Getting Started with Trieve: Build RAG-Powered Search in Minutes

Getting Started with Trieve: A Practical Tutorial for Building RAG-Powered Search

In this tutorial, we'll walk through deploying Trieve and using its API to build a semantic search engine with Retrieval-Augmented Generation (RAG). Trieve provides a powerful all-in-one API for search, recommendations, and RAG — making it an ideal choice for developers who want production-grade retrieval without managing multiple infrastructure components.

Prerequisites

• Docker and Docker Compose installed
• At least 4GB RAM recommended
• API key from dashboard.trieve.ai (or self-hosted instance)

Step 1: Quick Start with Docker Compose

Create a docker-compose.yml file:

version: '3.8'
services:
  trieve:
    image: trieve/server:latest
    restart: unless-stopped
    ports:
      - "8080:8080"
    volumes:
      - ./data/trieve:/data
    environment:
      - OPENAI_API_KEY=${OPENAI_API_KEY}

Run: docker compose up -d

Step 2: Upload Documents

Use the Trieve API to upload chunks for indexing. Here's an example using Python:

import requests

API_URL = "http://localhost:8080"
HEADERS = {"Authorization": "Bearer your-api-key"}

# Create a dataset
dataset = requests.post(
    f"{API_URL}/api/dataset",
    json={"name": "my-docs"},
    headers=HEADERS
).json()

# Upload a chunk
chunk = requests.post(
    f"{API_URL}/api/chunk",
    json={
        "chunk_html": "<p>Trieve is an all-in-one platform for search, recommendations, RAG, and analytics.</p>",
        "link": "https://docs.trieve.ai",
        "tag_set": ["search", "rag", "ai"]
    },
    headers=HEADERS
).json()

Step 3: Search with Hybrid Retrieval

Search your documents using hybrid search — combining dense vector search with sparse neural full-text search:

search = requests.post(
    f"{API_URL}/api/chunk/search",
    json={
        "query": "What can Trieve do?",
        "search_type": "hybrid",
        "limit": 5
    },
    headers=HEADERS
).json()

for result in search["chunks"]:
    print(f"Score: {result['score']:.3f}")
    print(f"Content: {result['chunk_html'][:100]}...")
    print("---")

Step 4: RAG with LLM Integration

Use Trieve's managed RAG routes to get LLM-generated answers based on your retrieved context:

rag_response = requests.post(
    f"{API_URL}/api/chunk/generate",
    json={
        "chunks": [chunk["id"] for chunk in search["chunks"]],
        "prompt": "Summarize what Trieve offers"
    },
    headers=HEADERS
).json()

print(rag_response["answer"])

Conclusion

Trieve makes it remarkably simple to build production-grade search and RAG systems. With its hybrid search approach, sub-sentence highlighting, and managed RAG routes, you can go from zero to a fully functional semantic search engine in under an hour. The self-hosting option ensures your data stays in your control while leveraging state-of-the-art retrieval techniques.