Practical notes for configuring and running Loomweave.
- Getting started — single-flow walkthrough: install, analyse a small repo, connect an MCP client, ask three questions, verify the secret-block. Target ≤15 minutes end-to-end.
- Language support — what each language plugin (Python, Rust) extracts and tags, side by side: entity/edge kinds, categorisation tags, and which tools work per language (e.g. dead-code is Python-only today).
- Rust analysis: known limitations — what Rust analysis does and does not resolve (macros, external edges, dead-code roots).
- OpenRouter LLM provider — API key, model ID, attribution headers, and token-ceiling configuration.
- Coding-agent LLM providers — Codex CLI and Claude CLI as local-login alternatives to API-key provider wiring.
- Runtime topology — supported
loomweave serveandloomweave analyzeconcurrency against one.weft/loomweave/loomweave.db. - Secret scanning — pre-ingest scanner behavior, baseline false-positive workflow, override confirmation, and audit queries.
- Guidance — authoring guidance sheets with the
loomweave guidanceCLI,--match/--scope-level/--expiressemantics, staleness findings, and the export/import team-sharing workflow. - Release handoff — retired Loomweave-owned GitHub ruleset enforcement and current standalone release sequence.
- Federation contracts — read-side HTTP contracts consumed by sibling products such as Filigree.