A pgpm workspace for building agentic, database-driven AI applications for the City of San Francisco.
# @sf-ai/rag-functions
RAG functions for import, export, and vector search.
## Overview
This package provides SQL functions for working with RAG collections:
- **Import/Export**: JSON format for portable collection data
- **Search**: Cosine similarity search using pgvector
- **CSV Export**: Flat format for embeddings
## Functions
### `rag.export_collection_json(collection_id uuid)`
Export a complete collection with all documents, chunks, and embeddings as JSON.
```sql
SELECT rag.export_collection_json('collection-uuid');
```
### `rag.import_collection_json(data jsonb, overwrite boolean)`
Import a collection from JSON format.
```sql
SELECT rag.import_collection_json('{...}'::jsonb, false);
```
### `rag.search_similar(collection_id uuid, query_vector vector, ...)`
Search for similar chunks using cosine distance.
```sql
SELECT * FROM rag.search_similar(
'collection-uuid',
'[0.1, 0.2, ...]'::vector,
'text-embedding-3-small', -- model name (optional)
10, -- limit
0.5 -- distance threshold (optional)
);
```
Also supports searching by collection name:
```sql
SELECT * FROM rag.search_similar(
'my-collection',
'[0.1, 0.2, ...]'::vector
);
```
### `rag.export_embeddings_csv(collection_id uuid, model_name text)`
Export embeddings in CSV-friendly format.
```sql
SELECT * FROM rag.export_embeddings_csv('collection-uuid', 'text-embedding-3-small');
```
## Dependencies
- `@sf-ai/rag-core`
## License
[SF License](https://github.com/city-of-san-francisco/license)