Skip to content

replit.md

Latest commit

 

History

History
73 lines (57 loc) · 3.01 KB

replit.md

File metadata and controls

73 lines (57 loc) · 3.01 KB
Raw

LCY3 AFM Dashboard — Floor Health

A Streamlit data visualization tool for monitoring Floor Health and Andon performance at an Amazon fulfillment center (LCY3). Built by Manish Karki.

Tech Stack

  • Python 3.12
  • Streamlit (port 5000, 0.0.0.0)
  • Pandas, Plotly, openpyxl, pyarrow

Project Structure

app.py              — Main Streamlit application
history_db.py       — Persistent storage layer (Parquet + JSON)
report_builder.py   — Excel report generation (daily + weekly .xlsx)
requirements.txt
.streamlit/config.toml

history/            — Persistent upload history (created at runtime)
├── daily/          — Per-day aggregated Parquet files
├── weekly/         — Per-week aggregated Parquet files
├── processed/      — Full raw DataFrames (one per upload)
└── metadata.json   — Index of all saved records

.andon_data/        — Legacy storage location (backward-compat, read-only)

Key Features

Dashboard Tabs

  1. Overview — bar chart of top resolvers, donut of andon types, daily trend spline
  2. Leaderboard — ranked resolver table with colour-coded performance badges
  3. AFM Profile — individual resolver deep-dive
  4. Root Cause — recurring issue analysis
  5. AFM Performance — team performance metrics
  6. By Andon Type / Equipment / Zone / Shift / Blocking / Equipment ID — pivot breakdowns (optional columns unlock these)
  7. Weekly Breakdown — weekly trend charts
  8. Hourly Trend / Heatmap — time-of-day analysis
  9. Raw Data — full filtered DataFrame view
  10. Export — download multi-sheet Excel reports
  11. History — browse saved daily, weekly, and processed datasets per upload

Persistent Storage (history_db.py)

Every uploaded file is saved to history/:

  • processed/<hash>.parquet — full raw data
  • daily/<hash>_daily.parquet — aggregated by date: total andons, avg resolve time, resolved count, top andon type, top resolver
  • weekly/<hash>_weekly.parquet — same metrics grouped by ISO week number
  • metadata.json — index with file name, upload date, week numbers, date range, andon count

Key functions: compute_hash, hash_exists, get_existing_name, record_upload, load_dataframe, load_daily, load_weekly, get_history, remove_entry, clear_history

Sidebar

  • Dark mode toggle
  • Saved Datasets list — Load/Unload and Remove buttons per file, ACTIVE badge, Clear All

Status Labels

  • ✅ On target
  • ⚠️ Above target
  • ⚠️ Average (was "🚨 Slow" — renamed)

Duplicate Detection

SHA-256 hash of uploaded file bytes; re-uploads are blocked and auto-pointed to the saved copy.

Session State Keys

  • dark_mode — boolean
  • saved_active_hashes — list of file hashes currently loaded into the dashboard
  • duplicate_warnings — list of warning messages

Required Input Columns

Status, Resolver, Andon Type, Dwell Time (hh:mm:ss), Time Created

Optional Columns (unlock extra tabs)

Equipment Type, Zone, Shift, Blocking, Equipment ID