How to Build Your First RAG Pipeline with RocketRide (Tutorial)

Hands-on: build your first RocketRide pipeline

In this tutorial we'll create a simple RAG pipeline in RocketRide — ingest a document, embed it into a vector store, and answer questions with an LLM — then run it from the VS Code extension and via the Python SDK.

Step 1: Install and start the engine

The fastest path is Docker: pull the engine image and start the stack with PostgreSQL + pgvector, Milvus, and Chroma. RocketRide's compose file wires everything together, and the engine exposes a health check on /ping. Minimum 4 GB RAM; 8 GB recommended for production workloads.

docker pull ghcr.io/rocketride-org/rocketride-engine:latest
docker compose up -d   # engine + postgres + milvus + chroma

Step 2: Open the IDE extension

Install the RocketRide VS Code extension and connect it to the running engine. You get a visual canvas where nodes are dragged and wired together. For our RAG pipeline we need: a document loader, a text splitter, an embedding node, a vector store node (pgvector), a retriever, and an LLM node.

Step 3: Wire the nodes and trace

Connect loader → splitter → embedding → pgvector to index your document. Then build the query branch: question → retriever → LLM. Hit Run and watch the tracing panel — every node's inputs, outputs, and latency are recorded in real time, which makes debugging failed steps trivial.

Step 4: Drive it from code

RocketRide ships TypeScript and Python SDKs. The same pipeline you drew in the IDE can be triggered from your application:

from rocketride import Client

client = Client("http://localhost:5565")
result = client.run("rag-pipeline", {"question": "What is RocketRide?"})
print(result.answer)

Review verdict

Bottom line: RocketRide is one of the most complete open-source LLM workflow engines available — fast, IDE-native, and fully self-hostable. If you want production-grade pipelines without cloud lock-in, it's an excellent pick.