How mohansaikokku/veritext-ai works
VeriText AI accepts text or PDF, DOCX, and TXT uploads through a web interface, sends them to a FastAPI backend, and returns AI-versus-human predictions with confidence and chunk details. File requests are temporarily saved, extracted, and removed; both input paths are cleaned, split into chunks, classified by a Hugging Face transformer, and aggregated. The graph separates those domain stages and the UI result presentation. Connections are limited to relationships shown in excerpts or documented in the README; unsampled HTML details are not inferred.
Main entry points: main.py FastAPI routes · script.js Input and results · analyzer.py Clean and chunk · index.html Analysis interface
Components 11
Web interface
- Analysis interface
index.html - Input and results
script.js
API and uploads
- FastAPI routes
main.py - Temporary uploads
main.py
Document extraction
- Format dispatch
text_extractor.py - PDF DOCX TXT readers
text_extractor.py
Text detection
- Clean and chunk
analyzer.py - Text classifier
model.py - Prediction summary
analyzer.py
Other components
- User
- Hugging Face model
Connections 12
- User → Analysis interface: provides input
README.md - Analysis interface → Input and results: uses interface
- Input and results → FastAPI routes: posts text or file
script.js - FastAPI routes → Input and results: returns results
script.js - FastAPI routes → Temporary uploads: saves and removes
main.py - FastAPI routes → Format dispatch: extracts uploaded text
main.py - Format dispatch → PDF DOCX TXT readers: selects reader
text_extractor.py - FastAPI routes → Clean and chunk: analyzes text
main.py - Clean and chunk → Text classifier: classifies chunks
analyzer.py - Text classifier → Hugging Face model: runs inference
- Clean and chunk → Prediction summary: summarizes predictions
analyzer.py - Prediction summary → Input and results: displays analysis
script.js