Vector Indexing

Agno-Go ships a pluggable VectorDB interface with a Chroma provider by default and an optional Redis provider (build tag redis). A simple migrations CLI is included.

Providers

• ChromaDB (default): HTTP client via chroma-go
• Redis (optional): build with -tags redis for a minimal provider without RediSearch

Migrations CLI

Create or drop collections:

# Chroma
go run ./cmd/vectordb_migrate --action up --provider chroma --collection mycol \
  --chroma-url http://localhost:8000 --distance cosine

# Redis (optional)
go run -tags redis ./cmd/vectordb_migrate --action up --provider redis \
  --collection mycol --chroma-url localhost:6379

--distance: l2|cosine|ip (defaults to l2 for Chroma; cosine in Redis provider).

Embeddings

Use OpenAI or VLLM embeddings with Chroma provider:

embed, _ := openai.New(openai.Config{APIKey: os.Getenv("OPENAI_API_KEY"), Model: "text-embedding-3-small"})
db, _ := chromadb.New(chromadb.Config{BaseURL: "http://localhost:8000", CollectionName: "docs", EmbeddingFunction: embed})
// db.Add / db.Query

VLLM (OpenAI-compatible /v1/embeddings):

emb, _ := vllm.New(vllm.Config{BaseURL: "http://localhost:8000/v1", Model: "bge-base"})
db, _ := chromadb.New(chromadb.Config{BaseURL: "http://localhost:8000", CollectionName: "docs", EmbeddingFunction: emb})

Notes

• Redis provider and tests are tag-gated: -tags redis
• Optional dependencies do not affect users who don't enable them