# Project Configuration (LTM)

*This file contains the stable, long-term context for the project.*
*It should be updated infrequently, primarily when core goals, tech, or patterns change.*

---

## Core Goal

Create a system to compare and reconcile mechanical manifest data (chutes and conveyors) against extracted device data from DWG files, presenting discrepancies and conflicts in a lightweight, searchable frontend.

---

## Tech Stack

* **Backend:** Python Flask on Linux
* **Frontend:** Lightweight SPA (e.g. React or Vanilla JS) with a data‐table library for tabular diff views
* **Data Sources:**

  * XLSX of mechanical manifest (chutes & conveyors)
  * XLSX of DWG‑extracted device list
  * Repository of SCADA object metadata (name properties)
* **Data Processing:** pandas (XLSX ingestion & diff logic)
* **Testing:** pytest
* **Linting/Formatting:** flake8, black

---

## Critical Patterns & Conventions

* **Manifest Ingestion:** Unified loader for XLSX sources, converting to a common schema before comparison
* **Diff Algorithm:** Align entries by identifier, flag missing, extra, and mismatched metadata
* **API Design:** RESTful endpoints under `/api/v1/`:

  * `GET /manifests` – list loaded manifests
  * `POST /compare` – run diff between specified sources
  * `GET /conflicts` – retrieve paginated conflict records
* **Frontend Components:**

  * Table view with search & filter by source
  * Conflict highlighting (added, removed, changed)
  * Modular, single‐file components to minimize bundle size
* **Metadata Loading:** SCADA object name properties loaded at startup from version‐controlled repo
* **Error Handling:** Return structured JSON errors with HTTP status codes
* **Commit Messages:** Conventional Commits format

---

## Key Constraints

* Must run on a Linux server (no Windows dependencies)
* Frontend bundle size < 200 KB gzipped
* Handle XLSX files up to 50 MB without timeouts
* All configuration and manifests stored in plain files or Git (no external DB)

---

## Tokenization Settings

* **Estimation Method:** Character-based
* **Characters Per Token (Estimate):** 4