Repository navigation
Release 2026-10-28: Arena Studio v0.100 - #283
Merged
Merged
Conversation
…rk between pairs) per Grace; smoke variant same shape The full run is now 24 pairs (3 duties x 4 colors x 2 reps), ~16 min; the 10 s pause is kept across block boundaries. The reversal is phase-continuous (15 s x 80 fps = 1200 frames, a multiple of the 20-frame period) and the camera timestamp moves to the start of each pair so its round trip never stretches the CW half. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ove camera + G6 dashboard plan Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… dashboard phases 1/1b/2) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…303) Live: bridge_node also publishes /arena/browser (the bag record is unchanged); the new stimulus_node turns it into the EXPECTED display (the trial's .pat frame x duty in the panels' LED colors, PNG) and the OBSERVED display (controller telemetry cf rows), plus the arena's lag, display rate and health, all on the controller's own clock. viz_node gains BIAS fps-from-count, drops, rtt and writer queue, and --sources. rig.py: --stimulus (included by --foxglove) and --no-fictrac. Layout: foxglove/arena-okey-cokey.json. Replay: foxglove_json.py --stimulus rebuilds the same topics from a finished bag (nodes/dashboard_replay.py, the same pure code) and /camera/image from the run's uFMF (tools/ufmf.py, BIAS's v4 layout), placed on the bridge clock by the getTimestamp replies. Panel colors are a Python port of js/panel-color.js, pinned cell by cell under Node. Tests: test-stimulus 54/54 (incl. a real 80 fps trial from the 2026-10-14 smoke run, every lag 0), test-stimulus-live 10/10, harness 54/54, root suite 71/71. No live camera (BIAS unchanged by decision). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…e dashboard steps, handover for the lab-PC test Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…art from it (no Python/git assumed) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- pixi.toml: viz env on win-64 prepends Library\lib to PATH; foxglove_bridge exited 0xC0000135 (foxglove.dll / foxglove_cpp_shared.dll not on the DLL path) - tools/bias-ufmf.ps1 + config/bias/okey-cokey-bfs-u3-51s5m.json: BIAS v0.73.1 restarts in jpg and rejects load-configuration until the camera has captured once; the script does connect -> 2 s capture -> load, run after every BIAS start - setup checklist + run sheet: arena on this rig is mounted upside down (protocol CW turns CCW seen from above; protocols unchanged, no direction test needed), Foxglove install URL/silent flags, 8766 not 8765, path placeholders (no D: on ww14), runlog-check pre-run "unknown" trials, known issues - .gitignore: stop ignoring tests/fixtures/*.jsonl.gz (the okey smoke fixture was never committed; still needs adding from ww18) - okey-cokey-pc-ww14-setup-2026-10-22.md: PC facts, tool versions, BIAS/camera, two passing smoke runs Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Run sheet step 5/6 on REISERLAB-WW14, 2026-10-22: - Full g6_color_optomotor_test with flies: runlog-check 48/48 pass, BIAS 58/58 ok, 70 953 frames at 73.3 fps. Results in okey-cokey-pc-results-2026-10-21.md. - The first full attempt died at minute 9 when the Claude desktop app restarted and took BIAS, the rig and the Studio server with it. New tools/okey-cokey-start.ps1 starts all three through WMI so they outlive the launcher; the run sheet points real runs at it. - The controller moved from COM3 to COM5 (a second Teensy is on COM3): setup notes and run sheet now say to pick it by the name G6_Arena. - A deliberate occlusion made the ufmf 11.8 GB instead of ~0.5 GB; a by-eye panel glitch report is recorded as open (telemetry is clean). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ws Terminal tabs Ran tools/okey-cokey-start.ps1 on ww14 with BIAS, the rig and the server all stopped: exit 0, ports 5010/7447/8765/8766/8092 up, every listener's parent chain runs through WmiPrvSE (none under the Claude app), BIAS on ufmf. The rig and server consoles open as tabs of one Windows Terminal window, which hid the rig tab from the operator at first; the script, run sheet and results now say so and how to stop them. