Cortex: Local-First Knowledge Graph for Developers
Cortex is designed specifically for developers who juggle multiple projects, enabling them to efficiently track, retrieve, and manage knowledge across diverse files and repositories. This local-first knowledge graph utilizes advanced language models (LLMs) to automatically extract key entities and relationships from project files, allowing natural language queries that bring context back to the forefront of development work.

Key Features

• Continuous Monitoring: Cortex proactively watches an array of project files (including Markdown, TypeScript, JavaScript, JSON, and YAML) for changes and updates.
• Entity Extraction: Automatically identifies and extracts crucial decision points, patterns, component relations, dependencies, and action items from the project files.
• Relationship Inference: Discovers and infers relationships between extracted entities across various projects, helping to identify conflicting decisions.
• Natural Language Queries: Users can query for information in natural language, receiving responses along with source citations.
• Privacy Focused: Ensures that restricted projects remain private by processing all data locally, with no sensitive information sent to the cloud.
• Interactive Web Dashboard: Provides a user-friendly interface for visualizing the knowledge graph, navigating through real-time updates, and exploring queries effortlessly.
• Multi-Language Model Support: Works with various LLM providers, adapting to different needs and protecting privacy accordingly.

How It Works

Cortex operates through a streamlined pipeline whenever a file change is detected:

1. Parse: Content is processed using a language-aware parser.
2. Extract: LLM identifies entities within the content.
3. Relate: Establishes connections between newly identified and existing entities.
4. Detect: Identifies contradictions and duplicates automatically.
5. Store: Saves all gathered data locally using SQLite and LanceDB.
6. Query: Facilitates natural language queries to synthesize answers from the knowledge graph.

Supported LLM Providers

• Anthropic Claude
• Google Gemini
• OpenAI-compatible providers
• Ollama for local model support

Benefits

Cortex centralizes knowledge from various projects, reducing redundancy and enhancing decision-making by providing a comprehensive overview of prior choices and patterns. Developers can quickly access pertinent information that would otherwise be spread across numerous files and repositories, thus minimizing context loss and improving productivity.

Getting Started

To begin using Cortex, it involves setting up the initial configuration and registering projects to monitor. Follow the step-by-step commands provided in the documentation for a seamless setup experience.

For more information, including detailed configuration options and command references, visit the Cortex documentation. Implement Cortex in your development workflow to enhance your productivity and maintain project context effortlessly.