Four Open-Source AI Projects for Agentic Workflows
Matthew Bermango watch the original →
the gist
A walkthrough of four open-source tools for AI agents, covering search, local document processing, engineering workflows, and context compression to reduce token costs.
Search and Knowledge Processing
Last 30 Days is a search tool that aggregates data from Reddit, Hacker News, GitHub, and PolyMarket based on human engagement metrics rather than traditional search algorithms. It allows users to query trending topics and generate concise summaries or HTML briefs by installing it as a skill in agentic platforms. Open Notebook provides a local, open-source alternative to Google's NotebookLM. It supports document ingestion for RAG-based Q&A and includes automated podcast generation, key insight extraction, and table-of-contents creation. Users can configure it with various LLM providers or run it locally via Ollama or LM Studio.
Engineering and Optimization
Agent Skills is a library designed to streamline the software engineering lifecycle for AI agents. It provides seven slash-commands that map to specific development stages: /spec, /plan, /build, /test, /review, /code, and /simplify. The tool includes an interactive interview mode to help define project requirements and generate structured markdown specifications. Headroom is a context compression tool that reduces token usage for LLM-based agents by compressing logs, RAG chunks, and conversation history. It reportedly achieves significant token savings, such as reducing 17,000 tokens to 1,400 in code search tasks, without degrading model performance. It also features a headroom learn command that analyzes failed sessions to suggest improvements for claw.md or agents.md configuration files.