The bus factor for the AI era. Map team knowledge blindspots before someone leaves.
Project description
blindspot
The bus factor for the AI era. Map your team's knowledge blindspots before someone leaves.
🚧 Pre-alpha. Codename: Blindspot. APIs and metrics will change.
Why
AI coding tools made engineering teams faster. But faster does not mean understood. Codebases now carry a new kind of risk: services that shipped quickly, owned by one person, reviewed by no one in depth — and that one person can leave tomorrow.
Existing tools measure velocity. blindspot measures resilience.
What it measures
- Ownership concentration — who actually understands each part of the codebase, weighted by recency and review depth
- Bus factor per service / folder — how many people would need to leave before knowledge is critically lost
- Review lineage — who reviews what, and where reviewer redundancy is dangerously thin
- Departure simulation — "If two senior devs on Payment leave next month, what coverage do we lose?"
- Knowledge decay — code volatility and contributor drift, projected 30 / 60 / 90 days forward
Quick start
pip install blindspot
blindspot scan /path/to/repo --output blindspot_report.html
blindspot simulate --person alice@example.com /path/to/repo
Output is a single self-contained HTML file. No server, no signup, no telemetry.
A rule-based narrator ships in the package — pip install blindspot && blindspot scan /repo produces a full report with an executive summary out of the box. Configure a cloud LLM key (Anthropic or OpenAI) for richer prose; the report itself shows you how.
Documentation
Full end-to-end documentation lives in docs/ — the algorithms (with formulas and parameters), the architecture, the CLI reference, configuration, and how to read every section of the report.
Design principles
- Service-first, not person-first. Default views show service-level risk. Individual views require explicit, justified access.
- Evidence over inference. AI-usage signals come from official telemetry (e.g. GitHub Copilot Usage API) when available — not from guessing.
- Reports, not surveillance. blindspot answers "is this service fragile?", not "is this person slacking?".
Roadmap
| Phase | Surface | Status |
|---|---|---|
| 1 | CLI + static HTML report | In progress |
| 1 | GitHub Action + Checks API output | Planned |
| 2 | Self-hosted dashboard | Planned |
| 2 | AI signal layer (Copilot Usage API) | Planned |
| 3 | Slack / Jira / incident integration | Planned |
License
MIT. See LICENSE.
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file blindspot-0.0.3.tar.gz.
File metadata
- Download URL: blindspot-0.0.3.tar.gz
- Upload date:
- Size: 155.7 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.12.13
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
cf5e64437c4c2b61bc03a8d1b6fd222fcfbfe82a70464eb37948dd3244827c2c
|
|
| MD5 |
8714430767c051fee232ce6d8dffd110
|
|
| BLAKE2b-256 |
202a400f20344caf0c96f086421ba6c84bf4b46089ad3c346f0241da6e10cdac
|
File details
Details for the file blindspot-0.0.3-py3-none-any.whl.
File metadata
- Download URL: blindspot-0.0.3-py3-none-any.whl
- Upload date:
- Size: 128.2 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.12.13
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
c50a244e617fe9acd4690fb89ba323c9618f0a71e78458dbb2a7fbd9e4004401
|
|
| MD5 |
d6429433e2b8a0ee569e28ea28254975
|
|
| BLAKE2b-256 |
4cd1ccecde39c9d8ee97f5cb5ace90f6b4ee85b5d5f0425b5853ce4b8fee7395
|