LiteLLM Price Mapper

A web server that matches the models stored in the LiteLLM database against the official LiteLLM price list (from GitHub).

On each request to /prices, the price list and the DB models are fetched fresh and returned as JSON in the price list format. The response contains all original models from the price list, plus the custom DB models keyed by their backend model name.

How it works

1. The price list is loaded from GitHub: https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json
2. All models from the LiteLLM_ProxyModelTable table are read from the database. The litellm_params are decrypted with the salt key to get the real backend name.
3. For each custom DB model, the OpenRouter API (/api/v1/models) is checked first. If the display name exists there, the OpenRouter data is used (converted to the LiteLLM price list format).
4. If the model is not found on OpenRouter, it is looked up in the LiteLLM price list by its display name (model_name).
5. The response contains all original price list entries, plus the custom DB models keyed by their backend name (litellm_params.model), with the price data from OpenRouter or the LiteLLM price list.

Configuration

Configuration is done via a .env file (see .env.example):

DATABASE_URL=postgresql://user:pass@postgres:5432/litellm
LITELLM_SALT_KEY=sk-...

Variable	Description
DATABASE_URL	Connection URL to the LiteLLM database (required)
LITELLM_SALT_KEY	Salt key to decrypt the backend names (required for correct names)
PRICE_LIST_URL	Price list URL (optional, default = GitHub)

Note: LITELLM_SALT_KEY must match the LITELLM_SALT_KEY / LITELLM_MASTER_KEY of your LiteLLM setup. Without it, backend names stay encrypted and cannot be matched correctly.

Endpoints

The server listens on port 8000.

Endpoint	Description
GET /prices	Price list of the models stored in the DB (JSON)
GET /health	Health check

GET /prices returns a JSON object in the price list format:

{
  "gpt-4o": {
    "input_cost_per_token": 2.5e-06,
    "output_cost_per_token": 1e-05,
    "max_input_tokens": 128000,
    "max_output_tokens": 16384,
    "litellm_provider": "openai",
    "mode": "chat"
  }
}

Run with Docker

cp .env.example .env
# adjust .env (set DATABASE_URL)

docker compose up --build -d
# server runs on http://localhost:8000
curl http://localhost:8000/prices

Run locally

cp .env.example .env
# adjust .env (set DATABASE_URL)

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

python -m price_mapper