# slvDev/esp32-ai

Repository: https://github.com/slvDev/esp32-ai
Canonical: https://ross.abutalabs.com/products/esp32-ai
Language: Python
License: MIT
License Family: permissive
Last push: 2026-08-08T09:54:28+00:00

## Health v2 (maintenance only)
Score: 56/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 96, release rhythm 35, longevity 2
- inputs: {"age_days": 41, "days_push": 25, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, young
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 4175, forks 553 (observed 2026-08-28T04:08:38.098080+00:00)

## What it is
A project demonstrating a 28.9M parameter language model running entirely on an ESP32-S3 microcontroller at ~9.88 tokens per second, using Per-Layer Embeddings (from Gemma 3n) to keep most weights in flash. It includes trained models (TinyStories, Barista) and serves as a reference for fitting LLMs onto memory-constrained hardware.

## Use cases
- run an LLM on an ESP32 microcontroller
- generate text offline on embedded hardware
- learn how Per-Layer Embeddings let large models fit in tiny memory
- experiment with on-device AI without cloud connectivity
- deploy a small language model on a microcontroller with 512KB SRAM

## When to choose
- you want to run or study LLM inference on ESP32-S3 or similar microcontrollers
- you need fully offline text generation on embedded hardware
- you want a working example of flash-resident embedding tables and tiered memory layouts

## When to avoid
- you need a model that answers questions, follows instructions, or writes code
- you want production-quality general-purpose LLM inference
- your target hardware lacks PSRAM and multi-megabyte flash

## Facets
- artifact type: learning-resource
- maturity: active
- function: llm-inference, embedded, machine-learning
- domain: embedded-systems, large-language-models, hardware
- platform: embedded, iot, python
- tags: esp32, microcontroller, tinyml, on-device-inference, per-layer-embeddings, gemma-3n, tiny-stories, edge-ai

## Member repositories
- slvDev/esp32-ai (main) score 56

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:38.098080+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:22:42.516386+00:00, confidence not recorded.
  - readme: https://github.com/slvDev/esp32-ai (fetched 2026-08-28T04:08:38.098080+00:00, sha c2c0c783cdd5)
- Data as of 2026-08-30T08:39:29.467469+00:00.