diff options
Diffstat (limited to 'backend/app/services/llm.py')
| -rw-r--r-- | backend/app/services/llm.py | 22 |
1 files changed, 11 insertions, 11 deletions
diff --git a/backend/app/services/llm.py b/backend/app/services/llm.py index 96b0514..87b9155 100644 --- a/backend/app/services/llm.py +++ b/backend/app/services/llm.py @@ -4,18 +4,18 @@ import openai import google.generativeai as genai from app.schemas import LLMConfig, Message, Role, Context -# Simple in-memory cache for clients to avoid re-initializing constantly +# Cache OpenAI clients by API key to avoid re-initializing constantly # In a real app, use dependency injection or singletons -_openai_client = None +_openai_clients: dict[str, openai.AsyncOpenAI] = {} def get_openai_client(api_key: str = None): - global _openai_client + global _openai_clients key = api_key or os.getenv("OPENAI_API_KEY") if not key: raise ValueError("OpenAI API Key not found") - if not _openai_client: - _openai_client = openai.AsyncOpenAI(api_key=key) - return _openai_client + if key not in _openai_clients: + _openai_clients[key] = openai.AsyncOpenAI(api_key=key) + return _openai_clients[key] def configure_google(api_key: str = None): key = api_key or os.getenv("GOOGLE_API_KEY") @@ -345,12 +345,12 @@ async def llm_streamer( yield f"Error calling LLM: {str(e)}" -async def generate_title(user_prompt: str, response: str) -> str: +async def generate_title(user_prompt: str, response: str, api_key: str = None) -> str: """ Generate a short title (3-4 words) for a Q-A pair using gpt-5-nano. Uses Responses API (required for gpt-5 series), synchronous mode (no background). """ - client = get_openai_client() + client = get_openai_client(api_key) instructions = """TASK: Extract a short topic title from the given Q&A. Do NOT answer the question - only extract the topic. @@ -413,7 +413,7 @@ Q: "What's the weather in NYC?" -> "NYC Weather\"""" return "New Question" -async def summarize_content(content: str, model: str) -> str: +async def summarize_content(content: str, model: str, openai_api_key: str = None, gemini_api_key: str = None) -> str: """ Summarize the given content using the specified model. Supports both OpenAI and Gemini models. @@ -434,7 +434,7 @@ Output only the summary, no preamble.""" from google.genai import types import os - key = os.getenv("GOOGLE_API_KEY") + key = gemini_api_key or os.getenv("GOOGLE_API_KEY") if not key: return "Error: Google API Key not found" @@ -456,7 +456,7 @@ Output only the summary, no preamble.""" else: # Use OpenAI - client = get_openai_client() + client = get_openai_client(openai_api_key) # Check if model needs Responses API responses_api_models = [ |