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…or glyph on several panels Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…nto claude/okey-cokey-rig # Conflicts: # rig/.gitignore # rig/HANDOVER-okey-cokey-2026-10-21.md # rig/config/bias/okey-cokey-bfs-u3-51s5m.json # rig/docs/okey-cokey-foxglove-dashboard-plan.md # rig/foxglove/arena-okey-cokey.json # rig/nodes/arena_display.py # rig/nodes/bias_health.py # rig/nodes/dashboard_replay.py # rig/nodes/stimulus_core.py # rig/nodes/stimulus_node.py # rig/okey-cokey-pc-2026-10-21.md # rig/okey-cokey-pc-results-2026-10-21.md # rig/okey-cokey-pc-setup.md # rig/okey-cokey-pc-ww14-setup-2026-10-22.md # rig/tests/fixtures/protocols-2026-10-14/g6_color_optomotor_smoke.yaml # rig/tests/test_stimulus.py # rig/tests/test_stimulus_live.py # rig/tools/bias-ufmf.ps1 # rig/tools/okey-cokey-start.ps1 # rig/tools/ufmf.py
…y reads (it had only ever existed in one worktree; the .gitignore exception was added 2026-10-22 but the file never was) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…d on the installed rig_nodes package (LAB-303, LAB-338) The shared half of rig/ moved to reiserlab/rig-nodes (PR #274); this merge re-expresses the okey-cokey additions through the installed package instead of the deleted local modules: * rig.py — the --no-fictrac / --stimulus / --stimulus-args options and the stimulus Child on top of the rig_nodes.supervise composition (fictrac via `python -m rig_nodes.fictrac_udp_adapter` only when not --no-fictrac; the udp 60000 preflight skipped the same way). * foxglove_json.py — the arena wrapper around rig_nodes.foxglove_json.convert keeps #274's `enrich` hook (frame derivations, FicTrac columns) and adds the --stimulus dashboard replay as the rewriter's `derive` hook (whole-stream → new topics, foxglove.CompressedImage by well-known schema name); dashboard_replay.run_streams is unchanged. Needs rig-nodes PR #1 (derive hook). * bias_health.py — moved to rig-nodes (rig_nodes.bias_health, rig-nodes PR #1); viz_node.py, dashboard_replay.py and the tests import it from there, test_stimulus's bias_health section moved with it (tests/test_pure.py in rig-nodes). * viz_node.py / stimulus_node.py / test_stimulus_live.py — `_common` → `rig_nodes._common`, the launcher primitives → `rig_nodes.supervise`. Until the next rig-nodes package set ships, run with `cd rig && pixi install -e dev` (a sibling ../../rig-nodes checkout on feat/bias-health or later) — the released 0.2.1 has no bias_health / derive. Tests (dev env): test-stimulus 50/50, test-viz, `pixi run test`; root `pixi run test` 72/72. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…ut; test_stimulus_live reads them from rig_nodes.zenoh `pixi install -e dev` installs rig_nodes EDITABLE (a .pth into ../../rig-nodes/src/rig_nodes), so the workspace-level ZENOH_ROUTER/SESSION_CONFIG_URI (…/site-packages/rig_nodes/config/…) named files that do not exist in that environment: the router and every rclpy process failed with "Error configuring Zenoh session" and `pixi run -e dev test` lost every check. `[feature.dev.activation.env]` (+ the win-64 target) now points both URIs into the checkout. (Pre-existing on feat/rig-nodes-extraction, where the dev environment had not been run through the harness.) test_stimulus_live.py read `rig/config/zenoh-*.json5`, deleted by the extraction; it takes the installed package's paths (`rig_nodes.zenoh.ROUTER_CONFIG` / `SESSION_CONFIG`) instead. Verified in the dev environment: `pixi run test` 62/62, `test-stimulus-live` 10/10. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…ency activation, bias_health from rig-nodes 0.3.0 (LAB-303) The dev environment uses git source dependencies since 2026-10-27 (the package is installed, not editable), so the sibling-checkout Zenoh URIs are gone. rig_nodes.bias_health and the derive hook ship in rig-nodes 0.3.0 (rig-nodes#1), which the candidate's re-pin PR installs. