What it does SurfSense ingests documents from 27+ sources—Google Drive, Notion, Slack, GitHub, Obsidian, even auth-protected web pages—and lets teams search, chat, and generate reports, podcasts, and presentations from that corpus. It offers both a managed cloud service and a self-hosted Docker deployment with a one-line install script.

The interesting bit The project is essentially a feature-by-feature rebuttal to NotebookLM’s limitations: no source caps, no notebook caps, no 200MB file ceiling, and no Gemini lock-in. You can plug in 100+ LLMs via LiteLLM, run local models through Ollama or vLLM, and collaborate in real time with RBAC and live chat. The desktop app extends this to any application via global shortcuts, text selection, and screenshot capture.

Key highlights

• Hybrid search — semantic + full-text with hierarchical indices and reciprocal rank fusion, not just vector similarity
• Agentic layer — scheduled and event-triggered automations via LangChain Deep Agents, with write-back to Notion, Slack, Linear, Drive
• Desktop integration — Quick Assist, Screenshot Assist, and local folder watch (Obsidian vault sync is explicitly called out)
• No-code automation builder — describe a workflow in chat and the system constructs it
• Self-hosting parity — cloud and on-prem versions share the same feature set; Docker Desktop is the only hard prerequisite

Caveats

• The README states the project is “not yet production-ready”
• Video and podcast generation are admitted weak spots: “NotebookLM is better here, actively improving”
• Real-time collaboration is marked Beta

Verdict Teams who’ve hit NotebookLM’s source limits or want to use their own LLM infrastructure should evaluate this now, but treat it as early-stage. Solo researchers who just need Google’s polished audio overviews should probably wait.