• plans/MOQ_MARMOT_AV_PLAN.md — Project plan (browser‑first: Text → Audio → Video)
• MOQ_MARMOT_AV_SPEC.md — Spec for MoQ + Marmot A/V (auth, directory, AEAD, tracks)
• MOQ_CHAT_SERVER.md — Binary Nostr relay (MoQ chat accelerator) design
• NOSTR_AUTH.md — Self‑issued caps + write‑proof auth (JWT alternative)

Notes

• Keep docs as the source of truth. Code should reference these docs instead of duplicating details.
• For implementation, create a new app in a separate folder (e.g., apps/marmot-moq-demo) and link back here.
• A reproducible toolchain is available via nix develop .#. The default shell exposes cargo, wasm-bindgen, and wasm-aware clang wrappers so cargo build -p marmot-chat --target wasm32-unknown-unknown succeeds without extra host setup.
• The marmot-chat crate now exposes the shared scenario fixtures plus identity-centric wasm bindings (create_identity, public_key, create_message, ingest_wrapper, accept_welcome, merge_pending_commit) via serde_wasm_bindgen so browser code can pass plain JS objects.
• Rust/wasm regression: just wasm-test (or nix develop .# -c wasm-pack test --node crates/marmot-chat) exercises a deterministic backlog in pure Rust, ensuring a new participant can ingest previously published wrappers with no JS glue.
• Browser regression suite: npm run test runs the Playwright specs (tests/step2-demo.spec.js for the BroadcastChannel flow plus tests/step4-chat.spec.js for the MoQ transport). Use just web-test to focus on the MoQ run.
• Build pipeline: npm run build (or just build) compiles the marmot-chat wasm bundle and the SolidJS UI in apps/chat-ui/dist/.
• Dev loop: just dev launches an ephemeral nostr-rs-relay, moq-relay, and the chat UI server with fresh builds. Visit http://127.0.0.1:8890/ (no query params) and follow the onboarding — choose NIP‑07 or a developer secret, create an invite, share it, then chat over MoQ.
• Targeted helpers:
  • just chat-dev — rebuild the UI and serve it without relays (useful when pointing at existing infra).
  • just relay-dev — run only moq-relay with self-signed TLS for manual experiments.