docs/issues.md

Issues & Task Tracking — Team Centurions

Below are the tracked development issues for PageParse, mapping assignees, estimated hours, due dates, and statuses.

---

Issue ID	Title / Task Description	Assignee	Estimate	Due Date	Status	Phase
PP-101	Core Ingestion & Preprocessing (OpenCV image load, crop, deskew, median filter)	Varun	8 hrs	2026-06-25	Completed	Phase 1/2
PP-102	ONNX Runtime Handwriting OCR Integration (TrOCR line model loading & decoder)	Varun	10 hrs	2026-06-26	Completed	Phase 2
PP-103	Tesseract Printed OCR Fallback Pipeline	Varun	4 hrs	2026-06-26	Completed	Phase 2
PP-104	SLM Transformation Layer (llama.cpp integration & GBNF grammar parser)	Varun	12 hrs	2026-06-27	Completed	Phase 2
PP-105	SQLite Persistence & Search Store (Sources + Records tables & sqlite-vec search)	Varun	8 hrs	2026-06-28	Completed	Phase 2
PP-106	Multi-lingual Translation Pipeline (Predefined Indian languages local mappings)	Varun	6 hrs	2026-06-28	Completed	Phase 2
PP-107	FastAPI Backend REST Endpoints & Web templates	Varun	8 hrs	2026-06-28	Completed	Phase 2
PP-108	TypeScript React Frontend PWA Dashboard with offline localStorage simulation	Varun	14 hrs	2026-06-29	Completed	Phase 2
PP-109	Live Telemetry & Resource Monitoring (psutil telemetry Panel)	Varun	4 hrs	2026-06-29	Completed	Phase 2
PP-110	CI/CD Quality Gates & Local GitLab Runner Audit configuration	Varun	6 hrs	2026-06-29	In Progress	Phase 3

---

Issues Detail

PP-101: Core Ingestion & Preprocessing

• Description: Implement input loading (JPG, PNG, PDF). Write OpenCV filters: adaptive binarization, rotation deskewing, and median denoise to feed clean line patches to OCR.
• Estimate: 8 hours

PP-102: ONNX Runtime Handwriting OCR

• Description: Implement TrOCR model loader using ONNX Runtime. Write tokenizer processor and greedy/beam search decoder for line recognition on CPU.
• Estimate: 10 hours

PP-103: Tesseract OCR Fallback

• Description: Write subprocess wrapper around Tesseract CLI as a fallback if TrOCR fails or processes printed documents.
• Estimate: 4 hours

PP-104: SLM Transformation Layer

• Description: Connect extracted text to llama.cpp/Ollama. Create grammar rule file (task.gbnf) to enforce output structure conforming to Pydantic schema.
• Estimate: 12 hours

PP-105: SQLite Persistence & Search

• Description: Design database. Build repository classes using sqlite3. Set up vector similarity search using sqlite-vec or numpy embeddings.
• Estimate: 8 hours

PP-106: Multi-lingual Translation

• Description: Build pre-defined mappings for Hindi, Telugu, and Tamil. Add free Google Translate API fallback for online usage.
• Estimate: 6 hours

PP-107: FastAPI Backend

• Description: Write API endpoints for upload, page list, record list, telemetry read, translate. Mount static web directory.
• Estimate: 8 hours

PP-108: TypeScript React Frontend

• Description: Build Vite TS PWA application. Implement dual-mode API: live server fetch with automatic localStorage simulation failover. Create modern dashboard.
• Estimate: 14 hours

PP-109: Live Telemetry

• Description: Add psutil helper to poll active CPU utilization and RAM footprint during inference steps. Serve telemetry API.
• Estimate: 4 hours

PP-110: CI/CD Quality Gates Setup

• Description: Configure pre-commit framework. Implement at least 10 pipeline checks in .gitlab-ci.yml running in a local GitLab environment.
• Estimate: 6 hours