cyberk-dev/Ultrareview
AI-powered code review: spec-aware Intent injection + GitNexus graph context + multi-step agent loop. Posts line-level comments on PRs.
View on GitHubTrust Signals
- Scorecard Score
- not yet scored
- Maintenance Recency
- Stale
- License
- None
Inputs
| name | description | required | default |
|---|---|---|---|
| ai-api-key | API key for AI proxy (cliproxy) | yes | — |
| ai-base-url | AI proxy base URL | no | https://api.openai.com |
| analysis-model | Model for deep analysis | no | gpt-5.4 |
| judge-model | Model for bug judging (should differ from analysis model) | no | gpt-5.2 |
| classify-model | Model for bug classification | no | gpt-5.4-mini |
| min-confidence | Minimum confidence threshold (0-1) | no | 0.7 |
| min-judge-score | Minimum judge score threshold (0-1) | no | 0.6 |
| timeout-minutes | Pipeline timeout in minutes (default 10, set 30 for thorough analysis) | no | 10 |
| feedback | Enable self-improving feedback loop (saves review data for prompt optimization) | no | true |
| gitnexus-enabled | Enable GitNexus code graph context (requires npx gitnexus in PATH) | no | true |
| gitnexus-secret-scan | Scan .gitnexus/ index for leaked secrets after indexing (fails with warning, never blocks PR) | no | true |
| intent-enabled | Enable Intent injection: inject PR title/body and classified spec files (OpenSpec, CK-Plan, docs) at the top of the analyzer prompt | no | true |
| flow-diagram-enabled | Embed a Mermaid flowchart of the bot's understanding at the top of the PR review summary (one extra cheap LLM call per PR) | no | true |
Outputs
no outputs