WeKnora

WeKnora is a multi-tenant retrieval-augmented generation (RAG) knowledge platform and autonomous AI agent framework. It transforms raw documents into queryable knowledge bases and integrates large language models with vector databases to provide grounded AI responses. The system also functions as a Model Context Protocol (MCP) tool server, exposing knowledge search and agentic capabilities to external AI clients.

The platform distinguishes itself through an autonomous agent framework that utilizes iterative reasoning, tool calling, and web search to solve multi-step tasks. It implements a standardized tool surface via the Model Context Protocol, allowing for the extension of agent capabilities through custom skill definitions and external service integration.

The system covers comprehensive data management areas, including recursive document chunking, hybrid search retrieval with cross-encoder reranking, and complex document parsing via OCR. It provides enterprise-grade infrastructure with multi-tenant data isolation, role-based access control, and OIDC authentication. Additional capabilities include the generation of structured wikis and knowledge graphs from ingested content, as well as integration with third-party messaging platforms.

The project can be deployed via Kubernetes or as a standalone lite distribution.

Features

Autonomous AI Agent Frameworks - Provides a framework for building self-directed agents that use iterative reasoning, tool calling, and web search.

Knowledge Base Management - Transforms raw documents and web content into searchable chunks, structured wikis, and knowledge graphs.

Conversational Interfaces - Enables users to query specific knowledge bases via a chat interface to receive grounded AI responses.

RAG Enabled Clients - Provides a RAG-enabled chat interface for querying specific knowledge bases to receive grounded AI responses.

Multi-Tenant AI Hosting - Hosts a multi-tenant AI platform on Kubernetes with integrated role-based access control and messaging channels.

Chunking Strategies - Implements recursive document chunking to optimize the accuracy of retrieval-augmented generation.

Hybrid Search Retrievers - Implements hybrid search retrievers that combine keyword and vector search to improve document recall and accuracy.

Agentic Reasoning Loops - Implements iterative reasoning loops where AI agents call tools and refine answers to solve multi-step tasks.

Agent Capability Extensions - Expands the capabilities of autonomous agents by integrating external tools through a standardized protocol.

Model Context Protocol - Exposes a standardized tool interface via the Model Context Protocol for compatible AI clients.

MCP Server Management - Provides interfaces for managing protocol tool services that extend the capabilities of autonomous agents.

Reasoning Process Monitors - Provides visualization and auditing of step-by-step reasoning chains and tool execution for token observability.

Retrieval-Augmented Generation - Transforms raw documents into queryable knowledge bases to provide grounded AI responses using retrieval-augmented generation.

MCP Protocol Integrations - Integrates with compatible editors via the Model Context Protocol to enable tool-based knowledge interaction.

Knowledge Ingestion Pipelines - Implements pipelines to upload and manage raw content and metadata used to power knowledge bases.

Document Chunking Strategies - Implements adaptive and recursive document chunking strategies to optimize retrieval accuracy.

MCP Servers - Implements an MCP server that exposes knowledge search and agent invocation to external AI clients.

Long-Form Answer Generators - Generates high-quality, grounded responses to user questions using retrieval-augmented generation.

Retrieval Augmented Generation Platforms - Provides a full platform for managing document collections and knowledge bases to ground generative language models.

Document Parsing and Extraction - Implements advanced extraction of text and structure from various document formats using OCR and multimodal processing for LLM ingestion.

Document Text Extractors - Extracts plain text from PDFs, Word files, and images using built-in tools or vision models for downstream processing.

Vector Storage - Integrates with external vector databases to store and retrieve high-dimensional embeddings for semantic search.

Multi-Format Document Ingestion - Ingests and normalizes various file formats including PDF, Word, Excel, and images for AI retrieval.

Semantic Knowledge Base Search - Uses vector embeddings to retrieve relevant information from indexed knowledge bases.

Vector Database Integrations - Connects large language models to various vector stores like pgvector, Milvus, and Qdrant for semantic search.

