Show HN: Kavim – Open-source, local-first AI brainstorming (kavim.deepelegant.com)

Kavim is an open-source, local-first “brainstorming canvas” that lets teams chat with AI, branch conversations into parallel idea paths, and keep all data on-device rather than in the cloud. Users can start a thread, fork it into multiple branches, add sketches or links, and invite teammates to collaborate live—every branch preserves its context and provenance so ideas aren’t lost or copied between documents. The app is downloadable and runs without sending your data to Kavim’s servers; when AI is used it connects only to whatever model you choose (local models or paid services you already subscribe to such as ChatGPT, Claude, or Gemini). For the AI/ML community this matters because it tackles privacy, reproducibility, and prompt engineering workflows in one UX: branching saves decision histories and experiment forks, local-first operation reduces attack surface and vendor lock-in, and its open-source nature lets teams inspect or adapt the code. Technically, Kavim’s model-agnostic design means it can act as a client for local LLMs or external APIs while preserving stateful conversation trees locally — a practical tool for product teams, researchers, and prompt engineers who need collaborative, auditable brainstorming without sacrificing control or incurring extra cloud costs.