Show HN: Cabinet – Kb+LLM (Like Paperclip+Obsidian)
| Source: HN | Original article
A new open‑source project called **Cabinet** landed on Hacker News today, positioning itself as a hybrid of the Paperclip LLM‑agent framework and the Obsidian note‑taking ecosystem. Its creator, Hilash, describes Cabinet as a “KB + LLM” platform that lets developers run autonomous agents, schedule heartbeats, and write directly to a personal knowledge base in formats ranging from Markdown and CSV to PDFs and web‑app data. The codebase is freely available on GitHub, and the initial release includes a web UI, a simple API for plugging in any LLM (including Claude Code, OpenAI, or locally hosted models via Ollama), and built‑in pipelines for common workflows such as lead tracking, email drafting, and data extraction.
The launch matters because it tackles two growing pains in the AI‑augmented productivity market. First, most existing solutions—Obsidian plugins, Notion AI, or proprietary assistants—require users to trust closed‑source back‑ends with their most sensitive notes. Cabinet’s open architecture lets users keep data on‑premises while still benefitting from LLM reasoning. Second, it bridges the gap between raw LLM calls and structured knowledge management, a niche that Paperclip pioneered for agents but never linked to a persistent, graph‑based note system. Early adopters, especially developers and researchers who already use Obsidian for personal knowledge graphs, can now prototype AI‑driven workflows without leaving their familiar markdown environment.
What to watch next is the community response. The project’s GitHub repository already shows a handful of pull requests aimed at adding Ollama support and a browser‑based WASM inference layer—an echo of the TurboQuant‑WASM effort we covered last week. If Cabinet gains traction, we may see integrations with existing AI‑toolchains, commercial backing from Nordic AI startups, or even a hosted SaaS variant that balances openness with managed scalability. For now, the open‑source crowd will be testing its agent reliability, data‑privacy guarantees, and how smoothly it can replace the patchwork of scripts many developers currently cobble together.
Sources
Back to AIPULSEN