RAG・知識ベース
trustgraph
The semantic intelligence layer for ontologies — governed, traceable knowledge for AI agents
★ 2.8k スターApache-2.0
ガイドを見る →Utopiaは、deepletheが公開しているオープンソースのRAG・知識ベースプロジェクトです（GitHub: deeplethe/utopia）。GitHubスターは8,372、フォークは1,159で、Apache-2.0ライセンスの下で公開されています。GitHubでの最終更新は2026年10月15日です。
Utopia — an open-source enterprise world model: a bitemporal knowledge substrate that learns from your documents and gives agents a governed, auditable memory
Clone the repo and bring the stack up with Docker Compose — the first account you register becomes the administrator.
Under Administration → Models, set the chat and embedding endpoints Utopia should use.
Start extracting business documents into the public knowledge base and watch the ontology evolve with time awareness.
Knowledge carries both valid time and record time, so past states of understanding can be replayed and audited.
Conflict detection, reasoning and decision support run against a governed ontology, not raw chunks.
Prebuilt images for x86-64 and ARM64 (Apple Silicon, Graviton) run entirely on hardware you control.
The roadmap adds checking agent calls against ontology rules and symbolic logic, plus agent memory over MCP.
Requirements: Docker; local development also needs Rust 1.85+, Node 20+ and pnpm.
git clone https://github.com/deeplethe/utopia.git
cd utopiaBrings up the app profile; then open http://localhost:1516 and register.
docker compose --profile app up -dUses the build overlay of the same Compose stack.
docker compose -f docker-compose.yml -f docker-compose.build.yml --profile app up -d --buildDeepLethe; the project site is utopia.bi.
Yes, Apache-2.0 licensed, with prebuilt container images on GHCR.
You configure your own chat and embedding model endpoints; Utopia itself has no licence fee.
A vector store retrieves similar chunks of present knowledge. Utopia models knowledge bitemporally over an ontology, so it can show what was believed, when it changed, and what a decision rested on.
Clone the repo, run docker compose --profile app up -d, open http://localhost:1516, register the admin account, and configure your model endpoints.
The semantic intelligence layer for ontologies — governed, traceable knowledge for AI agents
A private AI platform for agents, assistants and enterprise search over your documents
The open-source, permission-aware context layer that lets agents search all your company knowledge.
Skills that let coding agents search and read your local documents.
A trilingual e-commerce knowledge base agents can install.
A knowledge-base platform with visual AI workflows — build Q&A systems on your own documents
LibreAIカタログに掲載されたRAG・知識ベースプロジェクトの中で最もスターの多いもの — Utopiaが合わなかった場合に検討したい、現実的な候補リストです。
Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or…
RAGFlow is a leading open-source Retrieval-Augmented Generation (RAG) engine that fuses cutting-edge RAG with Agent capabilities to create a superior context…
Lightweight OCR toolkit turning PDFs and images into structured, LLM-ready data in 100+ languages
MinerU — turns PDFs and Office documents into LLM-ready Markdown and JSON for agents and RAG pipelines, Apache-2.0 plus terms
Stop renting your intelligence. Own it with AnythingLLM. Everything you need for a powerful local-first agent experience
カタログの中でUtopiaとスター数が近いRAG・知識ベースプロジェクト — 前後の棚もそのまま見て回れます。
Deep Lake — a multimodal datalake and vector store for AI, with version control and training streaming.
DeepSearcher — open-source deep research over your private data, with reasoning models and Milvus retrieval.
PaperQA2 — high-accuracy RAG for answering questions from scientific papers, with citations you can check.
Unstract — LLM-driven extraction of unstructured data, built for API deployments and ETL pipeline workflows.
Polyglot document intelligence in Rust — extract text, tables and structured data from 106 formats.
Vespa — the AI search platform: vector, text and tensor search at production scale.