How codie-ds/gradeflow-ai works
GradeFlow-AI lets educators define assignments, generate or edit rubric criteria, upload handwritten PDF submissions, and receive OCR-backed rubric scores with evidence and confidence indicators for human review. The main runtime path runs from the assignment/upload interface through FastAPI routes and a background grading pipeline to persisted results, which the interface retrieves for inspection. The graph preserves the documented multimodal OCR and audit capabilities; their implementation details are not fully shown in the sampled source, so those links are documented rather than asserted from code.
Main entry points: api.js HTTP API client · App.jsx Page routing · assignments.py Assignment endpoints · pipeline.py Submission pipeline
Components 19
Educator interface
- Page routing
frontend/src/App.jsx - Assignment dashboard
frontend/src/pages/Dashboard.jsx - Assignment setup
frontend/src/pages/CreateAssignment.jsx - Assignment workspace
frontend/src/pages/AssignmentPage.jsx - Results review
frontend/src/pages/ResultPage.jsx - HTTP API client
frontend/src/api.js - Status and score UI
frontend/src/components/ui.jsx
Application API
- Assignment endpoints
backend/app/routes/assignments.py - Submission endpoints
backend/app/routes/submissions.py - Request and response schemas
backend/app/schemas.py
Grading services
- Submission pipeline
backend/app/services/pipeline.py - PDF OCR extraction
backend/app/services/ocr.py - Rubric-based grading
backend/app/services/grader.py - Rubric generation
backend/app/services/rubric.py - Gemma JSON client
backend/app/services/gemma_client.py
Persistence
- Assignment and result data
backend/app/db.py
Other components
- Educator
- Supabase storage and database
- Qwen vision model
Connections 27
- Educator → Page routing: uses
README.md - Page routing → Assignment dashboard: routes to
frontend/src/App.jsx - Page routing → Assignment setup: routes to
frontend/src/App.jsx - Page routing → Assignment workspace: routes to
frontend/src/App.jsx - Page routing → Results review: routes to
frontend/src/App.jsx - Assignment setup → HTTP API client: calls
frontend/src/pages/CreateAssignment.jsx - Assignment dashboard → HTTP API client: calls
frontend/src/pages/Dashboard.jsx - Assignment workspace → HTTP API client: calls
frontend/src/pages/AssignmentPage.jsx - Results review → HTTP API client: calls
frontend/src/pages/ResultPage.jsx - Assignment workspace → Status and score UI: renders
frontend/src/pages/AssignmentPage.jsx - Results review → Status and score UI: renders
frontend/src/pages/ResultPage.jsx - HTTP API client → Assignment endpoints: HTTP requests
frontend/src/api.js - HTTP API client → Submission endpoints: HTTP requests
frontend/src/api.js - Assignment endpoints → Request and response schemas: validates with
backend/app/routes/assignments.py - Submission endpoints → Request and response schemas: serializes with
backend/app/routes/submissions.py - Assignment endpoints → Assignment and result data: reads and writes
backend/app/routes/assignments.py - Assignment endpoints → Rubric generation: generates criteria
backend/app/routes/assignments.py - Assignment endpoints → Submission pipeline: schedules
backend/app/routes/assignments.py - Submission endpoints → Assignment and result data: reads
backend/app/routes/submissions.py - Submission pipeline → Assignment and result data: loads and saves
backend/app/services/pipeline.py - Submission pipeline → PDF OCR extraction: extracts pages
backend/app/services/pipeline.py - Submission pipeline → Rubric-based grading: grades extraction
backend/app/services/pipeline.py - Rubric-based grading → Gemma JSON client: requests evaluation
backend/app/services/grader.py - Rubric generation → Gemma JSON client: requests criteria
backend/app/services/rubric.py - Assignment and result data → Supabase storage and database: queries and stores files
backend/app/db.py - PDF OCR extraction → Qwen vision model: extracts multimodal content
README.md - Educator → Results review: audits scores
README.md