Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/karpathy/llm-council/llms.txt

Use this file to discover all available pages before exploring further.

LLM Council intentionally keeps its configuration minimal and centralized. Almost everything you need to change lives in two files: backend/config.py for model and storage settings, and .env in the project root for secrets. The sections below explain each option, where it lives, and what to watch out for when changing it.

​
backend/config.py

This file is the single source of truth for which models sit on the council, which model acts as Chairman, where conversations are saved, and which OpenRouter endpoint is used. Open it in any editor to make changes.
backend/config.py
"""Configuration for the LLM Council."""

import os
from dotenv import load_dotenv

load_dotenv()

# OpenRouter API key
OPENROUTER_API_KEY = os.getenv("OPENROUTER_API_KEY")

# Council members - list of OpenRouter model identifiers
COUNCIL_MODELS = [
    "openai/gpt-5.1",
    "google/gemini-3-pro-preview",
    "anthropic/claude-sonnet-4.5",
    "x-ai/grok-4",
]

# Chairman model - synthesizes final response
CHAIRMAN_MODEL = "google/gemini-3-pro-preview"

# OpenRouter API endpoint
OPENROUTER_API_URL = "https://openrouter.ai/api/v1/chat/completions"

# Data directory for conversation storage
DATA_DIR = "data/conversations"
COUNCIL_MODELS and CHAIRMAN_MODEL are hardcoded in config.py and are read once at startup. After changing either value you must restart the backend for the new configuration to take effect. There is currently no way to update council membership through the UI at runtime.

​
Configuration Variables

​
COUNCIL_MODELS
list[str]
required
A Python list of OpenRouter model identifier strings. Every model in this list participates in Stage 1 (answering the query) and Stage 2 (peer-reviewing the other answers). The number of council members directly determines how many parallel API requests are made per message and how rich the cross-review will be.The default council is:
  • openai/gpt-5.1
  • google/gemini-3-pro-preview
  • anthropic/claude-sonnet-4.5
  • x-ai/grok-4
​
CHAIRMAN_MODEL
str
required
The OpenRouter model identifier for the Chairman — the model that runs Stage 3 and synthesizes the final answer from all council responses and peer-review feedback. The Chairman can be the same model as one of the council members (the default uses google/gemini-3-pro-preview for both) or a completely different model. Choose a model known for strong summarization and instruction-following.
​
OPENROUTER_API_KEY
str
required
Loaded at startup from the .env file via python-dotenv. This key authenticates all requests to the OpenRouter API. It is never hardcoded in config.py or any other source file — the line OPENROUTER_API_KEY = os.getenv("OPENROUTER_API_KEY") will return None if the .env file is absent or the variable is not set, causing every model query to fail.
​
OPENROUTER_API_URL
str
The full URL of the OpenRouter chat completions endpoint: https://openrouter.ai/api/v1/chat/completions. This follows the OpenAI-compatible API shape, so the same endpoint is used for every model regardless of the underlying provider. You should rarely if ever need to change this value.
​
DATA_DIR
str
The directory path (relative to the project root) where conversation JSON files are written. Defaults to data/conversations. Each conversation is stored as a separate JSON file named by its UUID. The directory is created automatically if it does not exist. You can change this to any path writable by the process running the backend.

​
Environment File (.env)

Create this file in the project root alongside pyproject.toml:
.env
OPENROUTER_API_KEY=sk-or-v1-...
Add .env to your .gitignore if it isn’t already there — committing your API key to a public repository will expose it immediately.

​
Choosing Council Models

Any model available on OpenRouter can be added to COUNCIL_MODELS or set as CHAIRMAN_MODEL. Use the format provider/model-name, for example mistralai/mistral-large, meta-llama/llama-3.1-70b-instruct, or cohere/command-r-plus. Browse the full catalogue at openrouter.ai/models. If you want to test a model identifier before committing to it, use the test_openrouter.py script included in the project root.

​
CORS Settings

The backend (backend/main.py) configures FastAPI’s CORS middleware to accept requests from two origins by default:
backend/main.py
app.add_middleware(
    CORSMiddleware,
    allow_origins=["http://localhost:5173", "http://localhost:3000"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)
If you run the Vite frontend on a different port (e.g., because 5173 is occupied), add the new origin to the allow_origins list and restart the backend. Without a matching entry, the browser will block API requests with a CORS error.

​
Port Configuration

ServiceDefault PortConfigured In
FastAPI backend8001backend/main.py (uvicorn call at bottom of file)
Vite frontend5173Vite default; referenced in frontend/src/api.js
The backend port is referenced in two places: the uvicorn.run() call in backend/main.py and the base URL constant in frontend/src/api.js. If you change the port, you must update both files and restart both servers, otherwise the frontend will send requests to the old port and receive connection errors.

Build docs developers (and LLMs) love

Get started for freeTalk to us