Zyrabit SLM
New Active Docs only Last checked 2026-11-16 Custom / mixed CPU OK Server / Docker
๐ฌ One weekly email: local RAG releases, hardware compatibility updates, and practical setup guides. Free. Unsubscribe anytime.
Updates on X: @mkjerman
What it is
Zyrabit SLM (also marketed as Zyrabit Runtime / Platform) is a local-first Docker stack for private document Q&A: FastAPI gateway, hybrid RAG (Chroma vector + BM25), optional Whisper audio ingest, a web UI on port 8080, and a PII "sandwich" that masks sensitive tokens before prompts reach the model. It expects a separate local inference engine (Ollama, MLX, llama.cpp, or vLLM / Tenstorrent vLLM-TT).
The GitHub repo and Docker Hub image zyrabitcore/zyrabit-slm are MIT open-core for evaluation. The company also sells fixed-price pilots and annual node licenses on zyrabit.com; those commercial offers are separate from the open image. README marks several capabilities as Beta; air-gap is described as validated via an isolated compose network. RAGShelf has not independently verified the air-gap claims or DGX Spark support.
- License
- MIT (open-core image); paid pilots/licenses for enterprise perimeter
- Latest release
- v2.4.9
- Release date
- 2026-11-12
- Hardware
- Docker; CPU OK (llama.cpp/Ollama), Apple Silicon Metal/MLX, NVIDIA CUDA, Tenstorrent NPU; entry ~8GB RAM, production 16โ32GB+ unverified
- Document types
- PDF, DOCX, CSV, text, audio (Whisper) into ChromaDB + BM25 unverified
- OCR / tables / retrieval
- Hybrid vector (Chroma) + BM25 keyword retrieval; PII sanitization before inference; cited answers claimed in product copy unverified
- Cloud calls
- None in air-gap/zero-egress profile once models are local; needs a local inference engine (Ollama, MLX, llama.cpp, vLLM/vLLM-TT) unverified
- Fully offline
- Yes
- GitHub stars
- 4
- Last commit
- 2026-11-12 (last push, GitHub API 2026-11-16)
Install
Copied from the official README. Check the project docs for options.
docker run -d --name zyrabit-platform -p 8080:8080 --add-host=host.docker.internal:host-gateway -e INFERENCE_PROVIDER=ollama -e SLM_URL=http://host.docker.internal:11434 -v zyrabit_vault:/app/db_data zyrabitcore/zyrabit-slm:latestRetrieval
Hybrid (vector + keyword) Agentic / iterative
Based on: README: Chroma vector + BM25 hybrid RAG; ReAct agent with MCP tool calling (checked 2026-11-05) (checked 2026-10-30). Vector vs. vectorless RAG explained โ
NVIDIA DGX Spark
DGX Spark: Unknown
No DGX Spark report found yet. We only mark a tool as working when there is a published source.
Last updated: 2026-11-16. License and release data from the GitHub API. Fields marked unverified have not yet been checked against official docs. How we know: Docs only (see the legend). Spot something wrong? Send a correction.
๐ฌ Get next week's tool
One new local RAG tool in your inbox every Monday. Free, unsubscribe anytime.
Sample issue: "AirLLM: run a 70B model on a 4GB GPU by loading one layer at a time. What it's good for (slow, private batch Q&A over your docs), what it isn't (fast chat), and the install command." Plus one new setup recipe or page, like our local vs. cloud cost calculator.