Show HN: Orchid – Local-first record and replay for AI agent debugging (github.com)

Orchid has launched a powerful debugging tool tailored for AI developers, enabling them to record and replay the network traffic of AI agents through a zero-instrumentation proxy. This innovative system allows users to analyze every aspect of their agent's operations, including prompt responses, token counts, and API costs, all while facilitating a "time travel" option to inspect runs step-by-step. The tool supports seamless integration with popular coding environments like Cursor and VS Code, boosting debugging efficiency by allowing AI assistants to directly query past runs. Notably, Orchid ensures that all recorded data remains local, processing everything within a SQLite database, which is a significant consideration for privacy-focused developers. This tool's implications for the AI/ML community are profound. By removing the need for cumbersome logging practices and enabling deterministic replay tests, Orchid simplifies the debugging process while eliminating API costs traditionally associated with LLM interactions. Furthermore, the built-in cost calculation and mock management features enhance testing accuracy and responsiveness, allowing teams to optimize their models effectively. With Orchid positioned as an open-source solution, it opens doors for collaborative improvements and potential contributions from the wider developer community, fostering innovation in AI agent development and debugging.