Quivr, Joute's review

The essentials

• Open source RAG platform for querying document bases in natural language
• From $20/month (self-hosting free)
• Document uploads (PDF, Word, web), natural language chat, AI agents, API
• Suited for developers and technical teams that want RAG on their own documents

What is Quivr?

Quivr is a French open source project that lets you build a "second brain": a personal or team knowledge base you can query in natural language. The tool accepts document uploads (PDF, Word, text, web URLs), indexes them in a vector database (Supabase pgvector), and lets you query them via a chat interface or API. Quivr agents can use multiple sources in parallel and cite their sources in responses. The project is actively developed on GitHub with a growing community. The cloud version is paid but self-hosting is entirely free.

Strengths

Open source with self-hosting possible

The source code is accessible and self-hosting is documented. For teams with confidentiality constraints, keeping data on your own infrastructure is crucial.

Multi-LLM integration

Quivr supports multiple LLMs (OpenAI, Anthropic, local models). Choose your LLM provider based on cost, performance, or privacy criteria.

API for integrations

Quivr's API lets you integrate RAG capabilities into existing applications. Use cases range from support chatbots to internal search engines.

Limitations

Technical setup for self-hosting

Setting up Quivr self-hosted requires skills in Docker, Supabase, and cloud service configuration. It's not turnkey for non-developers.

Variable performance depending on document quality

Quivr's response quality depends directly on the quality of indexed documents. Poorly formatted or contradictory documents degrade results.

Pricing

Cloud version from $20/month. Self-hosting entirely free. Check quivr.app for cloud plans.

Alternatives

For RAG without infrastructure: NotebookLM (Google) or Mem. For advanced RAG pipelines: LlamaIndex or LangChain. For team knowledge management: Dust.

Verdict

Quivr is one of the best open source options for teams that want RAG on their own documents with full data control. Self-hosting is accessible for technical teams. For non-developers, the cloud version simplifies getting started.

FAQ

Is Quivr really free for self-hosting?

The code is open source and free. Remaining costs are hosting (server), the Supabase database, and LLM API calls (OpenAI, etc.) based on usage.

Does Quivr support documents in languages other than French?

Yes, the underlying LLMs support multiple languages. Response quality on documents in any language depends on the chosen model.

Can you connect Quivr to sources like Notion or Confluence?

Integrations is supported via connectors. Check the Quivr GitHub repository for official connectors.

Does Quivr work with open source models (Llama, Mistral)?

Yes, Quivr supports local models via Ollama and other backends. Ideal for full self-hosting without dependency on proprietary APIs.

Joute may earn a commission on subscriptions taken out via links in this article. This doesn't affect our reviews.