Conversation
chore: sync templates + faqs from main → develop (re-authored as @saadqbal)
ci: add FR-pass comment + FR gate callers
Documents the training and inference pipeline for all nine supported use cases so a user evaluating tracebloc can reproduce a run locally and compare metrics against what the platform reports. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The file started with U+200B (UTF-8 e2 80 8b) before the opening bracket, which broke JSON parsing and caused mint dev to fail with a YAML parser error. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Lives between hyperparameters and model optimization, where users are already configuring a run and want to understand what the platform does with their model and data. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
mint dev was parsing .github/pull_request_template.md as MDX and failing on the HTML comments. The file claims .github is auto-ignored but some CLI versions still scan it; listing it explicitly is harmless and unblocks local preview. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Expand the page from a high-level overview into a reference users can match runs against. Each of the nine supported use cases (image classification, object detection, semantic segmentation, keypoint detection, text classification, tabular classification, tabular regression, time series forecasting, time-to-event prediction) now has a consistent plain-English breakdown of preprocessing, train/val split, training step, validation step, cycle metrics, and inference output — including the platform-side defaults and reproduction-load-bearing details (mask handling for SS, OD-vs-YOLO image-size pinning, augmentation pipeline behavior, frozen-in-cycle-1 preprocessing state, scaled-vs-original-target metric scales, etc.). Also adds a shared "Experiment parameters" table grounded in the SDK's actual starting defaults (SGD, lr=0.001, batch_size=16, epochs=10, dynamic per-dataset validation_split) and a tightened "Reproducing a run locally" checklist. .mintignore: consolidate the .github/ entries into a single block so the dev server stops tripping over GitHub PR-template HTML comments. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Clarify that reproducing a tracebloc run locally — even with everything matched — will produce small numerical variation, with the major sources called out: hardware/CUDA differences, GPU non-determinism, library versions, data-loader worker timing, federated averaging between cycles, stateful layer behavior, and mixed-precision rounding. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* docs: add "Bring your data" step to homepage journey The homepage's "How It Works" jumped from "Create your AI workspace" straight to "Define a use case", with "Select datasets" framed as a sub-action of use-case definition. That inverted the actual journey (ingest first, then define a use case from your prepared datasets) and contradicted /create-use-case, which gets it right. Adds an explicit "Bring your data" step between workspace creation and use-case definition, and rewords Step 2 + the Next Steps card so "Select datasets" no longer reads as a UI action performed during use-case creation. Closes #28 Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * docs: fix missed inversion in homepage hero card The hero CardGroup at the top of the homepage said "Define datasets, set metrics, invite contributors" — same inversion as the "How It Works" steps and the "Next Steps" card, framing datasets as something configured during use-case creation. The previous commit fixed the two visible spots but missed this one. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
docs: add "How training works" page
The setup-guide's "What's Next" sent data owners straight from "client running" to "create a use case", with "select datasets" framed as a sub-action of use-case creation. Same inversion pattern as the homepage bug fixed in #29 — found during a broader audit. Reword the data-owner bullet to route through ingestion first, then into use-case creation, with "pick from prepared datasets" replacing "select datasets" to match the homepage wording. Closes #30 Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
keypoint_detection and semantic_segmentation templates exist on tracebloc/data-ingestors develop (both with READMEs, scripts, sample data, sample CSVs) but three places on the docs site still claimed they were "coming soon" or "still in flight" and pointed users at support@tracebloc.io instead of the actual templates. - prerequisites.mdx: replace "Template coming soon" with [Example] links for both Keypoint Detection and Semantic Segmentation, matching the format used by image_classification and object_detection. - templates.mdx: add the 2 missing rows to the Available templates table; delete the "still in flight" sentence underneath. Grep `coming soon|still in flight|template.*soon` across develop returns zero hits after this change (the legitimate "if your use case isn't yet supported, contact us" fallbacks on prerequisites.mdx:6 and define.mdx:14,30 are different copy and stay). Closes #32 Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
saadqbal
requested review from
LukasWodka,
aptracebloc,
divyasinghds,
saqlainsyed007 and
shujaatTracebloc
divyasinghds
approved these changes
May 11, 2026
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
3 participants
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
Related
Type of change
Test plan
Screenshots / recordings
Deployment notes
Checklist
Note
Low Risk
Low risk documentation-only changes; main risk is broken links/navigation or Mintlify build issues due to the large new MDX page and updated ignore rules.
Overview
Adds a new
Join a Use Casedoc page,How training works, detailing the end-to-end training/inference lifecycle and per–use-case preprocessing/metrics so runs can be reproduced locally.Updates docs to reflect newly available ingestion templates for keypoint detection and semantic segmentation (replacing “template coming soon” language) and adjusts onboarding copy to explicitly include a dataset-ingestion step before creating a use case.
Hardens the Mintlify dev/build experience by explicitly ignoring
.github/**in.mintignore, wires the new page intodocs.jsonnavigation, and fixes a stray character at the start ofevals.json.Reviewed by Cursor Bugbot for commit d0d10b9. Bugbot is set up for automated code reviews on this repo. Configure here.