Multi-Tenant Isolation Layers - Enforces strict data separation between independent workspaces to ensure multi-tenant isolation.

Tenant-Specific Knowledge Workspaces - Provides isolated knowledge workspaces for different organizations with individual document repositories and RBAC.

Tenant ID Assignments - Assigns requests to unique tenant identifiers to maintain rigorous data isolation.

Tenant Identifiers - Ensures multi-tenant data isolation by segregating resources using unique tenant identifiers and role-based access matrices.

Permission-Based Access Control - Controls resource ownership and access through a structured four-tier role permission matrix.

Role-Based Access Control - Enforces a multi-tier role-based access control matrix to restrict system and resource access.

Agent Skill Integrations - Integrates pre-installed external capabilities to extend the functionality of autonomous agents.

AI Agent Builders - Provides tools for constructing and configuring custom AI agents with specific models, knowledge scopes, and skills.

Agent Skill Definitions - Supports defining custom AI agent skill sets using structured markdown and scripts.

AI Agent Servers - Functions as an AI agent server that allows users to invoke custom server-side agents for task execution.

Model Context Protocol Integrations - Implements the Model Context Protocol to expose internal knowledge and agent tools to external AI clients.

Streaming Chat Responses - Delivers chat and session answers as a line-delimited JSON event stream for incremental token processing.

AI Model Configurations - Allows configuration of model parameters and sources to control the behavior and creativity of generated responses.

Model Registrations - Provides a mechanism for registering and managing the specific models used for processing and generating responses.

Conversation History Managers - Maintains and organizes the sequence of messages within a conversation to provide stateful AI interactions.

Cross-Encoder Rerankers - Uses a cross-encoder architecture to score retrieved document chunks and optimize the final ranking of results.

Web Search Integrations - Integrates real-time web search to supplement internal knowledge bases with current external data.

Cross-Tenant Collaboration - Enables cross-tenant collaboration by granting members of different tenants access to shared organizational spaces.

AI Agent Session Managers - Manages the lifecycle of AI agent sessions, including listing history and recovering interrupted event streams.

Team Member Management - Manages organizational team structures to facilitate sharing of knowledge bases and AI agents.

AI-Powered Wiki Generators - Uses AI to distill raw documents into an interlinked markdown knowledge base and structured graph.

External System Synchronization - Automatically imports and synchronizes content from third-party platforms using external system connectors.

Web Content Ingestion Tools - Provides utilities to scrape and ingest web pages, text, and images from a browser into a knowledge base.

Standalone Distribution Bundlers - Offers a standalone lite distribution that bundles the frontend and removes external dependencies.

Dependency Health Verification - Implements active health verification for the dependency chain of knowledge bases and agents.

Kubernetes Deployments - Provides the necessary configurations and tooling to deploy the platform and its dependencies onto Kubernetes clusters.

Messaging Platform Integrations - Delivers RAG question-and-answer capabilities directly to users through messaging platforms like Slack and Telegram.

Knowledge Graph Visualizations - Transforms raw documents into an interactive graph to visualize conceptual relationships and narratives.

Audit Logging - Records role assignments and access attempts in searchable security logs for audit and compliance.

Credential Encryption - Protects sensitive API keys using AES-256-GCM encryption with support for key rotation.

Creator Restrictions - Restricts the ability to modify resources to only the original creator or an administrator.

API Request Authentication - Validates API request identity using security keys provided in HTTP headers.

OIDC Authentication Plugins - Secures user and agent access by integrating OpenID Connect identity providers for authentication.

Private Knowledge Base Imports - Extracts and stores information from web URLs to populate searchable private knowledge bases.

User Account Management - Handles user registration, login, and identity flows using OpenID Connect for account management.

Environment Health Diagnostics - Includes diagnostics to validate server reachability, authentication status, and credential storage health.

TencentWeKnora

Features

Star history