Add image_count metadata field to dataset entries

[Copilot]

Track the number of screenshot images in each dataset instance's problem statement directory as metadata.

Changes

• EntryMetadata: Added image_count: int | None = None field
• DatasetEntry.count_images(): Method to count images in problem statement dir
• count_images_in_directory(): Utility function supporting .png, .jpg, .jpeg, .gif, .webp, .bmp, .svg
• bcbench dataset update-metadata: CLI command to calculate and update image counts across all entries
• Dataset: All 55 entries populated with actual image counts (0-20 range)

Usage

# Preview changes
bcbench dataset update-metadata --dry-run

# Apply updates
bcbench dataset update-metadata

entry = load_dataset_entries(path, entry_id="microsoftInternal__NAV-227219")[0]
entry.metadata.image_count  # 8
entry.count_images()  # Recalculate from disk

Original prompt

Calculate number of images (screenshots) included in each dataset instances. It should be included as another metadata field, with type int | None.

You tasks: calculate the existing ones programatically

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.