Source Tenders

Scrape and analyze Taiwan government procurement tenders from web.pcc.gov.tw.

Quick Start

# 1. Scrape tenders (last 7 days by default)
uv run scripts/scraper.py > tenders.json

# 2. Analyze and generate HTML report
uv run scripts/analyze.py

# 3. Open the report
open analysis.html

Scraping

# Date options
uv run scripts/scraper.py --days 14              # Last 14 days
uv run scripts/scraper.py --date 2026-01-27      # Specific date
uv run scripts/scraper.py --start 2026-01-20 --end 2026-01-27

# Output formats
uv run scripts/scraper.py --output json          # Default
uv run scripts/scraper.py --output table         # Human-readable
uv run scripts/scraper.py --output csv           # CSV format

Output Fields

Field	Description
`notice_date`	Publication date
`deadline`	Bid submission deadline
`agency`	Procuring agency
`tender_id`	Tender case number
`title`	Tender title
`tender_method`	公開招標/選擇性招標/限制性招標
`procurement_type`	工程類/財物類/勞務類
`budget`	Budget amount (TWD integer)
`url`	Detail page link
`is_amendment`	Amendment notice flag

Analysis

# Analyze tenders.json → analysis.html + filtered_tenders.json
uv run scripts/analyze.py

# Custom input/output
uv run scripts/analyze.py --input data.json --output report.html

Presets

Analysis supports multiple scoring presets:

Preset	Focus	High-value Keywords
Software	Web/app/system development	網站, APP, API, 系統, 軟體
AI	AI/ML/intelligent systems	人工智慧, 機器學習, 智慧, 深度學習

HTML Report Features

Switch presets (matched/unmatched updates dynamically)
Filter by: procurement type, budget range, tags (multi-select)
Sort by any column
Click rows to expand details with all preset scores

Typical Workflow

# In a context folder
cd projects/pcctw/contexts/YYYYMMDD-sourcing/

# Scrape
uv run ../../skills/source-tenders/scripts/scraper.py \
  --start 2026-01-20 --end 2026-01-27 > tenders.json

# Analyze
uv run ../../skills/source-tenders/scripts/analyze.py

# Review
open analysis.html

Technical Notes

Source: web.pcc.gov.tw/prkms/tender/common/basic/readTenderBasic
Pagination: 100 items/page, auto-fetches all
Rate limiting: 1.5s delay between pages (adjustable with --delay)
Volume: ~1,500-2,000 tenders/day across all categories

Output Format (tenders.json)

{
  "scraped_at": "2026-01-31T21:53:38.790223",
  "date_range": {"start": "2026-01-26", "end": "2026-01-31"},
  "total_count": 5013,
  "tenders": [
    {"tender_id": "...", "url": "https://web.pcc.gov.tw/...", ...},
    ...
  ]
}

Used by: /pcctw-tender-research-review reads tenders.json to add direct PCC detail links to the review page.

source-tenders