Qdrant Setup Guide¶

Vector database options for semantic search.

Option 1: Local Qdrant via Docker¶

# From repo root
docker compose up -d qdrant
# Verify
curl http://localhost:6333/collections
# Configure
QDRANT_URL=http://localhost:6333
QDRANT_API_KEY=

Data persists in ./qdrant_data if configured in compose.

Option 2: Local Qdrant via WSL 2¶

Run the container in WSL Docker, find the WSL IP (e.g., ip addr show eth0), then configure:

QDRANT_URL=http://<wsl-ip>:6333
QDRANT_API_KEY=

Option 3: Qdrant Cloud (Free Tier) — recommended for dev¶

Create free 1 GB cluster at https://cloud.qdrant.io (≈100k vectors, ~2.6k tools at 384-dim).
Configure:

QDRANT_URL=https://<cluster>.aws.cloud.qdrant.io
QDRANT_API_KEY=<key>

Create collection toolweaver_tools (384 dims, cosine, dense vector default; sparse not required by code).

Option 4: Azure Container Instances (self-hosted cloud)¶

az group create --name rg-toolweaver --location eastus
az container create --resource-group rg-toolweaver --name qdrant-instance --image qdrant/qdrant:latest --cpu 1 --memory 2 --ports 6333 6334 --dns-name-label toolweaver-qdrant --location eastus --restart-policy Always
az container show --resource-group rg-toolweaver --name qdrant-instance --query "ipAddress.fqdn" -o tsv

Config:

QDRANT_URL=http://toolweaver-qdrant.eastus.azurecontainer.io:6333
QDRANT_API_KEY=<key-if-set>

Capacity & monitoring¶

Free tier: 1 GB (~100k vectors @384 dims). Upgrade when you need more.
Use Qdrant Cloud dashboard metrics; for self-hosted, expose /metrics if desired.

ToolWeaver behavior¶

VectorToolSearchEngine falls back to in-memory search if Qdrant is unavailable; index the catalog after configuring Qdrant for best results.