Weave CLI

git clone https://github.com/maximilien/weave-cli.git
cd weave-cli
./build.sh
# Binary available at bin/weave

# Interactive configuration - fastest way to get started
weave config create --env

# Follow prompts to enter:
# - WEAVIATE_URL
# - WEAVIATE_API_KEY
# - OPENAI_API_KEY

# Verify setup — diagnose everything in one command
weave doctor

# Or check just database health
weave health check

# List configured databases
weave vdb list

# Show detailed database info
weave vdb info weaviate-cloud

# Install sentence-transformers (one time)
pip install sentence-transformers

# Create collection with OSS embedding model
weave docs create MyDocs_OSS data/documents.pdf \
  --embedding sentence-transformers/all-mpnet-base-v2

# Re-embed existing collection (20x faster than re-ingestion!)
weave collection reembed MyCollection \
  --new-embedding sentence-transformers/all-mpnet-base-v2 \
  --output MyCollection_OSS

# Query automatically uses collection's embedding model
weave cols query MyDocs_OSS "search query" --top-k 5

# Compare OpenAI vs OSS embeddings
weave collection compare MyCollection MyCollection_OSS \
  --query "search query" \
  --report comparison.md

# List all available embedding models
weave embeddings list

# Output shows providers with dimensions and support status:
# OpenAI Models:
#   text-embedding-3-small     (1536 dims) ✅
#   text-embedding-3-large     (3072 dims) ✅
#   text-embedding-ada-002     (1536 dims) ✅
#
# sentence-transformers (OSS):
#   all-mpnet-base-v2          (768 dims)  ✅ Recommended
#   all-MiniLM-L6-v2           (384 dims)  ✅ Fast
#
# Ollama (Local):
#   nomic-embed-text           (768 dims)  ✅
#   mxbai-embed-large          (1024 dims) ✅

# Start MinIO (one time)
./scripts/start-minio.sh

# Ingest images with automatic external storage
weave docs create AuctionImages data/images/ \
  --image-storage minio \
  --minio-bucket weave-images \
  --milvus-local

# Bucket auto-created if it doesn't exist (v0.9.22+)
# Images >47KB automatically uploaded to MinIO
# Thumbnails stored in Milvus for fast preview
# Full-resolution URLs: http://localhost:9000/weave-images/...

# Set AWS credentials (or use --s3-access-key/--s3-secret-key)
export AWS_ACCESS_KEY_ID=your-key
export AWS_SECRET_ACCESS_KEY=your-secret

# Ingest with S3 storage
weave docs create ProductCatalog images/ \
  --image-storage s3 \
  --s3-bucket my-product-images \
  --s3-region us-west-2 \
  --milvus-cloud

--image-storage s3|minio|local  # Storage backend
--s3-bucket <name>              # S3 bucket name
--s3-region <region>            # AWS region (default: us-east-1)
--minio-bucket <name>           # MinIO bucket name
--minio-endpoint <host:port>    # MinIO endpoint (default: localhost:9000)
--local-storage-path <path>     # Local storage directory (default: ./storage)
--store-pdf                     # Also store PDFs in external storage

# Ingest PDF with external storage
weave docs create KnowledgeBase document.pdf \
  --image-storage minio \
  --minio-bucket my-docs \
  --store-pdf \
  --milvus-local

# Each chunk gets these metadata fields:
# - pdf_minio_key: Storage key (e.g., "pdfs/KnowledgeBase/document.pdf")
# - pdf_minio_url: Access URL (e.g., "http://localhost:9000/bucket/pdfs/...")

# Set credentials (one time)
export AWS_ACCESS_KEY_ID="your-access-key"
export AWS_SECRET_ACCESS_KEY="your-secret-key"

# Create backup and upload to S3
weave backup create MyCollection --output backup.weavebak \
  --remote-storage s3 \
  --s3-bucket my-backups \
  --s3-region us-east-1

# Restore from S3
weave backup restore backup.weavebak.gz \
  --remote-storage s3 \
  --s3-bucket my-backups

# Start MinIO locally
docker run -d -p 9000:9000 -p 9001:9001 \
  --name minio \
  -e MINIO_ROOT_USER=minioadmin \
  -e MINIO_ROOT_PASSWORD=minioadmin \
  minio/minio server /data --console-address ":9001"

# Create backup and upload to MinIO
weave backup create MyCollection --output backup.weavebak \
  --remote-storage minio \
  --s3-bucket weave-backups \
  --s3-endpoint localhost:9000 \
  --s3-access-key minioadmin \
  --s3-secret-key minioadmin \
  --s3-no-ssl

# Restore from MinIO
weave backup restore backup.weavebak.gz \
  --remote-storage minio \
  --s3-bucket weave-backups \
  --s3-endpoint localhost:9000 \
  --s3-no-ssl

