kalmantic/PeakInfer
Analyze LLM inference points in your code for cost, latency, throughput, and reliability issues
View on GitHubTrust Signals
- Scorecard Score
- not yet scored
- Maintenance Recency
- Stale
- License
- None
Inputs
| name | description | required | default |
|---|---|---|---|
| path | Path to analyze | no | ./src |
| peakinfer-token | PeakInfer API token (required). Get your token at https://peakinfer.com/dashboard. Pass via secrets.PEAKINFER_TOKEN. | yes | — |
| github-token | GitHub token for PR comments. Defaults to github.token if not provided. | no | — |
| runtime-source | Runtime data source: helicone, langfuse | no | — |
| runtime-api-key | API key for runtime source | no | — |
| events-file | Path to runtime events file (JSONL) | no | — |
| events-map | Field mappings for non-standard formats | no | — |
| include-benchmarks | Compare to InferenceMAX benchmarks | no | true |
| benchmark-framework | Benchmark framework: api, vllm, tgi | no | api |
| evals-source | Evals source: braintrust, evalica | no | — |
| evals-api-key | API key for evals source | no | — |
| fail-on-critical | Fail the action if critical issues are found | no | false |
| comment-mode | When to post comments: always, on-issues, never | no | always |
| show-enhancement-prompts | Show prompts for adding more layers | no | true |
Outputs
| name | description |
|---|---|
| verdict | PASS, OK, REVIEW, BLOCK, PAUSED, SKIP, ERROR |
| critical-count | Number of critical issues |
| warning-count | Number of warnings |
| drift-count | Number of drift detections |
| layers-used | Comma-separated list of active layers |
| credits-used | Credits consumed (paid mode) |
| credits-remaining | Credits remaining (paid mode) |
| inference-points | Number of inference points found |