feat: upgrade to async extractor, add RAG processing, link healing and Docker support

AGENTS.md

@@ -0,0 +1,29 @@
+## How to investigate
+Read highest-value sources first:
+- `README*`, root manifests, workspace config, lockfiles
+- build, test, lint, formatter, typecheck, and codegen config
+- CI workflows and pre-commit / task runner config
+- existing instruction files (`AGENTS.md`, `CLAUDE.md`, `.cursor/rules/`, `.cursorrules`, `.github/copilot-instructions.md`)
+- repo-local OpenCode config such as `opencode.json`
+
+If architecture is unclear, inspect representative code files to find entrypoints and flow. Prefer executable sources of truth over prose. Trust executable source over conflicting docs.
+
+## What to extract
+Look for high-signal facts:
+- Exact developer commands (especially non-obvious ones)
+- How to run a single test or focused verification step
+- Required command order (e.g., `lint -> typecheck -> test`)
+- Monorepo/multi-package boundaries and entrypoints
+- Framework/toolchain quirks (codegen, build artifacts, dev servers)
+- Repo-specific style or workflow conventions
+
+## Questions
+Only ask if the repo cannot answer something important, such as undocumented team conventions or missing setup prerequisites.
+
+## Writing rules
+Include only high-signal, repo-specific guidance:
+- Exact commands and shortcuts that agents would otherwise guess wrong.
+- Architecture notes not obvious from filenames.
+- Conventions differing from language/framework defaults.
+- Setup requirements, environment quirks, and operational gotchas.
+Exclude generic advice or long tutorials. Be concise with short sections and bullets.