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…every run's bag A generic node (nothing FicTrac- or arena-specific) that publishes, once a second, CPU per core (window mean + busiest 100 ms sub-window, P/E-core class, sampler lateness, context switches / interrupts / DPCs), memory + page file + hard faults, log-volume free space, per-disk and per-adapter rates, NVIDIA GPU (incl. NVENC/NVDEC) and the top processes plus pinned groups (FicTrac, BIAS, MATLAB, browser, Foxglove, router, each graph node). The method is atlas-bench tools/cpu_sampler.py: cycle counters and one NtQuerySystemInformation call for all processes on Windows; GetIfEntry2 for network; NVML through ctypes; psutil for the rest and on Linux/macOS. - rig/nodes/host_sampler.py (no ROS) + host_monitor.py: rig-nodes material, prototyped here. Typed rig_interfaces/HostStats when installed (requested for rig-nodes 0.3.0; spec in rig/docs/host-stats-request.md), else String JSON with the same field names. - bridge_node.py records /rig/host into the bag; rig.py starts the monitor by default (--no-host-monitor, --host-monitor-args). - Cost on ww18: 0.9 % of one core (median); 4.9 % of the harness run's compressed payload, about 1.0 MB/h. Gated in the harness (<= 2 % of a core, <= 5 % of the payload). - tests/test-host-sampler.py (root suite, no ROS/psutil): fields and types against the .msg spec, rates, CPU mean vs peak, process selection, strict JSON. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…latform (#280, LAB-338) Manifest only; the re-lock with pixi 0.83 follows once the v0.3.0 channel is tagged. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…rig-nodes The generic host monitor (host_sampler.py, host_monitor.py, its unit test, design note and the HostStats message spec) belongs in reiserlab/rig-nodes as rig_nodes.host_monitor; it is being ported there from branch rig/host-monitor-port (f5cf5a6). What stays here is the arena rig's composition and gateway: - bridge_node.py records /rig/host into each run's bag (typed HostStats when rig_interfaces has it, else String). - rig.py starts `python -m rig_nodes.host_monitor --log-dir rig/runlogs --py-root rig/` when the installed rig_nodes ships it (a note otherwise); --no-host-monitor, --host-monitor-args. - test_ros2_poc.py runs it the same way and checks samples, fields, per-core sanity, pinned graph nodes, its CPU (<= 2 % of a core) and storage (<= 5 % of the compressed payload); skipped with an info line when the installed rig_nodes has no host monitor. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…`channel` feature (default/typed/viz), dev from git alone Channel read at the payload commit 5e20dee (the v0.3.0 tag sits one commit early; moving it is a one-liner for Michael). `dev` no longer includes `typed` and the rig_nodes pin is no longer a workspace dependency: one package name from a git source and the channel does not solve once the released version is ahead of rig-nodes main. Verified on the Mac with the sibling rig-nodes checkout renamed away: `pixi install --locked` (default + typed), check-typed lists the eight message types, `pixi run test` 59/59, `pixi run -e typed test` 62/62. Closes #280. LAB-338. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…7942b8 # Conflicts: # rig/nodes/rig.py # rig/tests/test_ros2_poc.py
…after the BIAS node, docstring) Coordinated with the session building the candidate with #269 (stimulus node): its hunks edit the --sim argparse neighbourhood, the --foxglove docstring line and the lines after nodes.append(bridge). Ours now sit after --sim-args, after the BIAS node append and after the --sim docstring example, so the two PRs merge without conflict. No behaviour change. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…on record §3.3 amendment) Michael's decision, 2026-10-28: everything specific to the 2P imaging experiments lives in reiserlab/scanimage-2026-bergamo `2p-rig-nodes/` (a pixi workspace in the rig-nodes layout on the public rig_nodes package; PROVENANCE.md there lists the source commit of every file). Copied with provenance, not rewritten in history. Removed here: - tiff2mcap (#273): rig/tools/tiff_run_to_mcap.py, check_tiff2mcap.py, rig/tests/test_tiff2mcap_runlog.py and the tiff2mcap / test-tiff2mcap / check-tiff2mcap tasks (patfile.py stays: arena code bound for pyDisplayTools) - tools/led_sweep_si.py, tools/led_sweep_quicklook.py - records: rig/SCANIMAGE-DAY.md, rig/scanimage-day-2026-10-16.md, docs/development/2p-scanimage-signal-logging.md, 