Merge develop to main -- Annotate Those Steps presentation#145
Merged
Conversation
* Fix BYOK vision provider mismatch and update default models - Fix: annotation provider (anthropic) was passed to vision agent in BYOK mode, causing provider mismatch crash on annotate-image - Fix: vision_provider not passed through LocalExecutionBackend - Update vision/eval defaults to qwen/qwen3.5-122b-a10b with alibaba - Keep annotation model as claude-haiku-4.5 with anthropic provider - Add 504 gateway timeout handler to CLI client - Update docker-compose, .env.example, deploy docs, frontend display - Bump version to 0.7.9.dev0 Closes #140 * Address PR review findings - Add vision_provider field to ImageAnnotationRequest for full decoupling - Wire vision_provider_override through all 4 API code paths - Add X-OpenRouter-Vision-Provider to CORS allowed headers - Fix stale docstring: GPT-OSS-120B -> Qwen3.5-122B in create_openrouter_workflow - Fix stale eval_provider description: None -> alibaba in config - Update provider examples: Cerebras/deepinfra -> anthropic/alibaba - Align .env.example annotation model with code defaults (claude-haiku-4.5) - Make vision provider comments durable (architectural intent, not config) * Address silent-failure-hunter findings - Add logging.exception() to BYOK agent init error paths - Include server-provided detail in 503 handler instead of discarding
This reverts commit 11a7579.
# Conflicts: # pyproject.toml # src/version.py
Adds the "Annotate those steps -- how to curate analysis-ready MoBI datasets" presentation (20 slides) under presentations/annotate-those-steps/. - Reveal.js-based deck driven by annotate-those-steps.json (validated) - 14 hand-coded SVG icons (assets/icons) for the MoBI gap narrative - 2 AI-generated nature-style figures (assets/figures) for slides 4 & 6: experiment-scene and gait-cycle, substrate via gpt-image-2 + programmatic label overlay - Case study data: ds003739 (Peterson and Ferris 2018) events.tsv/json for all four sessions plus the source PDF - OSA-style landing pages: presentations/annotate-those-steps.md (with embedded iframe, 171-word abstract, talk outline, links) and presentations/index.md (table of presentations) Positions HEDit v0.6 strictly as natural-language -> HED tag converter; hedify (paper PDF reader, GUI, auto-suggest) is explicitly marked as roadmap.
Removes the vendored reveal.js runtime (KaTeX, Mermaid, Prism bundles, fonts), the presentation.html viewer, the casual-vibers logo, and the top-level index.md from presentations/annotate-those-steps/. The viewer and runtime now live in the annot-garden/documentation site. What remains here is the talk source: annotate-those-steps.md (expanded blog body), annotate-those-steps.json (deck source), 14 hand-coded SVG icons, 2 AI-generated figures with their substrate PNGs + labels, the ds003739 events files, and the Peterson and Ferris 2018 PDF. Also regenerates the experiment-scene substrate with a three-quarter oblique view so the mediolateral pull cable is geometrically correct (perpendicular to the walking direction, not anterior-posterior). Label overlay on the new substrate is deferred to a follow-up.
Slide 4 (experiment-scene): re-runs the label overlay on the corrected mediolateral substrate. Drops the in-figure panel headers (the slide title already announces them), keeps EEG cap, safety harness, F_pull arrow on the cable, motor + load cell, narrow beam, Oculus Rift DK2, and 20° visual rotation as crisp SVG callouts on top of the AI substrate. Slide 2 (MoBI stack): replaces the hand-coded stick-figure subject with a /ai-full-figure-generated subject wearing the full sensor suite (EEG cap, VR headset, harness, EMG patches, motion-capture markers). Hand- authors the surrounding SVG with six modality cards (EEG, EMG, motion capture on the left; VR / visual stim, load cell + motor, task / behavior on the right), dashed connectors from each card to the relevant body region, and the bottom red callout about what reaches the shared events.tsv. Theme bible (agi-theme.json) and label overlay JSONs kept under assets/figures/ so the figures are regenerable.
Slide 4: re-runs the label overlay with correct anchor pixel-coords for the new mediolateral substrate. EEG cap, safety harness, F_pull, motor + load cell, narrow beam, Oculus Rift DK2, and 20° visual rotation now actually point to their features. Slide 2: drops the card-based composition. The card design pointed at modalities (Load cell + motor, Task / behavior) that have no anatomical anchor in the figure, which was misleading. New layout uses simple text labels with dashed connector lines from the text edge to a marker dot on the actual visible feature: EEG cap, EMG (neck + leg), safety harness, motion-capture markers, narrow beam, treadmill belt, VR headset, harness D-ring (where the load-cell pull attaches during a trial), calf EMG patch, ankle marker. Slides 13 and 16 align the deck with HEDit v0.6 actual scope -- one detailed natural-language description per event level, no composite tags for co-occurring events. Slide 16 demo input now shows two paragraph-grade descriptions (pull_left and rotate_cw) with explicit recipe in the callout. Source markdown (annotate-those-steps.md) updated to match.
Deploying hedit with
|
| Latest commit: |
66a55f8
|
| Status: | ✅ Deploy successful! |
| Preview URL: | https://badffa43.hedit.pages.dev |
| Branch Preview URL: | https://develop.hedit.pages.dev |
Codecov Report✅ All modified and coverable lines are covered by tests. 📢 Thoughts on this report? Let us know! |
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
1 participant
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.
Summary
Brings the Annotate Those Steps MoBI annotation talk into main, plus an interim BYOK vision-provider fix already on main is reconciled.
presentations/annotate-those-steps/with 20-slide deck, 14 SVG icons, 2 AI-generated figures (substrate + label overlay), case-study event files, and the Peterson and Ferris 2018 PDFDevelop commits since 0.7.9a1:
Test plan