ruview vs laya
Two of the top projects in Open models, side by side: who leads on each score part and who each one suits, then setup, license, activity and what each review found.
ruview
WiFi CSI sensing on ESP32 for presence, breathing, heart rate and pose
laya
Single-pass multilingual classifier for typed choice, score and yes/no decisions
Verdict
ruview ranks higher, 95 to 86. Of the 4 score parts we measure for both, ruview leads on 2, laya on 1 and 1 is even.
- Adoptionruview leads by 13 points, 98 to 85.
- FreshnessEven at 100.
- Maintenancelaya leads by 5 points, 93 to 88.
- Agent-readyruview leads by 55 points, 85 to 30: a wide gap.
Pick ruview if you are
Self-hosters and researchers needing camera-free presence and vitals sensing
- Runs on roughly $9 ESP32 nodes; the pretrained encoder fits in 8 KB at 4-bit
- Publishes 21 Home Assistant MQTT entities per node, plus Matter and Apple Home bridges
Pick laya if you are
Engineers needing fast multilingual triage, routing and moderation classifiers
- One forward pass: 33 ms per question, 7.2 ms/question batched, measured on a T4
- No generation, so no output parsing; answers are typed choice, score or probability
| What we compare | ruview | laya |
|---|---|---|
| Score parts, out of 100 | ||
| Adoption | 98, widely used | 85, widely used |
| Freshness | 100, active | 100, active |
| Maintenance | 88, healthy | 93, healthy |
| Agent-ready | 85, ready | 30, minimal |
| Facts from GitHub and the README | ||
| Stars | 101.9k | 34.8k |
| License | MIT (permissive) | Apache-2.0 (permissive) |
| Last commit | Oct 2026 | Nov 2026 |
| Last release | Oct 2026 | Nov 2026 |
| Language | Rust | Python |
| Docker | Yes | Yes |
| GPU | Optional | Optional |
| arm64 or Apple Silicon | Mentioned | Mentioned |
ruview
RuView reads Channel State Information from ESP32-S3/C6 nodes and derives presence, person count, breathing (6–30 BPM), heart rate (40–120 BPM), falls and 17-keypoint pose without cameras. It is written in Rust, with pretrained models on Hugging Face, Home Assistant, Matter and HomeKit integration, and an npm MCP toolkit for AI-guided setup. Several accuracy figures are self-reported, and the README says the on-device pose model is a first cut.
Who it is for: Self-hosters and researchers needing camera-free presence and vitals sensing
Strengths
- Runs on roughly $9 ESP32 nodes; the pretrained encoder fits in 8 KB at 4-bit
- Publishes 21 Home Assistant MQTT entities per node, plus Matter and Apple Home bridges
- Pretrained weights on Hugging Face, plus a phase-variance presence fallback needing no model
- README retracts an earlier accuracy claim and flags unvalidated vitals and synthetic RF accuracy
Weaknesses
- Live sensing needs ESP32-S3/C6 CSI firmware; the Docker image runs simulated data
- Committed on-device pose model scores PCK@20 of 3.0% and its runtime path is a stub
- Through-wall range is about 5 m and signal-dependent
- Very large README and feature surface; hard to tell which parts are production-ready
- GPU optional
- Docker + Compose
- Needs ESP32-S3 or ESP32-C6 boards, Cognitum Seed (optional), Home Assistant / MQTT (optional)
- Models: safetensors, RVF, Candle
- port 3000
laya
Laya answers typed questions (choice, score, noul) about a text, email, ticket or JSON state in one forward pass of an encoder model, with no text generation. A Router picks between three checkpoints (ModernBERT-large English, mmBERT-base multilingual for 100+ languages, and a typed-decisions fine-tune). It ships as a Python package with a CLI, optional HTTP and MCP servers, and LangChain, LlamaIndex and CrewAI integrations.
Who it is for: Engineers needing fast multilingual triage, routing and moderation classifiers
Strengths
- One forward pass: 33 ms per question, 7.2 ms/question batched, measured on a T4
- No generation, so no output parsing; answers are typed choice, score or probability
- Router selects English or multilingual checkpoint per request; 100+ languages covered
- Fine-tuning notebooks provided; typed-decisions checkpoint scores 0.766 vs 0.362 for base English
Weaknesses
- Zero-shot accuracy is low on domain tasks; fine-tuning is needed for good results
- Multilingual checkpoint ships with a 1,024-token limit; beyond about 4,000 tokens accuracy varies
- Only classifies fixed question types; cannot generate or explain text
- Checkpoints download from Hugging Face on first use, so offline setups need preparation
- GPU optional
- Docker + Compose
- Needs Hugging Face Hub (checkpoint download), PyTorch, transformers
- Models: ModernBERT-large (421M), mmBERT-base (322M), ONNX Runtime
More in Open models
Building with these?
We help teams pick an open-source AI stack and run it in production. Client work never changes a rank.