Inspiration

Developers often inherit undocumented or poorly structured codebases, making onboarding and maintenance a challenge. AIDocs was created to automate in-depth project understanding by transforming GitHub repositories into complete, readable documentation with architecture diagrams, workflow mappings, and service-level insights.

What it does

AIDocs converts any public GitHub repository into a full-featured documentation suite. The platform enables:

Selection of specific files or entire repositories.
AI-powered analysis (via AgentForce) to extract:
- Service logic
- Database schemas
- Controller layers
- API route structures
Automatic generation of:
- Structured markdown documentation
- Logical summaries and module responsibilities
- Code flow explanations
- Workflow diagrams
- Architecture diagrams
- Database entity-relationship (ER) diagrams
- Service/API interdependency maps

All outputs are pushed to a new GitHub repository with a generated permalink for easy access and sharing.

How we built it

Frontend:
- React.js and TailwindCSS for an interactive UI to browse repositories and select files.
AgentForce-Powered Backend:
- Uses LLM-based agents to:
- Parse and understand code context
- Extract responsibilities, logic flow, and internal dependencies
- Chain prompts to map code structure and generate summaries
Diagram Generator:
- Generates:
- Mermaid/Graphviz-based flowcharts
- Sequence diagrams for API routes
- ER diagrams from database schema files
- Component-based architecture maps
Documentation Engine:
- Builds markdown files with internal navigation, linked modules, and embedded diagrams.
- Supports selective or full-repository documentation.
GitHub Integration:
- Uses GitHub API to push documentation into a new repository.
- Provides direct access links.

Challenges we ran into

Parsing heterogeneous code structures into a unified format for AI processing.
Generating consistent diagrams across non-standardized projects.
Mapping deeply nested services and dependencies into clear documentation.
Keeping latency low while performing in-depth code understanding and diagram generation.

Accomplishments that we're proud of

Built an end-to-end system that transforms raw code into readable, visual documentation.
Successfully integrated AgentForce to perform semantically rich code analysis.
Enabled developers to instantly visualize architecture and understand logic flows.
Completed the doc + diagram generation cycle in under 30 seconds per repository.

What we learned

Diagrams aren’t just visuals—they’re essential to understanding architecture.
AI documentation tools must balance technical depth with readability.
Prompt engineering and fallback logic are critical when using LLM agents like AgentForce.
Developers prioritize structured insight over plain summaries.