# List collections (all configured VDBs)
weave cols ls

# List collections from specific database types
weave cols ls --weaviate            # Weaviate only
weave cols ls --qdrant-local        # Qdrant local only
weave cols ls --qdrant-cloud        # Qdrant cloud only
weave cols ls --milvus-local        # Milvus local only
weave cols ls --milvus-cloud        # Milvus cloud (Zilliz) only
weave cols ls --chroma-local        # Chroma local only
weave cols ls --chroma-cloud        # Chroma cloud only
weave cols ls --supabase            # Supabase only
weave cols ls --neo4j-local         # Neo4j local only
weave cols ls --neo4j-cloud         # Neo4j cloud (Aura) only
weave cols ls --mongodb             # MongoDB Atlas only
weave cols ls --pinecone            # Pinecone only
weave cols ls --opensearch-local    # OpenSearch local only
weave cols ls --opensearch-cloud    # OpenSearch cloud (AWS) only
weave cols ls --elasticsearch-local # Elasticsearch local only
weave cols ls --elasticsearch-cloud # Elasticsearch cloud (Elastic) only
weave cols ls --mock                # Mock database only
weave cols ls --all                 # All configured databases

# Create a collection
weave cols create MyCollection --text

# Add documents
weave docs create MyCollection document.txt
weave docs create MyCollection document.pdf

# Search with natural language
weave cols q MyCollection "search query"

# AI-powered Read, Evaluate, Print, Loop (REPL) mode or agent mode
weave
> show me all my collections
> create TestDocs collection
> add README.md to TestDocs

# Or doing one query at a time
weave query "show me all my collections"

# List available embeddings
weave embeddings list
weave emb ls --verbose

# Create collection with specific embedding (used as default for all documents)
weave cols create MyCollection --embedding text-embedding-3-small
weave cols create MyCollection -e text-embedding-ada-002

# Get AI-powered schema suggestions for your documents
weave schema suggest ./docs --collection MyDocs --output schema.yaml

# Get AI-powered chunking recommendations
weave chunking suggest ./docs --collection MyDocs --output chunking.yaml

# List all configured databases
weave vdb list                    # Show all databases
weave vdb ls                      # Alias for list
weave vdb list --cloud            # Show only cloud databases
weave vdb list --local            # Show only local databases

# Show detailed database information
weave vdb info weaviate-cloud     # Connection details, auth, collections
weave vdb info milvus-local       # Vector settings, endpoints

# Check database health
weave vdb health                  # Directs to 'weave health check'
weave vdb health weaviate-cloud   # Check specific database

# Aliases for discoverability
weave db list                     # Same as 'weave vdb list'
weave database info my-db         # Same as 'weave vdb info my-db'

# Quick timeout for fast connectivity test
weave health check --timeout 5s

# Longer timeout for cloud operations
weave cols query MyDocs "search" --timeout 30s

# Very long timeout for bulk imports
weave docs batch --dir ./data --collection Docs --timeout 600s

# Configure per-database in config.yaml
timeout: 30  # seconds

# Single file with parallel chunk processing
weave docs create MyCol document.txt --workers 3

# Glob pattern with parallel processing (batch mode)
weave docs create MyCol "docs/*.pdf" --workers 5

# Multiple text files with progress tracking
weave docs create MyCol "/path/to/files/*.txt" --workers 4

# Complex glob patterns
weave docs create MyCol "**/*.md" --workers 3  # Recursive

# Traditional batch command (file-level parallelism)
weave docs batch --dir ./docs --collection MyCol --parallel 3

# Sequential (workers=1, default)
weave docs create MyCol large.pdf --chunk-size 5000
# ⏱ 45 seconds for 10 chunks

# Parallel (workers=3)
weave docs create MyCol large.pdf --chunk-size 5000 --workers 3
# ⏱ 15 seconds for 10 chunks (3x faster)
🚀 Processing 10 chunks with 3 workers...
🔍 Processing 10 chunks in parallel
   [==>                 ]  10% (1/10, ETA: 23s)
   [====>               ]  20% (2/10, ETA: 13s)
   ...
   [====================] 100% (10/10)
✅ All 10 chunks processed successfully (15.2s)

# Try any command - you'll get prompted to configure interactively
weave cols ls

# Or install latest release of weave-mcp for REPL mode
weave config update --weave-mcp

# Create configuration in global directory
weave config create --env --global

# Sync local configuration to global directory
weave config sync

# View which configuration location is being used
weave config show

# Single database setup - no flags needed!
weave docs create MyCollection doc.txt       # Uses your only configured DB

# Multiple databases with VECTOR_DB_TYPE set
export VECTOR_DB_TYPE=weaviate-cloud
weave docs create MyCollection doc.txt       # Uses weaviate-cloud (default)
weave docs delete MyCollection doc123        # Uses weaviate-cloud (default)

# Override default with --vdb (short) or --vector-db-type (long)
weave docs create MyCollection doc.txt --vdb weaviate-local
weave docs create MyCollection doc.txt --vector-db-type supabase

