​
Inference

Explore notebooks and projects showcasing how to run generative AI models on Fireworks, demonstrating both third-party integrations and innovative applications with industry-leading speed and flexibility.

​
LLMs

Dive into examples that utilize Fireworks for deploying and fine-tuning large language models (LLMs), featuring integrations with popular libraries and cutting-edge use cases.

Notebooks

Notebook: Fireworks Model Comparison App

(Python) An interactive Streamlit app for comparing LLMs on Fireworks with parameter tuning and LLM-as-a-Judge functionality.

Notebook: Structured Response with Llama 3.1

(Python) Demonstrates structured responses using Llama 3.1, covering Grammar Mode and JSON Mode for consistent output formats.

Notebook: Llama 3.1 Synthetic Data Generation

(Python) Explores generating synthetic data with Llama 3.1 models on Fireworks, including structured outputs for quizzes.

Notebook: Reasoning JSON Mode - Computer Specifications

(Python) Uses DeepSeek V3 & R1 to generate structured PC specifications while explaining component choices using Reasoning JSON Mode.

Notebook: Reasoning JSON Mode - Healthcare Records

(Python) Demonstrates structured patient record generation using Reasoning JSON Mode to explain treatment recommendations.

Apps

App: Transcription Chat

A Next.js app for real-time transcription chat using Fireworks and Vercel integration.

​
Visual-language

Discover projects combining vision and language capabilities using Fireworks, integrating external frameworks for seamless multimodal understanding.

​
Audio

Explore real-time audio transcription, processing, and generation examples using Fireworks’ advanced audio models and integrations.

Notebooks

Notebook: Streaming Audio Transcription

A notebook demonstrating real-time audio transcription using Fireworks’ whisper-v3-large compatible model. The project includes streaming audio input and getting transcription messages, making it ideal for tasks requiring accurate and responsive audio processing.

Python: Streaming Audio Transcription

Stream audio to get transcription continuously in real-time.

Node.js: Streaming Audio Transcription

Stream audio to get transcription continuously in real-time.

​
Voice Agents

Build complete conversational AI systems with Fireworks’ end-to-end voice solution, combining speech-to-text, language models, and text-to-speech in a unified pipeline.

Key Differentiators:

  • Single API call for complete speech-to-speech interaction
  • Low latency for real-time interaction
  • Built-in context management and conversation state
  • Integrated function calling for complex workflows

Cookbook

Python: Building a Dentist Voice Agent

Example on building a dentist office voice agent to schedule appointments, enroll new patients and more.

​
Image

Experiment with image-based projects using Fireworks’ models, enhanced with third-party libraries for innovative applications in image creation, manipulation, and recognition.

​
Multimodal

Learn from complex multimodal examples that blend text, audio, and image inputs, demonstrating the full potential of Fireworks combined with external tools for interactive AI experiences.

​
Fine-tuning

Access notebooks that demonstrate efficient model fine-tuning on Fireworks, utilizing both internal capabilities and third-party tools like Axolotl for custom optimization.

​
Multi-LoRA

Explore notebooks showcasing the integration and utilization of multiple LoRA adapters in Fireworks. These resources demonstrate advanced techniques for merging, fine-tuning, and deploying multi-LoRA configurations to optimize model performance across diverse tasks.

Notebooks

Notebook: Revolutionizing Product-Specific AI with Fireworks' Multi-LoRA

(Python) An interactive guide showcasing the integration of Multi-LoRA adapters on Fireworks, enabling fine-tuned responses for diverse product domains such as beauty, fashion, outdoor gear, and baby products.

​
Function calling

Explore examples of function-calling workflows using Fireworks, showcasing how to integrate with external APIs and tools for sophisticated, multi-step AI operations.

Notebooks

Notebook: Fireworks LangChain Tool Usage

Demonstrates Function-Calling with LangChain integration, including custom tool routing and query handling. (Python)

Notebook: Fireworks & LangChain Tools Integration

Explore the integration of Fireworks’ function-calling model with LangChain tools. This notebook demonstrates building basic agents using firefunction-v1 for tasks like answering questions, retrieving stock prices, and generating images with the Fireworks SDXL API (Javascript).

Notebook: Fireworks LangGraph Tool Usage

Showcases Function-Calling with LangGraph integration for graph-based agent systems and tool queries. (Python)

Notebook: Fireworks Function-Calling QA with OpenAI

Uses Fireworks’ Function-Calling for structured QA with OpenAI, featuring multi-turn conversation handling. (Python)

Notebook: Fireworks Function-Calling Demo

Demonstrates querying financial data using Fireworks’ Function-Calling API with integrated tool setup. (Python)

Notebook: Fireworks Function-Calling for Information Extraction

Extracts structured information from web content using Fireworks’ Function-Calling API. (Python)

Notebook: Fireworks AutoGen Stock Chart Demo

Generates stock charts using Fireworks’ Function-Calling API with AutoGen integration. (Python)

Apps

App: Functional Chat Demo

A demo app showcasing chat with function-calling capabilities for dynamic service invocation.

​
RAG

Build retrieval-augmented generation (RAG) systems with Fireworks, featuring projects that connect with vector databases and search tools for enhanced, context-aware AI responses.

Notebooks

Notebook: Simple RAG with Chroma for League of Legends

A basic RAG implementation using ChromaDB with League of Legends data, comparing responses across multiple models. (Python)

Notebook: RAG Paper Title Generator

An agentic system using RAG for generating catchy research paper titles with embeddings and LLM completions. (Python)

Notebook: MongoDB RAG Movie Recommender

A movie recommendation system using Fireworks’ function-calling models and MongoDB Atlas for personalized, real-time suggestions. (Python)

Apps

App: RAG Chat App with SurrealDB

A RAG chatbot using SurrealDB for vector storage and Fireworks for real-time, context-aware responses.

​
Integration partners

We welcome contributions from integration partners! Follow these steps:

  1. Clone the Repo: Fireworks Cookbook repo
  2. Create Folder: Add your company/tool under integrations
  3. Add Examples: Include code, notebooks, or demos
  4. Use Template: Fill out the integration guide
  5. Submit PR: Create a pull request
  6. Review: Fireworks will review and merge
Need help? Contact us or open an issue.

​
Support

For help or feedback:

Resources: