Mozilla.ai

What we're working on:

• any-guardrail: any-guardrail is meant to provide the minimum amount of access necessary to implement the guardrails in your pipeline.
• mcpd: requirements.txt for agentic systems. Manage MCP servers and tools with one config.
• any-llm: A single interface to use different llm providers.
• any-agent: A single interface to use and evaluate different agent frameworks.
• llamafile: Distribute and run LLMs with a single file.

📘 Blueprints

Blueprints are customizable workflows that help developers build AI applications using open-source tools and models. See our collection of Blueprints below:

• WASM Agents: Blueprint demonstrating how to run AI agents directly in the browser using WebAssembly (WASM).

• Surf Spot Finder: Blueprint for comparing agent frameworks on finding the best surf spot.

• Document-to-podcast: Convert input documents into a podcast featuring multiple speakers.

• Document-to-Markdown: Transform documents into markdown format with Docling, developed in collaboration with EleutherAI.

• Structured-QA: A simple LLM workflow to answer questions based on structured documents.

• Speech-To-Text: Transcribe audio files using Speaches, developed in collaboration with EleutherAI.

• Speech-To-Text Alignment: Align / Bias Speech-to-Text models to a predefined set of words or phrases.

• Speech-To-Text Finetune: Create your own Speech-to-Text model tailored for your specific language & use-case.

• Fake-Audio-Detection: Train and deploy an ML model that detects synthetic audio, developed in collaboration with UncovAI

• OpenStreetMap AI Helper: Map Features in OpenStreetMap with Computer Vision.

• Federated Finetuning: Finetune a language model with federated AI, developed in collaboration with Flower.

• Build Your Own Timeline Algorithm: An approach to personal, local timeline algorithms that people can either run out-of-the-box or customize.

Also, check out the Blueprints created by the community:

• LLM Document Parser: A locally runnable pipeline for parsing structured data from scanned or digital documents using open-source OCR and LLMs.

Would you like to build your Blueprints? Use our template:

• Blueprint template

🐊 Lumigator

Lumigator is an open-source platform developed to help users select the most suitable language model for their specific needs