2p-line-sync-rig-checklist.md - the arena composition's ScanImage node: rig.py --scanimage / --si-mock / --scanimage-args, the scanimage-node / scanimage-cmd / si-mock tasks, and the harness's ScanImage stand-in + its 5 checks (rig-nodes 0.4.0 drops the node, rig-nodes #4; re-pinning rig/ to 0.4.0 must not precede this change) Kept public (the boundary): the gateway's connector dispatch and its scanimage CONNECTOR_TOPICS entry, mcap_convert's connector derivation, the Studio's ScanImagePlugin and the emulator's --scanimage-cmds path; the 2P protocols and the Bergamo arena rig profile stay for now (comments repointed). Also: the dated §3.3 amendment in rig-software-repos-and-runners.md, the orientation skill's rig-nodes row, CLAUDE.md's rig line, and tests/test-flasher-catalog.js (its exception for the moved checklist), the expected harness counts in rig/WINDOWS-TEST.md. Tests on the Bergamo PC: root `pixi run test` 72/72 files; rig/ harness `pixi run test` 54/54 and `-e typed` 57/57 (main: 59 / 62, minus the 5 ScanImage checks; rig/WINDOWS-TEST.md updated), run on rig_nodes 0.2.1 with rig/pixi.lock unchanged (pixi 0.82 needed an unlocked install, the known #280 lock-format issue the next candidate re-locks). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ateway half of #261 / #266) The 2P rig's nodes are private since 2026-10-28 (reiserlab/scanimage-2026-bergamo `2p-rig-nodes/`; decision record §3.3 amendment). What stays public is the gateway's side, ported from #266's head (a4442ad, which includes #261) onto main: - bridge_node: the `sipm` connector (/sipm/cmd|reply|status in CONNECTOR_TOPICS, every bag's topic map and hello_ack), and the ScanImage frame-stream STREAM_TOPICS (/scanimage/frame_meta, frame_stats, preview/meta, preview/ch1-4 as sensor_msgs/CompressedImage) recorded on arrival, never derived into the run log - mcap_convert: /sipm/reply|status → {"type":"connector","src":"sipm",…} events - tests/test_gateway_registration.py (`pixi run test-gateway`, 5 checks): the gateway checks of #266's test_sipm (D1-D3) and #261's test_scanimage_frames (§G), which the private harnesses no longer carry Not ported: the nodes, their harnesses and tasks, the rig.py flags (all private now), and #261's rig_nodes._common SIGBREAK fix (_common lives in rig-nodes; the private package carries it as a shim). Tests on the Bergamo PC: test-gateway 5/5; harness `pixi run test` 59/59 (same as main); rig/pixi.lock unchanged (installed unlocked for the test, #280). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… recorded schema The converter's pure CDR decoder reads no nested message arrays (HostStats carries DiskStats[], NetStats[], GpuStats[], ProcStats[]), and the converter never needs /rig/host, so the typed check deserializes with ROS itself. The info line now says which type was recorded. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…+ G6 Foxglove dashboard, uFMF tools, ww14 run records (LAB-303)
…e rig repo; gateway keeps the 2P topics (LAB-303, LAB-319) # Conflicts: # rig/nodes/bridge_node.py # rig/nodes/rig.py # rig/pixi.toml # rig/tests/test_ros2_poc.py
mbreiser
force-pushed
the
release/2026-10-28
branch
from
October 28, 2026 22:41
037a5cc to
06bad57
Compare
Contributor
Author
|
✅ Validated Mac (osx-arm64, pixi 0.83), candidate SHA ww18 ( ww14 (okey-cokey rig), #269 on /next/: pixi 0.83 already; Bergamo PC, #284: root suite 72/72, harness 54/54 (+ 57/57 typed) on the removal branch, test-gateway 5/5; |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Release candidate
2026-10-28rc2 · candidate commit06bad57a02· built on main8d5b952a9cLive on Next: https://reiserlab.github.io/webDisplayTools/next/
2218d39e63e9dd994f74db51439a58e25914ebbcPromotion checklist
pixi run releasechecks)Validated 06bad57 …(pixi run tiers -- validate --rig … --by …). Any change to a constituent PR = rebuild (rc+1) = validate again.releaselists recent rig runs — advisory)fictrac-bridge/changed, rig PCsgit pull+ restart the bridgePromote with
pixi run release(merges with a merge commit — never squash — and tagsrelease-2026-10-28). Manifest:releases/2026-10-28.json.🤖 Generated with Claude Code release tooling (
scripts/tiers/candidate.py)