mbreiser · Oct 28, 2026 · Oct 19, 2026 · Oct 27, 2026 · Oct 27, 2026 · Oct 27, 2026
diff --git a/.agents/skills/g6-orientation/SKILL.md b/.agents/skills/g6-orientation/SKILL.md
@@ -42,7 +42,7 @@ interface.
 |---|---|
 | **reiserlab/maDisplayTools** | **MATLAB** tools (G4+). Pattern math reference — **web output must match MATLAB exactly** (CI validates, tol 1e-4). Its protocol **runner is G4.1-only and frozen** (decision 2026-10-26); protocol authority = `webDisplayTools/docs/development/v3-spec.md` + conformance fixtures. (Its `docs/development/G4G6_ROADMAP.md` is largely historical now — use the umbrella docs for current G6.) |
 | **reiserlab/LED-Display_pyDisplayTools** (pyDisplayTools) | **Python** tools for the displays (G4 pattern + experiment-folder tools). **Accepted destination** for the Python arena library and the **headless runner** (`arena-run`), a peer of the web runner (decision 2026-10-26). |
-| **reiserlab/rig-nodes** | **Accepted destination, not yet created:** shared ROS 2 infrastructure for every lab rig (`rig_interfaces`, device nodes for FicTrac / Kinefly / BIAS / ScanImage / SiPM, MCAP recording and supervision primitives). Today that code lives in `webDisplayTools/rig/` on the 2026-10-20 release candidate. Record: `webDisplayTools/docs/development/rig-software-repos-and-runners.md`. |
+| **reiserlab/rig-nodes** | **Current location (since 2026-10-26)** of the shared ROS 2 infrastructure for every lab rig: `rig_interfaces` (FicTracFrame, ArenaFrame, ConnectorReply), the device adapters + simulators (FicTrac, Kinefly, BIAS, ScanImage; SiPM and the ScanImage frames node follow when their PRs land), the MCAP recorder/indexer/CDR decoder/Foxglove rewriter, the supervision primitives and the Zenoh configs, as the `rig_nodes` + `rig_interfaces` packages (one package-set version, conda channel at a release tag). `CONTRACT.md` there is the graph contract (topics, QoS, envelope, invariants). webDisplayTools' `rig/` installs the set and keeps only the arena gateway, composition and visualization. Record: `webDisplayTools/docs/development/rig-software-repos-and-runners.md`. |
 
 ### G6 firmware (what the web connects to over USB)
 | repo | role |
@@ -120,13 +120,15 @@ interface.
   (`CE_DISPLAY_ACTIVE`); STOP also blanks the panels (they latch the last frame).
 - **"Web is the runner"** (CSHL) — the browser **executes** the protocol, host-timed:
   `js/arena-runner-g6.js`.
-- **FicTrac closed loop** — `fictrac_sim.py` (or real FicTrac) → **`fictrac-bridge/bridge.py`**
+- **FicTrac closed loop** — the FicTrac simulator (`rig_nodes.sims.fictrac_sim`, rig-nodes; or real FicTrac) → **`fictrac-bridge/bridge.py`**
   (WebSocket `ws://localhost:8765` ← UDP `:60000`) → browser. `pixi run bridge` + `pixi run
   sim` (start the bridge first). The web runner executes only the `fictrac` and `log` plugins;
   `backlight`/`camera`/`temperature` are declared for MATLAB and **skipped on the web**.
-- **ROS 2 graph (LAB-303, candidate 2026-10-20)** — `rig/` replaces the bridge with a graph
+- **ROS 2 graph (LAB-303, released 2026-10-20/26)** — `rig/` replaces the bridge with a graph
   (rmw_zenoh, MCAP record, BIAS + ScanImage connectors) behind the SAME WebSocket; the Studio is
-  unchanged. Where the pieces will live from here: `rig-software-repos-and-runners.md` (above).
+  unchanged. Since 2026-10-26 the device nodes, simulators and recording primitives come from the
+  installed `rig_nodes` package (reiserlab/rig-nodes); `rig/` keeps the gateway (`bridge_node.py`),
+  the composition (`rig.py`) and the arena visualization. Record: `rig-software-repos-and-runners.md`.
 
 ## Pointers (go here for depth)
 

diff --git a/.github/workflows/rig-interfaces.yml b/.github/workflows/rig-interfaces.yml
diff --git a/CLAUDE.md b/CLAUDE.md
@@ -107,11 +107,18 @@ When editing web tools, audit all buttons and controls in the modified code sect
 - Dependencies via CDN or vendored under `js/vendor/` (e.g. `yaml`); no runtime npm
 - Web outputs must match MATLAB outputs exactly
 
