🚀 Getting Started with LLMosaic¶

LLMosaic is a developer-friendly AI platform designed to accelerate the creation of Retrieval-Augmented Generation (RAG) applications. It provides a unified set of REST APIs for generating embeddings, performing LLM completions, storing and searching vector data, and working with structured and semi-structured enterprise data — all backed by a PostgreSQL engine extended with vector support.

This "Getting Started" guide introduces the platform’s core functionality and prepares you to complete a series of hands-on tutorials and walkthroughs.

✅ What You’ll Learn¶

1. 5‑Minute RAG Quickstart¶

2. Generate Embeddings via OpenAI-Compatible API¶

LLMosaic provides a /v1/embeddings endpoint that accepts raw text input and returns high-dimensional vector embeddings. The endpoint is fully compatible with OpenAI client libraries, making it easy to integrate into existing workflows.
You’ll learn how to: - Submit text or batched input for embedding - Interpret the embedding vector output - Use the output for semantic search and retrieval

3. Perform LLM Completions via OpenAI-Compatible API¶

LLM completions can be performed using the /v1/chat/completions endpoint. The API supports both standard chat-based interactions and system-instructed prompts.
You’ll learn how to: - Create and format completion requests - Use system and user roles in prompts - Parse responses for integration into your app logic

4. Store and Search Embeddings Using PostgreSQL + pgvector/vectorchord¶

LLMosaic’s vector database layer is built on PostgreSQL with the pgvector and vectorchord extensions.
You’ll learn how to: - Store embedding vectors in tenant-isolated tables - Create and manage vector indexes - Perform similarity search using REST-based queries

5. Query and Join Structured Data via Relational and JSON Schemas¶

The platform supports relational tables and JSON documents via a PostgREST-inspired REST API. Multi-tenant schemas ensure isolation and scalability.
You’ll learn how to: - Create and manage PostgreSQL tables and JSONB columns - Insert and retrieve structured or semi-structured data - Perform filtered queries and table joins via HTTP

6. Load Enterprise Data Using the REST API¶

Enterprise data from external systems can be loaded into your tenant’s PostgreSQL schema using simple REST calls.
You’ll learn how to: - Create schemas tied to tenant identity - Upload structured (relational) and JSON-based data - Organize application data using named schemas

🧑‍💻 Who Is This For?¶

• AI developers integrating LLM and vector search into business applications
• Backend engineers building RAG pipelines using PostgreSQL
• Teams migrating from OpenAI to a flexible, self-hosted alternative
• Developers working with structured and semi-structured enterprise data

ℹ️ Note: This guide focuses on currently supported features. ETL pipelines, fine-tuning APIs, and optimized data pathways will be covered in future releases of the platform.