Configuration

All options use the YAPA_ prefix and are set as environment variables in your MCP host config. Under the DSH plugin, the same values live in the cordis row config: or the hot-reloaded yapa: section of ~/.dsh/settings.yaml (camelCase: chromaUrl, syncEnabled, …). See .env.example for the full list.

Variable	Description	Default
`YAPA_STORAGE`	`chroma` \| `local` (embedded store, no server)	`chroma` for MCP; the DSH plugin defaults to `local`
`YAPA_LOCAL_STORE_PATH`	Root dir for the embedded store	`~/.local/share/yapa/store`
`YAPA_CHROMA_URL`	ChromaDB server URL (when `YAPA_STORAGE=chroma`)	`http://localhost:8000`
`YAPA_USERNAME`	Username for task ID prefixes	`user`
`YAPA_EMBEDDING_PROVIDER`	Embedding provider — `chromadb` is in-process MiniLM (zero-config, no server call); `fireworks`/`openai`/`voyage`/`ollama` use HTTP APIs	`chromadb`
`YAPA_SALIENCE_DECAY_RATE`	Daily decay multiplier	`0.98`
`YAPA_SALIENCE_RANKING_WEIGHT`	How much salience influences retrieval ranking (0.0 = pure distance, higher = salience-dominant)	`0.3`
`YAPA_TRAINING_PIPELINE`	Expose the 18 ML-ops tools (curation/buckets/training/eval/adapter)	`false`
`YAPA_SYNC_ENABLED`	Enable remote sync	`false`
`YAPA_SYNC_DATABASE_URL`	PostgreSQL connection string	(none)
`YAPA_SYNC_INTERVAL_MS`	Background sync interval in ms	`300000` (5 min)
`YAPA_SYNC_SIMILARITY_THRESHOLD`	Cosine similarity threshold for dedup	`0.95`

ML-ops configuration (curation models, bucket thresholds, training backend, eval holdout, …) lives in docs/training-pipeline.md.

yapa

AI Analysis

Install

README

Configuration