README
¶
pgvector — RAG with PostgreSQL + pgvector
Interactive RAG chat with pgvector as a persistent vector store. The database schema is created automatically on first run via MigratePGVector (idempotent).
Quick start
1. Start PostgreSQL with pgvector:
docker compose up -d
2. Run the example:
cd examples/pgvector && cp .env.example .env && go run .
For mock mode this is sufficient. For a real LLM, set LLM_PROVIDER=ollama|openai|anthropic and the corresponding keys.
Environment variables
| Variable | Default | Description |
|---|---|---|
LLM_PROVIDER |
mock |
LLM provider (mock, ollama, openai, anthropic) |
EMBEDDING_DIM |
1536 |
Vector dimension (must match the model) |
PGVECTOR_DSN |
— | Required. DSN for connecting to PostgreSQL |
TABLE_NAME |
draftrag_chunks |
Table name for storing chunks |
For LLM_PROVIDER=ollama:
| Variable | Default | Description |
|---|---|---|
OLLAMA_HOST |
http://localhost:11434 |
Ollama URL |
OLLAMA_EMBED_MODEL |
nomic-embed-text |
Embedding model |
OLLAMA_LLM_MODEL |
llama3.2 |
LLM model |
For LLM_PROVIDER=openai:
| Variable | Default | Description |
|---|---|---|
OPENAI_API_KEY |
— | Required. API key |
OPENAI_BASE_URL |
https://api.openai.com |
Base URL |
OPENAI_EMBED_MODEL |
text-embedding-3-small |
Embedding model |
OPENAI_LLM_MODEL |
gpt-4o-mini |
LLM model |
For LLM_PROVIDER=anthropic:
| Variable | Default | Description |
|---|---|---|
ANTHROPIC_API_KEY |
— | Required. API key |
ANTHROPIC_LLM_MODEL |
claude-3-5-sonnet-latest |
LLM model |
Vector dimension
The dimension must match the embedding model used:
| Model | EMBEDDING_DIM |
|---|---|
text-embedding-ada-002 |
1536 |
text-embedding-3-small |
1536 |
text-embedding-3-large |
3072 |
nomic-embed-text (Ollama) |
768 |
If the dimension changes after the first run, you need to recreate the table or use a different TABLE_NAME.
Migrations
MigratePGVector creates the table and index on first run. Re-running is safe — migrations are idempotent.
For production, it is recommended to apply SQL migrations as a separate deployment step:
psql $PGVECTOR_DSN -f pkg/draftrag/migrations/pgvector/0000_pgvector_extension.sql
psql $PGVECTOR_DSN -f pkg/draftrag/migrations/pgvector/0001_chunks_table.sql
psql $PGVECTOR_DSN -f pkg/draftrag/migrations/pgvector/0002_metadata_and_indexes.sql
Local mode (Ollama)
ollama pull nomic-embed-text
ollama pull llama3.2
PGVECTOR_DSN="postgres://draftrag:draftrag@localhost:5432/draftrag?sslmode=disable" \
EMBEDDING_DIM=768 \
LLM_PROVIDER=ollama \
OLLAMA_HOST=http://localhost:11434 \
go run ./examples/pgvector/
Documentation
¶
Overview ¶
@sk-task docs-and-examples#T2.2: pgvector example — RAG-чат с PostgreSQL + pgvector (AC-001). Использует публичный API draftrag напрямую. Shared только для mock/print.
Быстрый старт с Docker:
docker compose up -d PGVECTOR_DSN="postgres://draftrag:draftrag@localhost:5432/draftrag?sslmode=disable" \ go run ./examples/pgvector/
Source Files