# --weaviate tries both weaviate-cloud and weaviate-local
weave docs ls MyCollection --weaviate        # Searches both for collection
weave cols delete MyCollection --weaviate    # Searches both for collection

# Read operations work with specific or all databases
weave cols ls --weaviate                     # All Weaviate databases
weave cols ls --supabase                     # Supabase only
weave cols ls --all                          # All configured databases (default)

# Query multiple databases at once
weave cols query MyCollection "search" --weaviate --supabase

# Collections summary across all databases (default for multiple VDBs)
weave cols ls                    # Shows summary table by default
weave cols ls -S                 # Explicit summary flag (shorthand)
weave cols ls --summary          # Explicit summary flag (long form)

# Health check with progressive output (new in v0.7.2)
weave health check               # Shows summary table, results appear
weave health check -S            # Same as above (shorthand)

# Filter databases by deployment type (new in v0.7.2)
weave config list --cloud        # Show only cloud databases
weave config list --local        # Show only local databases

weave health check --cloud       # Check only cloud databases
weave health check --local       # Check only local databases
weave health check --local -S    # Local databases summary

# Force detailed view for single database
weave cols ls --weaviate         # Detailed list (default for single VDB)
weave health check weaviate      # Detailed health check

# Collections summary also supports filtering
weave cols ls --cloud            # Collections from cloud databases only
weave cols ls --local -S         # Local collections summary

weave doctor                    # Full diagnostic scan
weave doctor --verbose          # Show all checks (including passing)
weave doctor --fix              # Scan + show auto-fix suggestions
weave doctor --section config   # Only check config section
weave doctor --json             # Machine-readable JSON output

# Use RAG agent with any vector database
weave cols query MyDocs "What is machine learning?" --agent rag-agent
weave cols query MyDocs "Summarize main topics" --agent summarize-agent --db qdrant
weave cols query MyDocs "Answer this question" --agent qa-agent --db milvus

# Combine with database selection
weave cols q MyDocs "AI overview" --agent rag-agent --qdrant-local
weave cols q MyDocs "Quick summary" --agent summarize-agent --chroma-local
weave cols q MyDocs "Detailed analysis" --agent rag-agent --mongodb

# Show progress during agent execution
weave cols q MyDocs "complex query" --agent rag-agent --progress

# JSON output with progress (JSON Lines format)
weave cols q MyDocs "query" --agent rag-agent --json --progress

# Multi-collection queries (query multiple collections at once)
weave cols query WeaveDocs WeaveImages "weave cli" --agent rag-agent --top_k 3
weave cols query AuctionsDocs AuctionsImages AuctionResults "vintage cars" --agent rag-agent

# Query 2 collections (returns top 3 from EACH)
weave cols query Collection1 Collection2 "search query" --agent rag-agent --top_k 3

# Query 3+ collections (useful for multi-modal data)
weave cols query Docs Images Audio "query" --agent summarize-agent --top_k 5

# Query collections from different VDBs
weave cols query WeaveDocs:weaviate-local WeaveImages:milvus-local "weave cli" --agent rag-agent

# AuctionsMax.ai example: Query across MongoDB, Weaviate, and Milvus
weave cols query \
  AuctionListings:mongodb-cloud \
  AuctionResults:weaviate-cloud \
  AuctionImages:milvus-cloud \
  "vintage Leica cameras" \
  --agent rag-agent --top_k 3

# Mixed: explicit VDB + default from flags
weave cols query WeaveDocs WeaveImages:milvus-local "query" --weaviate-local --agent rag-agent

# Chain agents: RAG analysis → QA extraction
weave cols query MyDocs "machine learning algorithms" --agents rag-agent,qa-agent

# Triple chain: Search → Comprehensive RAG → Summarization
weave cols query LargeDocs "annual report" --agents search-agent,rag-agent,summarize-agent

# Multi-collection + multi-agent
weave cols query Docs Images "product features" --agents rag-agent,summarize-agent --top_k 10

# With progress tracking
weave cols query MyDocs "complex query" --agents rag-agent,qa-agent --progress

# Run evaluation with baseline dataset
weave eval run --agent rag-agent --dataset baseline

# Create custom test dataset from template
weave eval datasets create my-tests --template baseline

# List all available datasets
weave eval datasets list

# Show evaluation results
weave eval show eval-20260126-143022

# Compare multiple agents on same dataset
weave eval benchmark --agents rag-agent,qa-agent --dataset baseline

# Default: Local evaluators (uses OpenAI API key)
weave eval run --agent rag-agent --dataset baseline

# Opik: Cloud-based with dashboards & cost tracking
weave eval run --agent rag-agent --dataset baseline --use-opik

# List available custom evaluators
weave eval list-evaluators

# Create a new evaluator from template
weave eval create-evaluator citation_check --type regex

# Validate evaluator definition
weave eval validate-evaluator evals/evaluators/citation_check.yaml