-**Repository boundaries (decision 2026-10-26):** `docs/development/rig-software-repos-and-runners.md`
-is the record. The browser tools and their rig-side gateway (`rig/nodes/bridge_node.py`) stay
-here; the Python arena library + headless runner go to `LED-Display_pyDisplayTools`; shared ROS 2
-infrastructure (interfaces, device nodes, recording, supervision) goes to `rig-nodes`, which does not
-exist yet. Until the extraction session, every file is where the record's inventory says it is today.
+**Repository boundaries (decision 2026-10-26; extraction done the same day):**
+`docs/development/rig-software-repos-and-runners.md` is the record. The browser tools, their rig-side
+gateway (`rig/nodes/bridge_node.py`), the arena composition (`rig/nodes/rig.py`, `rig/pixi.toml`) and the
+arena visualization stay here; the Python arena library + headless runner go to
+`LED-Display_pyDisplayTools` (LAB-337); shared ROS 2 infrastructure lives in **`reiserlab/rig-nodes`**
+(`rig_interfaces`, the device adapters + simulators, `rig_nodes.recorder` / `mcap_index` / `cdr` /
+`foxglove_json`, `rig_nodes.supervise`, the Zenoh configs; contract: its `CONTRACT.md`) and is INSTALLED
+into `rig/` as the pinned `rig_nodes` + `rig_interfaces` packages from that repo's channel. A node fix
+is tested here with `pixi install -e dev` (rig-nodes `main` built from source — git source dependencies,
+NEVER path dependencies on a sibling checkout: pixi validates every environment's source paths before
+installing any environment, which broke `pixi install` on rig PCs on 2026-10-27); a release of the set
+re-pins `rig/pixi.toml`. Never re-add a device node or recorder copy under `rig/`.
 
 ### Dev environment (pixi — the only tool you install)
 

diff --git a/arena_studio.html b/arena_studio.html
@@ -3426,7 +3426,7 @@ <h2 id="modalTitle">Import error</h2>
 </div>
 
 <footer id="footer">
-  <span class="foot-left">Arena Studio v0.98 | 2026-10-26 18:11 ET · <a href="https://github.com/reiserlab/webDisplayTools" target="_blank" rel="noopener" title="webDisplayTools on GitHub: source, issues, and release notes (docs/development/arena-studio-release-notes.md)">GitHub</a></span>
+  <span class="foot-left">Arena Studio v0.99 | 2026-10-27 15:21 ET · <a href="https://github.com/reiserlab/webDisplayTools" target="_blank" rel="noopener" title="webDisplayTools on GitHub: source, issues, and release notes (docs/development/arena-studio-release-notes.md)">GitHub</a></span>
   <!-- Course-repo quick-links: open protocols / logs / patterns in a new tab.
        Hrefs built from the configured repo + bench id (updateGhQuickLinks). -->
   <span id="ghQuickLinks" title="Open the course repo on GitHub (new tab)">

diff --git a/docs/development/arena-studio-release-notes.md b/docs/development/arena-studio-release-notes.md
@@ -4,6 +4,14 @@ The Studio's footer used to carry the full changelog inline; it now shows one li
 history lives here. Newest first. (Per-session engineering detail stays in
 `arena-studio-handover.md` and the design docs — this file is the user-facing what-changed list.)
 
+## v0.99 (2026-10-27) · rig/ installs the rig-nodes v0.2.1 package set; closed-loop-report applies start_frame; tiff2mcap --runlog
+
+<!-- #274 -->
+none (no user-visible change; rig PCs: `cd rig && pixi install` re-locks onto the rig-nodes channel — the Studio's `hello_ack` now carries `packages`).
+
+<!-- #273 -->
+none (offline analysis tool)
+
 ## v0.98 (2026-10-26) · Ros2-poc: the FicTrac record in the bag — one /arena/frame message per frame with 10 columns (LAB-303) · Assert sticky controller settings before every run: controller: block, rig defaults, run header (LAB-230) · Rename ros2-poc/ → rig/ and arena_interfaces → rig_interfaces (LAB-303 cleanup, no behavior change)
 
 <!-- #264 -->

diff --git a/docs/development/closed-loop-bias.md b/docs/development/closed-loop-bias.md
@@ -370,8 +370,16 @@ idx = round((coupling * rel + offset + b) / deg_per_frame) % n_frames
 # bridge <= 3.2 logs: rel = ((hd_deg - hd0_deg + 180) % 360) - 180;  idx = round((rel + offset + b) / gain) % n_frames
 ```
 
+`offset` is NOT simply the `config` echo's `offset` field on bridge 3.4: when the epoch's
+config carries `start_frame` (the trialParams `frame_index` the epoch opens on), the bridge
+sets `offset = (start_frame % n_frames) * deg_per_frame` for that message while the logged
+browser config still says `offset: 0`; the next config's `offset` replaces it
+(`docs/development/closed-loop-start-frame.md`). The bridge's `[cfg] applied` line goes to
+its stderr, not the run log, so readers replay the config echoes in order.
+
 `scripts/closed-loop-report.py` does exactly this (it picks the formula from whether the
-`config` echo carries `coupling`) and reports the mismatch fraction per epoch.
+`config` echo carries `coupling`, and replays `start_frame` the way the bridge applies it)
+and reports the mismatch fraction per epoch.
 
 Because the waveform is analytic and deterministic, this recovers the exact value the
 bridge used — no per-frame bias column is needed.