Quick Start Sandbox

The sandbox lets you try Weaviate Studio against a local Weaviate instance before connecting a production cluster. For the query-profiling release, use the dedicated profiling configuration first: it creates searchable collections with a local text vectorizer, so the ordinary Text (Semantic) workflow remains available without a cloud embedding API key.

Choose a Sandbox

Sandbox	Best for	Start command
Query Profiling sandbox	Testing Profile, View profile, Text (Semantic) search, and per-shard timing	docker compose -f docker-compose.profiling.yml up -d
Extended sandbox	Exploring the older RAG, nested-object, cross-reference, and bulk demo data	docker compose up -d

Query Profiling Sandbox

Prerequisites

Install Docker with Docker Compose and Node.js. From a fresh repository clone, run npm install once so the seed script can use the project dependencies.

1. Start the Services

Run these commands from the repository’s sandbox directory:

docker compose -f docker-compose.profiling.yml up -d
docker compose -f docker-compose.profiling.yml ps

The first start downloads the local embedding model. Wait until the text2vec-transformers container is healthy before seeding. The seed script also waits for Weaviate readiness before creating data.

2. Seed Searchable Demo Collections

node seed-query-profiling.cjs

The script creates or refreshes the following focused demo collections:

Collection	Recommended use
ProfileTest	A small, focused smoke-test collection for the profiling workflow.
TravelGuide	Ordinary Text (Semantic) search with travel-oriented content.
ProductCatalog	Ordinary Text (Semantic) search with product-oriented content.

Re-running the script refreshes only these three collections.

3. Connect Weaviate Studio

Setting	Value
Endpoint	http://localhost:8080
API key	test-key-123

Open the Weaviate Studio side-bar view in VS Code, add the connection, then open Data Explorer for one of the seeded collections.

4. Profile a Search

1. Select Vector Search.
2. Use Text (Semantic) for ProfileTest, TravelGuide, or ProductCatalog, then enter a natural-language query such as independence day of india.
3. Check Profile beside Run Vector Search.
4. Run the search. Once results arrive, select View profile to reveal the per-shard Timing breakdown.

The setting is remembered across collection changes and Data Explorer panel reopens. Select Hide profile when you want to return to the compact results view.

Text (Semantic) search requirement: A collection needs a text vectorizer. The profiling sandbox supplies a local text2vec-transformers service, so its three demo collections support this mode without a cloud API key.

Reset the Profiling Sandbox

docker compose -f docker-compose.profiling.yml down -v --remove-orphans
docker compose -f docker-compose.profiling.yml up -d
node seed-query-profiling.cjs

Extended Sandbox

The standard docker compose.yml workflow remains available for the larger legacy data set. It covers nested properties, cross-references, multi-collection RAG, generative queries, and bulk sample data.

What’s Included

• Nested object properties — Author.address, Book.metadata
• Cross-references — Book → Author, GitHubRepo → GitHubUser
• Multi-collection RAG — Query across Books and Podcasts
• Generative queries — AI-powered answers grounded in retrieved data

Start and Populate

docker compose up -d
pip install weaviate-client requests
python3 populate.py

For a smaller or specialized import, the script supports --rag-only, --legacy-only, --skip-github, and --verify-only.

Example Generative Queries

• "Find highly rated fantasy books"
• "What topics do these podcasts cover?"
• "Compare rationality-related podcast topics with highly rated nonfiction books"
• "What stories involve mystery or investigation?"

Troubleshooting

Situation	What to do
The profiling seed fails immediately	Run docker compose -f docker-compose.profiling.yml ps, wait for the transformer service to become healthy, then rerun node seed-query-profiling.cjs.
The endpoint is not ready	Check docker compose -f docker-compose.profiling.yml logs weaviate and wait for Weaviate readiness.
Text (Semantic) search cannot vectorize input	Confirm that the profiling compose file, rather than the manual-vector-only setup, is running and that the selected collection is one of the seeded profiling collections.
You need a clean start	Use the reset commands above; they remove the profiling sandbox’s local volumes before reseeding.