sh1zen/reql — reverse-engineered prompt

Build me a local Python tool called REQL that gives coding assistants a reliable memory of a codebase. I want it to scan a repo and supported docs, understand code across a lot of common languages, and save that into a local graph so an agent can ask focused questions like what files matter, what symbols are related, what tests touch this area, and where the evidence came from. It should feel like something an agent quietly uses in the background, not a separate manual workflow.

Please include a command line app and a small Python API. It should be able to compile a project, refresh incrementally when files change, return compact context instead of dumping whole files, and keep a separate agent workspace for notes, tasks, findings, decisions, risks, links, session maps, and handoffs. Keep it deterministic and local first, with no required cloud service or external graph database. I also want useful exports and reports, plus an optional MCP style server for integrations. Check current docs online if you need to.