[DEPRECATED] Programmatic access

This document is deprecated, refer to https://readme.fireworks.ai/docs/querying-text-models instead

Fireworks.ai offers a REST API that is OpenAPI-compatible for /completion and /chat/completion. There are several ways to interact with it:

Fireworks Python client library

The Fireworks Python client library provides a convenient API for accessing Fireworks supported LLMs. See the API reference for a full description of the library.

Installation

Install with Pip:

pip install --upgrade fireworks-ai

Authentication

Generate an API key following the Quickstart instructions and set it in Python

import fireworks.client

fireworks.client.api_key = "<API_KEY>"

or provide as an environment variable

export FIREWORKS_API_KEY=<API_KEY>

Usage

Completion

completion = fireworks.client.Completion.create(
  model="accounts/fireworks/models/llama-v2-7b-w8a16",
  prompt="Say this is a test",
)
print(completion)

Async completion

import asyncio

async def main():
  response = await fireworks.client.Completion.acreate(
    model="accounts/fireworks/models/llama-v2-13b",
    prompt="Once upon a time in an iterative fashion")
  print(response.choices[0].text)
asyncio.run(main())

REST

The base URL for the REST endpoint is https://api.fireworks.ai/inference/v1. You can view all available REST APIs and their options on the API reference page.

curl https://api.fireworks.ai/inference/v1/completions \
  -H 'Authorization: Bearer <YOUR_API_KEY>' \
  -H 'Content-Type: application/json' \
  -d '{
  "model": "accounts/fireworks/models/llama-v2-7b-chat",
  "prompt": "Hello there!"
}'