r0han7/gitmemory
Git-native AI memory: recall relevant decisions/gotchas on new PRs & issues, and distill new memories on merge — powered by a local model via LM Studio.
View on GitHubTrust Signals
- Scorecard Score
- not yet scored
- Maintenance Recency
- Stale
- License
- None
Inputs
| name | description | required | default |
|---|---|---|---|
| mode | Operation: 'recall' (comment on new activity) or 'ingest' (distill + reconcile on merge/close). | yes | — |
| source | Provenance label for ingest, e.g. 'PR#231'. Required for ingest. | no | "" |
| content | Text to recall against or ingest from (PR/issue title + body + discussion). | no | "" |
| store | Path to the in-repo memory file. | no | .gitmemory/memories.json |
| base-url | LM Studio OpenAI-compatible base URL. | no | http://localhost:1234/v1 |
| chat-model | Chat model identifier loaded in LM Studio. | no | local-model |
| embedding-model | Embedding model identifier loaded in LM Studio. | no | text-embedding-nomic-embed-text-v1.5 |
| min-score | Minimum cosine similarity to surface a memory on recall. | no | 0.15 |
Outputs
| name | description |
|---|---|
| comment | Markdown comment produced by recall (empty if nothing relevant). |