Home Getting Started Architecture Integration Configuration Examples Contributing

DeepNext Architecture

DeepNext operates through a pipeline that handles the entire software engineering process:

Pipeline Components

1. Gather Project Knowledge

• Inputs: Repository root directory
• Process:
  • Creates comprehensive project description by analyzing READMEs, configuration files, and code structure
  • Generates project map showing directory organization and component relationships
• Outputs: Consolidated project knowledge description

2. Create Action Plan

• Inputs: Project knowledge and issue description
• Process:
  • Selects relevant files through multiple parallel analysis cycles
  • Uses AI to deduce which files need modifications
  • Combines results from multiple analysis passes for higher accuracy
  • Creates detailed, structured solution design with specific file changes
• Outputs: Action plan with ordered steps for implementation

3. Implement Changes

• Inputs: Action plan and access to the files for modification
• Process:
  • Develops edits for each file with proper context and formatting
  • Applies changes using exact matching first, then fuzzy matching for resilience
  • Generates comprehensive git diff for review
• Outputs: Complete implementation as git diff or pull/merge request

Project Structure

The project is structured as a monorepo with specialized components:

• Apps: Main DeepNext application for Github/GitLab integration (app/)
• Libs: Core processing logic, connectors for external services, and common utilities
  • core/: Main logic for analyzing repositories and generating solutions
  • connectors/: Integration with external services
  • common/: Shared utilities and configurations

Back to Home