2.3.48 Satellite Airweave

Airweave

Handle: airweave
URL: http://localhost:34481

A lightweight, fast, and extensible vector database for AI applications. Airweave provides persistent storage and efficient similarity search for embeddings, making it ideal for RAG, chat history, and knowledge base use cases.

Starting

# [Optional] Pre-build the image
harbor build airweave

# Start the service
# --open is optional, to open in browser
harbor up airweave --open

• Airweave exposes a REST API and a web UI for managing collections and documents
  • Run harbor url airweave-backend to get API URL
• Data is persisted in the services/airweave/data folder by default, configurable via airweave.workspace

Usage

You will mostly interact with Airweave via its Web UI or API. Run harbor open to access the UI.

# Open Airweave Web UI in the browser
harbor open airweave

# Get a URL
harbor url airweave

# Get a QR for the phone
harbor qr airweave

Configuration

See the official documentation for more details.

Following options can be set via harbor config:

# The port on the host where Airweave endpoint will be available
HARBOR_AIRWEAVE_HOST_PORT      34481

# Ports on the host for other services in Airweave group
HARBOR_AIRWEAVE_BACKEND_HOST_PORT    34482
HARBOR_AIRWEAVE_POSTGRES_HOST_PORT   34483
HARBOR_AIRWEAVE_REDIS_HOST_PORT      34484
HARBOR_AIRWEAVE_QDRANT_HOST_PORT     34485
HARBOR_AIRWEAVE_EMBEDDINGS_HOST_PORT 34486

# Git ref to build the image from
# You can adjust the branch/tag/commit to use by adjusting part after the #
HARBOR_AIRWEAVE_GIT_REF        https://github.com/airweave-ai/airweave.git#main

# Location of the Airweave workspace on the host
# This is where service data is stored
# Should be either relative to $(harbor home) or an absolute path
HARBOR_AIRWEAVE_WORKSPACE      ./services/airweave/data

# Required by the upstream backend (change from defaults for real deployments)
HARBOR_AIRWEAVE_STATE_SECRET          harbor-airweave-state-secret-change-me-32ch
HARBOR_AIRWEAVE_SVIX_JWT_SECRET       harbor-airweave-svix-jwt-secret-change-me

# Embedding pipeline. local_minilm runs fully locally via the bundled
# text2vec container; EMBEDDING_DIMENSIONS must match the embedder (384
# for local_minilm). Alternatives: openai_text_embedding_3_small/large,
# mistral_embed (require the matching API key).
HARBOR_AIRWEAVE_DENSE_EMBEDDER        local_minilm
HARBOR_AIRWEAVE_EMBEDDING_DIMENSIONS  384
HARBOR_AIRWEAVE_SPARSE_EMBEDDER       fastembed_bm25

Note: the backend requires GET /health on the embeddings service, which the upstream-specified text2vec image does not serve — Harbor routes it through a small nginx shim (airweave-embeddings-proxy) that rewrites /health to /.well-known/ready.

See environment configuration guide to set arbitrary environment variables for the service.