Emergency

Survival Co-pilot

Handheld LoRa radios relay survival questions to an off-grid laptop that answers with zero internet

ragembeddingstext-generation

People in a disaster zone or remote wilderness carry handheld Meshtastic LoRa radios with no cell signal and no internet. Survival Co-pilot puts a base station on that mesh - a laptop or SBC running a fully on-device LLM and RAG through the QVAC SDK - that auto-answers survival questions the moment they come in over the radio.

Two local processes, no cloud anywhere

A QVAC server exposes the LLM and embedding model on localhost, and a Python base station owns the Bluetooth radio link, retrieves the top-k matches from a bilingual RAG corpus, calls the model, and chunks the reply to fit under 200 bytes per LoRa segment before sending it back over the mesh.

architecture (README.md)

[Handheld Meshtastic] --LoRa--> [mesh relays] --LoRa--> [LoRa / Meshtastic Node]
                                                                  │ BLE
                                                                  ▼
                          Base station (laptop / SBC) - 100% local
                          basestation.py --HTTP--> qvac serve openai
                           owns BLE, RAG retrieve,    @qvac/sdk: co-pilot (LLM) · embed-mlm
                           chunk ≤200B

Demo video

Watch the Survival Co-pilot demo →

Why it refuses before it hallucinates

If the best RAG match scores below 0.40 similarity, it replies 'out of scope' instead of guessing
Retrieval is cross-lingual - a Chinese query can retrieve relevant English survival chunks and back
Every reply is chunked to fit a 200-byte LoRa segment with no cloud round trip possible anyway

Ship your own local AI

One install in JavaScript or Python. 10+ AI tasks. Private, offline and free - on the hardware you already own.

Read the docs