Turn PDFs into Smart Search Engines with WeKnora’s Open‑Source LLM Framework

Overview

WeKnora is an open‑source framework from Tencent that integrates large language models (LLM) with multimodal parsing, semantic indexing and retrieval‑augmented generation (RAG). It enables natural‑language queries over complex documents such as PDFs, Word files and images, turning manual document search into conversational interaction.

Key Features

Agent mode : Implements ReACT‑style agents that can invoke built‑in knowledge bases, MCP tools and web search, iterating to produce comprehensive reports.

Precise understanding : Extracts structured content from PDFs, Word and images and builds a unified semantic view.

Intelligent reasoning : Uses LLMs for context comprehension and multi‑turn dialogue.

Multiple knowledge bases : Supports FAQ‑type and document‑type stores with folder, URL import, tagging and online entry.

Flexible pipeline : Decoupled stages (parsing → embedding → retrieval → generation) allow easy integration of custom components.

Hybrid retrieval : Combines keyword, vector and knowledge‑graph search, enabling cross‑knowledge‑base queries.

MCP tool integration : Extensible agents via MCP, including uvx and npx launch tools.

Conversation control : Configurable agent model, normal model, retrieval thresholds and prompts for fine‑grained dialogue management.

User‑friendly interface : Web UI and standard RESTful API require minimal technical expertise.

Secure and controllable : Deployable on‑premise or in private cloud, keeping data fully under user control.

Architecture

Document processing layer : Parses PDFs, Word, images and performs preprocessing.

Knowledge modeling layer : Performs vectorization, chunking and knowledge‑graph construction for deep content representation.

Retrieval engine layer : Mixes keyword, vector and graph‑based strategies for efficient, accurate recall.

Reasoning generation layer : Leverages LLMs for deep understanding and answer generation, with optional agent reasoning.

Interaction layer : Provides a web UI and RESTful API.

The design allows interchangeable LLMs (e.g., Ollama, Qwen, DeepSeek) and vector databases, while preserving full control for private deployments.

Quick Start

Prerequisites

Docker

Docker Compose

Git

Installation steps

Clone the repository and enter the project directory

git clone https://github.com/Tencent/WeKnora.git
cd WeKnora

Copy the example environment file and edit .env with your settings cp .env.example .env Start all services (includes Ollama) ./scripts/start_all.sh or make start-all Stop services when needed ./scripts/start_all.sh --stop or

make stop-all

Service URLs after startup

Web UI: http://localhost Backend API: http://localhost:8080 Jaeger tracing UI:

http://localhost:16686

Open‑Source Repository

https://github.com/Tencent/WeKnora