hirokawaguchi/open-genai

Description and use

**Brief Summary**
Open‑GenAI is an unofficial, fully‑local implementation of Japan’s Government AI “GENAI”, providing SAML‑based authentication, RAG with Qdrant, Whisper transcription, Stable Diffusion image generation, and team‑wide knowledge management—all run on local LLM/OpenAI‑compatible models. Written in TypeScript, the project is actively maintained (last commit 2026‑07‑03) and already has a modest community (≈113 ★, 8 forks).

**Value Proposition**

• **Unified Knowledge Access** – By indexing internal documents in a local vector store (Qdrant) and exposing them through a searchable API/CLI, teams can retrieve precise, up‑to‑date information without exposing data to external services.
• **End‑to‑End AI Stack** – The same repository bundles authentication (Keycloak/SAML), speech‑to‑text (Whisper), text generation (local LLM or OpenAI‑compatible API), and image creation (Stable Diffusion), eliminating the need to stitch together multiple third‑party tools.
• **Data Sovereignty & Cost Control** – All components run on‑premises, guaranteeing compliance with government‑level data regulations while avoiding recurring cloud‑AI fees.

**Practical Adoption Path**

1. **Pilot Setup** – Deploy the Docker‑compose stack in a sandbox environment, connect it to an existing Keycloak realm for SAML SSO, and point Qdrant at a sample knowledge base.
2. **Integration** – Use the provided TypeScript SDK or REST endpoints to embed the RAG service into internal chat‑bots, ticketing systems, or custom front‑ends.
3. **Scale & Harden** – Replace the placeholder LLM with a production‑grade local model (e.g., Llama 3 or Mistral) and configure Whisper/Stable Diffusion GPU resources. Add monitoring, RBAC policies in Keycloak, and backup routines for the Qdrant index.

**Production Readiness**

• **Activity & Community** – Recent commits (within days), 113 ★, and clear documentation indicate healthy momentum.
• **Architecture** – Modular services (auth, RAG, transcription, image generation) are containerized and communicate via standard HTTP/REST, simplifying deployment in Kubernetes or on‑prem VMs.
• **Security & Compliance** – SAML integration via Keycloak satisfies enterprise SSO requirements; however, a formal security audit (dependency scanning, license verification) is still advisable before a full production rollout.

Overall, Open‑GenAI offers a production‑grade, self‑hosted AI platform that can be evaluated quickly and, with modest hardening, be promoted to a mission‑critical knowledge‑assistant for organizations that must keep data in‑house.

Use in a workflow

hirokawaguchi/open-genai helps make internal knowledge searchable and usable by assistants.

• index knowledge bases
• improve search over documents
• ground assistant answers

Limitations and review notes

High for an OSS candidate: recent activity, adoption, and ecosystem signals are strong enough for a serious pilot.