voyageai import VoyageAIEmbeddings
-
                 _embeddings = VoyageAIEmbeddings(
-                    voyage_api_key=os.environ["VOYAGE_API_KEY"],
                     model=model,
-                    **embdding_kwargs,
+                    voyage_api_key=os.getenv("VOYAGE_API_KEY"),
+                    **embedding_kwargs,
                 )
+                
             case "dashscope":
                 from langchain_community.embeddings import DashScopeEmbeddings
-
-                _embeddings = DashScopeEmbeddings(model=model, **embdding_kwargs)
+                _embeddings = DashScopeEmbeddings(model=model, **embedding_kwargs)
+                
             case "bedrock":
                 from langchain_aws.embeddings import BedrockEmbeddings
-
-                _embeddings = BedrockEmbeddings(model_id=model, **embdding_kwargs)
+                _embeddings = BedrockEmbeddings(model_id=model, **embedding_kwargs)
+                
             case "aimlapi":
                 from langchain_openai import OpenAIEmbeddings
-
                 _embeddings = OpenAIEmbeddings(
                     model=model,
-                    openai_api_key=os.getenv("AIMLAPI_API_KEY"),
-                    openai_api_base=os.getenv("AIMLAPI_BASE_URL", "https://api.aimlapi.com/v1"),
-                    **embdding_kwargs,
+                    openai_api_key=os.getenv("OPENAI_API_KEY", "custom"),
+                    openai_api_base=base_url or os.getenv("AIMLAPI_BASE_URL"),
+                    **embedding_kwargs,
                 )
             case _:
                 raise Exception("Embedding not found.")