marco-jardim/Changelog Analyze Action
Analyzes changeset data using LLMs to produce executive-friendly insights (business impact, risks, highlights)
View on GitHubTrust Signals
- Scorecard Score
- not yet scored
- Maintenance Recency
- Stale
- License
- None
Inputs
| name | description | required | default |
|---|---|---|---|
| changeset_path | Path to the changeset.v1.json file produced by changelog-collect-action | yes | — |
| provider | LLM provider to use. One of: fireworks, openai, anthropic, ollama | yes | — |
| model | Model name/ID to use. Defaults per provider: fireworks: accounts/fireworks/models/llama-v3p1-70b-instruct openai: gpt-4o-mini anthropic: claude-3-5-haiku-20241022 ollama: llama3.2 | no | "" |
| api_key | API key for the LLM provider. Not required for ollama. | no | "" |
| base_url | Base URL override. Used for ollama (default: http://localhost:11434) or custom endpoints. | no | "" |
| max_tokens | Maximum tokens for LLM response | no | 2000 |
| prompt_profile | Prompt profile to use. One of: executive, technical, brief. Default: executive | no | executive |
| language | Output language. Supported: en, pt-BR. Default: en | no | en |
| fallback_on_error | If true, falls back to heuristic analysis when LLM fails. Default: true | no | true |
| group_by_date | If true, generate separate LLM insights for each calendar day of commits. Default: true | no | true |
| output_path | Path where insights.v1.json will be written. Default: insights.v1.json | no | insights.v1.json |
Outputs
| name | description |
|---|---|
| insights_path | Absolute path to the generated insights.v1.json file |
| provider_used | The LLM provider that actually generated the insights (may differ if fallback was used) |
| model_used | The model that was used for analysis |
| fallback_used | Whether the heuristic fallback was used instead of the LLM |