| title | SWE-Wiki |
|---|---|
| emoji | 📙 |
| colorFrom | blue |
| colorTo | purple |
| sdk | gradio |
| sdk_version | 5.50.0 |
| app_file | app.py |
| hf_oauth | true |
| pinned | false |
| short_description | Track GitHub wiki statistics for SWE assistants |
SWE-Wiki ranks software engineering assistants by their real-world GitHub wiki editing activity.
No benchmarks. No sandboxes. Just real wiki edits tracked from public repositories.
Most AI coding assistant benchmarks use synthetic tasks and simulated environments. This leaderboard measures real-world activity: how many wiki pages is the assistant editing? How active is it across different projects? Is the assistant's usage growing?
If an assistant is consistently editing wikis across different projects, that tells you something no benchmark can.
Key metrics from the last 180 days:
Leaderboard Table
- Assistant Name: Display name of the assistant
- Website: Link to the assistant's homepage or documentation
- Total Wiki Edits: Total number of wiki pages edited by the assistant
Monthly Trends
- Wiki edit volume over time (bar charts)
- Activity patterns across months
We focus on 180 days to highlight current capabilities and active assistants.
Data Collection We mine GitHub activity from GHArchive, tracking:
- Wiki pages edited by the assistant (
GollumEventdata)
Regular Updates Leaderboard refreshes daily
Community Submissions
Anyone can submit an assistant. We store metadata in SWE-Arena/bot_metadata and results in SWE-Arena/leaderboard_data. All submissions are validated via GitHub API.
Planned improvements:
- Repository-based analysis (which repos are assistants documenting)
- Extended metrics (wiki page types, edit actions)
- Organization and team breakdown
- Wiki editing patterns (page creations, updates, deletions)
Open an issue for bugs, feature requests, or data concerns.