It started because I could not get the range I expected with Meshtastic on my Heltec WiFi LoRa 32 V4.3 boards, and actually, instead of searching out why, I got interested in writing some basic communication app, so I stripped things down to the sx1262 communication and built a minimal testbed.

TrailText sends encrypted text messages through this chain: phone → BLE → ESP32 → LoRa → ESP32 → BLE → phone

The goal: text when networks fail.

The repo also includes a LoRa Ping tool to test antennas, RSSI, SNR, round-trip time and range before moving to actual messaging.

Comments URL: https://news.ycombinator.com/item?id=48442923

Points: 1

# Comments: 1

The index is built on Linux from the GoogleNews Word2Vec model, then copied to an SD card. The ESP32-S3 runs search-only for the demo: it receives a 300-dimensional float32 query vector over TCP, searches the nn20db HNSW graph local from SD card, and shows the results.

The demo supports normal word queries and vector arithmetic such as:

Paris - France + Poland

The full GoogleNews model contains around 3 million vectors. In my current test, one search on the ESP32-S3 takes about 12 seconds (ef 15).

This is not meant to beat server/vector-db performance. The point is to explore what useful ANN/vector search can look like on tiny offline hardware with limited RAM and cheap storage.

Video: https://www.youtube.com/watch?v=uMlM1yzEbDw

Comments URL: https://news.ycombinator.com/item?id=48095785

Points: 2

# Comments: 1

nn20db takes a different approach: it is a small C-based vector search engine for Linux and embedded devices, designed to search large persistent HNSW indexes from low-cost storage, including on devices such as the ESP32-P4.

The trade-off is intentional: nn20db favors large indexes, up to million-vectors, offline operation, low hardware cost, and good recall over maximum raw query speed. It is built for cases where the index is larger than available RAM and the device still needs to search locally.

Comments URL: https://news.ycombinator.com/item?id=47930529

Points: 3

# Comments: 1