ForkFree

Developer tools

picollm

by Picovoice

picoLLM is a software kit for running compressed large language models directly on your own device, without sending data to the cloud. It lets developers add offline chat and text generation to apps on computers, phones, browsers and Raspberry Pi boards.

Worth a lookEstablished, with a few things to check (listed below).

View on GitHubProject website

Stars
318
Licence
Apache-2.0
Runs on
Your computer
Setup
Moderate
Language
Python
Age
937 days

What you can do with it

  • Run open-weight models such as Llama 3, Gemma, Mistral and Phi locally on a laptop, phone or Raspberry Pi
  • Add offline chat or text completion to a Python, Node.js, .NET, Android, iOS or web app
  • Build a private voice assistant that works without an internet connection for the model itself

Make money with it

A freelance developer could build and sell custom offline AI assistants or voice-enabled devices for clients who need privacy.

An idea, not financial advice. Check the licence before charging for anything built on it.

Ways a business could use it

  • Build customer-facing apps that keep user data on the device, avoiding per-request cloud AI fees
  • Add offline AI features to kiosks, Raspberry Pi gadgets or mobile apps for clients

Can you use it commercially?

Free to use, change and sell, including in paid products. Keep the copyright notice.

Not legal advice. Read the project's licence file before building a product on it.

Getting it running

Install the SDK for your language (for example pip3 install picollmdemo for Python), then get an AccessKey and a model file from Picovoice Console.

Follow the official instructions in the project README.

Things to check

  • Requires a Picovoice AccessKey and internet access to validate it, even though inference runs offline
  • Free use has account limits; continued or larger use needs a commercial arrangement with Picovoice
  • Models must be downloaded from Picovoice Console, not freely distributed
  • The SDK relies on Picovoice's own compiled engine, so the Apache-2.0 licence may not cover everything

Who's talking about it

Checked 03 Nov 2026. Summary drafted by AI from the project's README.

More in Developer tools

taste-skill Leonxlnx
Worth a look

Taste Skill is an add-on for AI coding agents such as Claude Code, Codex and Cursor. It guides the agent to design more distinctive websites and app screens by choosing an art direction, using credited real photographs and checking its own work in a real browser.

109k stars +756 a day MIT
gpu-unlock Zagitta
Check the licence

This tool unlocks the FP/BF16 instruction issue rate on Nvidia 3000-series cards, doubling the speed of certain operations.

14.8k stars +690 a day Unclear
graphify Graphify-Labs
Worth a look

Graphify turns a codebase, along with its documents, PDFs, SQL schemas and configuration files, into a graph of connected concepts that you can search and question. It works as a /graphify command inside AI coding assistants such as Claude Code, Cursor, Codex and Gemini CLI, so they can look things up in the graph instead of reading every file.

141.4k stars +672 a day Apache-2.0