TL;DR: This repository is not the official Microsoft project. It’s my personal fork used to showcase a new Evaluation tab concept I built as a proof-of-concept (POC).
If you’re looking for the real project, go here → microsoft/AI-Dev-Gallery.
I cloned the upstream repo and opened a couple of symbolic PRs in this fork to document and demonstrate the work I did (mostly front-end). Those PRs capture the design, UX, and implementation approach for an Evaluation experience inside AI Dev Gallery.
- 🤖 Built with AI coding agents: I used Claude Code and GitHub Copilot to accelerate scaffolding, refactors, and test stubs during this POC.
- 📦 This fork is for show & tell only (a POC / “vibe-coded” exploration).
- 🧭 For source of truth, docs, issues, and downloads, use the official repo linked above.
The two PRs below summarize the core of this POC and include descriptions/screenshots of the feature set.
- PR #5 – Evaluation Core Feature with Evaluation Wizard Dialog
Link: #5
Adds a complete Evaluation workflow and the new Evaluation tab foundation.- 6-step Evaluation Wizard (workflow → type → model → dataset → metrics → review)
- Three evaluation modes: Test Model, Evaluate Responses, Import Results
- Evaluation management views (card + list), search/filter, bulk selection, and status tracking
- Local data persistence and sample data generation
- Keyboard navigation & screen-reader support (accessibility improvements)
Evaluation Home
Step 1 - Select Eval Type
Step 2 - Choose Workflow
Step 3 - Configure Model with Image Description API Chosen
Step 4 - Dataset Input
Step 5 - Evaluation Criteria
Step 6 - Evaluation Confirmation
- PR #6 – Evaluation Insights, Analytics, and Comparison Features
Link: #6
Builds on the core to add analytics and comparison.- Insights view with score distributions and criteria breakdowns
- Compare up to 4 evaluations side-by-side
- Basic statistics (mean/median/std dev), trends, and item-level analysis
- Export evaluation results (JSON/CSV)
- Continued focus on performance and accessibility
Evaluation Insights
Printable Report
Compare Evaluations
Tip: In each PR, click the “Files changed” tab to review the implementation details.
⚠️ This fork isn’t maintained as a product. Please don’t open issues here for upstream bugs.- 👉 Use microsoft/AI-Dev-Gallery for releases, docs, and contributions.
- Upstream project: Microsoft — AI Dev Gallery
- AI coding agents used in this POC: Claude Code and GitHub Copilot
- My contributions in this fork are for demonstration purposes only and do not imply Microsoft endorsement.
If you want to build this fork locally, follow the official instructions from the upstream repo. The authoritative README and setup steps live there.
If you have questions about this POC or want to discuss the Evaluation feature, feel free to reach out via GitHub.