User Guide#

Welcome to the Elsai Core User Guide. This section provides comprehensive documentation for all Elsai components and modules.

Components Overview#

Large Language Models & MCP

  • elsai-model: Connects to and manages LLMs from OpenAI, Azure OpenAI, and Amazon Bedrock.

  • elsai-model-hub: OpenAI-compatible API for hosted Models Hub endpoints (for example Gemma-4 E4B and Phi-4-mini).

  • elsai-mcp: Automatically converts any backend Swagger (OpenAPI) API into a Model Context Protocol (MCP) server, allowing LLMs to interact with APIs as tools.

Document Processing

  • elsai-text-extractors: Extracts raw text from structured documents like PDF, DOCX, CSV, and Excel.

  • elsai-ocr-extractors: Extracts text from scanned PDFs and images using OCR services like Azure, Textract, and Vision AI.

  • elsai-parsers: Parses and queries Excel files using LLMs with natural language queries, structured responses, and support for both small and large files.

  • elsai-utilities: offers a collection of helper classes designed to power intelligent document and conversation processing workflows. It includes tools for chunking and converting documents for use in retrieval-augmented generation (RAG) and vector database pipelines, as well as advanced conversational intelligence features such as follow-up question generation, action item detection, and topic/intent classification

Data and Storage

  • elsai-db: Executes natural language queries over SQL databases using LLM-powered connectors.

  • elsai-embeddings: Generates vector embeddings from text using Azure OpenAI embedding models.

  • elsai-vectordb: Stores and retrieves document embeddings using ChromaDB or Pinecone for semantic search.

  • elsai-cloud-connectors: Enables file operations across cloud platforms like AWS S3, Azure Blob, SharePoint, and OneDrive.

Retrieval and Search

  • elsai-retrievers: Combines semantic and keyword-based retrieval to enhance document search accuracy.

  • elsai-nli: Lets users query structured data like CSV files using natural language and LLMs.

  • elsai-graph-query-rag: Enables natural language question answering by combining knowledge graph traversal with vector similarity search.

Audio and Prompts

  • elsai-stt: Provides comprehensive audio processing capabilities including Speech-to-Text (STT), Text-to-Speech (TTS), and Speech-to-Speech conversion using Azure OpenAI Whisper and TTS models.

  • elsai-prompts: Retrieves and manages prompt templates from Pezzo for version-controlled prompt engineering.

Chat History

  • elsai-chat-history: simplifies managing LLM-compatible chat histories with full async support, multi-backend storage, and pluggable memory strategies. Ideal for building scalable and memory-aware AI chat applications.

Knowledge Graphs

  • elsai-graph-constructor: Automatically builds knowledge graphs from unstructured text by identifying entities and discovering their relationships.

  • elsai-graph-generator: Manages Neo4j connections, stores graph data, and handles vector embeddings for similarity searches.

  • elsai-graph-query-rag: Enables natural language question answering by combining knowledge graph traversal with vector similarity search.

Getting Started#

To get started with Elsai Core, we recommend beginning with elsai-model to understand how to connect to and manage language models, which forms the foundation for most other components.