Multi-LLM-Backend Support

README.md

@@ -45,7 +45,7 @@ This two-phase approach ensures that the solutions are not just generic LLM resp
 
 - **140 Thinking Models**: A comprehensive library of thinking models, from SWOT analysis to second-order thinking.
 - **Two-Phase Query Processing**: Enhances LLM responses with structured problem-solving methodologies.
-- **OpenAI-Compatible API**: Integrates with any OpenAI-compatible LLM API.
+- **Multi-LLM-Backend Support**: Integrates with multiple LLM providers, including OpenAI and Google Gemini.
 - **CLI & Web Interfaces**: Access the system through a command-line interface or a modern web application.
 - **Real-time Updates**: Get live feedback during query processing via WebSockets.
 - **Model Browser**: Explore, search, and filter thinking models through the web UI.
@@ -60,7 +60,7 @@ This two-phase approach ensures that the solutions are not just generic LLM resp
 
 - Python 3.8+
 - `pip` for package management
-- An OpenAI-compatible LLM API endpoint
+- An API endpoint and key for an OpenAI-compatible service or Google Gemini.
 
 ### Installation
 
@@ -79,9 +79,13 @@ This two-phase approach ensures that the solutions are not just generic LLM resp
 
 3. **Configure your LLM API:**
 
-   Create a `.env` file in the project root and add your API credentials:
+   Create a `.env` file in the project root. Below are examples for configuring OpenAI and Gemini providers.
 
+   **For OpenAI-Compatible APIs:**
    ```env
+   # Set the provider to openai (this is the default)
+   LLM_PROVIDER=openai
+
    # Required
    LLM_API_URL=https://your-llm-api-endpoint.com
 
@@ -90,6 +94,18 @@ This two-phase approach ensures that the solutions are not just generic LLM resp
    LLM_MODEL_NAME=gpt-3.5-turbo
    ```
 
+   **For Google Gemini:**
+   ```env
+   # Set the provider to gemini
+   LLM_PROVIDER=gemini
+
+   # Required - your Gemini API key
+   GEMINI_API_KEY=your-gemini-api-key
+
+   # Optional - specify a Gemini model
+   LLM_MODEL_NAME=gemini-1.5-flash
+   ```
+
 ---
 
 ## Usage

config.py

@@ -18,6 +18,8 @@ class Config:
     # LLM API Settings
     llm_api_url: Optional[str] = None
     llm_api_key: Optional[str] = None
+    gemini_api_key: Optional[str] = None
+    llm_provider: str = "openai"
     llm_model_name: str = "gpt-3.5-turbo"
     llm_temperature: float = 0.7
     llm_max_tokens: int = 2000
@@ -53,6 +55,8 @@ def from_env(cls) -> 'Config':
             # LLM API Settings
             llm_api_url=os.getenv('LLM_API_URL'),
             llm_api_key=os.getenv('LLM_API_KEY'),
+            gemini_api_key=os.getenv('GEMINI_API_KEY'),
+            llm_provider=os.getenv('LLM_PROVIDER', 'openai'),
             llm_model_name=os.getenv('LLM_MODEL_NAME', 'gpt-3.5-turbo'),
             llm_temperature=float(os.getenv('LLM_TEMPERATURE', '0.7')),
             llm_max_tokens=int(os.getenv('LLM_MAX_TOKENS', '2000')),

requirements.txt

@@ -1,6 +1,7 @@
 # Core dependencies
 requests>=2.28.0
 dataclasses-json>=0.5.7
+google-generativeai>=0.8.5
 
 # CLI dependencies
 click>=8.0.0
@@ -9,7 +10,7 @@ rich>=13.0.0
 
 # Web dependencies
 fastapi>=0.100.0
-uvicorn>=0.23.0
+uvicorn[standard]>=0.23.0
 jinja2>=3.0.0
 python-multipart>=0.0.5
 

src/cli/main.py

@@ -29,7 +29,7 @@
 sys.path.insert(0, str(Path(__file__).parent.parent))
 
 from core.model_parser import ModelParser
-from core.llm_client import LLMClient, LLMConfig
+from core.llm_client import get_llm_client
 from core.query_processor import QueryProcessor, QueryResult
 
 # Initialize Rich console for pretty output
@@ -64,44 +64,27 @@ def from_env(cls):
 def setup_processor(config: CLIConfig) -> Optional[QueryProcessor]:
     """Initialize the query processor with given configuration"""
     try:
-        # Load models
         if config.verbose:
             console.print(f"[blue]Loading thinking models from {config.models_dir}...[/blue]")
-        
+
         model_parser = ModelParser(config.models_dir)
         models = model_parser.load_all_models()
-        
+
         if config.verbose:
             console.print(f"[green]✓ Loaded {len(models)} thinking models[/green]")
-
-        # Setup LLM client
-        if not config.api_url:
-            console.print("[red]Error: LLM_API_URL environment variable must be set[/red]")
-            return None
-
-        llm_config = LLMConfig(
-            api_url=config.api_url,
-            api_key=config.api_key,
-            model_name=config.model_name,
-            temperature=config.temperature,
-            max_tokens=config.max_tokens
-        )
-
-        llm_client = LLMClient(llm_config)
-
-        # Test connection if in verbose mode
+
+        # Get LLM client from factory
+        llm_client = get_llm_client()
+
         if config.verbose:
             console.print("[blue]Testing LLM connection...[/blue]")
             if llm_client.test_connection():
                 console.print("[green]✓ LLM connection successful[/green]")
             else:
                 console.print("[yellow]⚠ LLM connection test failed, but continuing anyway[/yellow]")
-
-        # Create query processor
-        processor = QueryProcessor(model_parser, llm_client)
-
-        return processor
-
+
+        return QueryProcessor(model_parser, llm_client)
+
     except Exception as e:
         console.print(f"[red]Error initializing processor: {str(e)}[/red]")
         return None