shield_lock
金大哥
← 返回 Skill 列表
extension
分类: 其它需要 API Key

bidding-tool

Generate complete bidding technical proposals (投标技术方案) in Word (.docx) format from tender documents, BOQ, and drawings using AI analysis.

person作者: user_d7dc7dcchubcommunity
download下载资源包

Bidding Tool - 投标技术方案自动编写

Overview

This skill generates professional construction/engineering bidding technical proposals (施工技术方案) in .docx format. The pipeline is:

Tender docs (PDF/DOCX) + BOQ (XLSX/CSV) → AI Analysis → Per-chapter AI Generation → Word Document Assembly

Project Location

The bidding tool code lives at C:\Users\lenovo\bidding-tool\. All commands assume this working directory.

Prerequisites (one-time setup)

cd C:\Users\lenovo\bidding-tool

# 1. Install dependencies (run once)
"C:/Users/lenovo/AppData/Local/Programs/Python/Python312/python.exe" -m pip install -e "."

# 2. Initialize config files
"C:/Users/lenovo/AppData/Local/Programs/Python/Python312/python.exe" -m bidding_tool.main init

# 3. Edit .env to add ANTHROPIC_API_KEY
# 4. Edit config/company.yaml with real company info
# 5. Edit config/projects.yaml with real past project cases

Workflow

Step 1: Parse input files (dry-run to verify extraction)

python -m bidding_tool.main parse --tender <招标文件.pdf> --boq <工程量清单.xlsx> --verbose

Parse handles:

  • PDF files (via PyMuPDF): text + tables + images
  • DOCX files (via python-docx): paragraphs + headings + tables
  • Excel BOQ (via openpyxl): auto-detect column mapping by header keywords (序号/名称/工程量/单价...)
  • CSV BOQ (via pandas): same column detection

Step 2: Analyze tender with AI

python -m bidding_tool.main analyze \
  --tender <招标文件.pdf> \
  --boq <工程量清单.xlsx> \
  --output analysis.json

This extracts structured data from tender documents:

  • Project name, location, scale, structure type
  • Scope of work, technical requirements, quality/safety/environmental requirements
  • Schedule requirements, key milestones
  • Applicable standards, owner priorities
  • Identifies key risks and special conditions
  • Auto-detects project type (highway / municipal_construction / building_maintenance / landscaping / property_service / general)

Step 3: Full generation (one command)

python -m bidding_tool.main generate \
  --tender <招标文件.pdf> \
  --boq <工程量清单.xlsx> \
  --output-dir ./output \
  --model <claude-model-name>

Optional flags:

  • --sections ch4_methods,ch5_quality - Generate only specific chapters
  • --no-cache - Skip AI response cache
  • --max-tokens 8192 - Increase output length for complex sections

Step 4: Regenerate a single section

python -m bidding_tool.main section \
  --section-id ch10_key_points \
  --analysis-file analysis.json \
  --format docx \
  --output ./output/revised_chapter10.docx

Generated Document Structure (13 Chapters)

| Chapter | Title | Key Content | |---|---|---| | 1 | 工程概况 | Project overview, scope, site conditions, standards | | 2 | 施工总体部署 | Deployment principles, construction zones, sequence | | 3 | 施工进度计划及保障措施 | Schedule plan, milestones, guarantee measures | | 4 | 主要施工方案与技术措施 | Earthwork, foundation, concrete, steel, MEP, decoration | | 5 | 质量管理体系与保证措施 | Quality objectives, system, process control, inspection | | 6 | 安全管理体系与保证措施 | Safety objectives, hazard control, emergency plan | | 7 | 文明施工与环境保护措施 | Site management, noise/dust, waste, water/soil | | 8 | 劳动力/材料/设备投入计划 | Labor, material supply, equipment plan | | 9 | 施工总平面布置 | Overall layout, traffic, utilities, staging | | 10 | 重点难点分析及对策 | Key risks identification, countermeasures (★ critical for evaluation) | | 11 | 项目管理机构配置 | Organization structure, key personnel, responsibilities | | 12 | 新工艺/新技术应用 | New technology adoption, expected benefits | | 13 | 售后服务及保修承诺 | Warranty period, service commitment, emergency response |

AI Quality Features

  • Project type classification: Auto-detects 6 industry types, injects domain-specific glossary into generation prompts
  • Anti-hollow writing: Detects 62 generic/empty phrases, enforces specific-data-density requirements (concrete dimensions, quantities, dates, standards)
  • Quality auto-retry: Scores each chapter (0-100), auto-regenerates up to 3 times with targeted fix instructions if score < threshold
  • Domain terminology injection: Highway/municipal/building glossaries loaded per project type, injected into AI prompts
  • Content quality checks: Sub-section completeness, table presence, keyword coverage, data point density

Document Output Features

  • Cover page: Project name, company name, date
  • Table of Contents: Auto-generated TOC field (press F9 in Word to update)
  • Preface: Documented methodology statement
  • Headers/Footers: Company name in header, page numbers in footer
  • Professional formatting: SimHei headings, SimSun body, 1.5 line spacing, first-line indent
  • Tables: Properly formatted with alternating row shading

Cache Management

# View cache stats
python -m bidding_tool.main cache-stats

# Clear all cached AI responses
python -m bidding_tool.main cache-clear

# Clear cache for a specific section
python -m bidding_tool.main cache-clear --section ch4_methods

Config Management

# View current config
python -m bidding_tool.main config-show

# Validate config files
python -m bidding_tool.main config-validate

Key Files

  • config/company.yaml - Company profile, qualifications, personnel, standard phrases
  • config/projects.yaml - Historical project case library
  • config/reference/project_types.yaml - 6 industry type definitions with detection keywords
  • config/reference/glossaries.yaml - Domain terminology databases (highway, municipal, building, etc.)
  • config/reference/anti_hollow_rules.yaml - 14 hollow-writing categories + 62 generic phrases + 7 self-test questions
  • src/bidding_tool/ai/prompts/analyzer.py - AI analysis prompt template
  • src/bidding_tool/ai/prompts/generator.py - Per-chapter generation prompt templates + SECTION_INSTRUCTIONS
  • src/bidding_tool/ai/prompts/enhancer.py - Prompt enhancement with glossary + anti-hollow rules
  • src/bidding_tool/ai/quality.py - Content quality scoring and fix suggestion generation
  • src/bidding_tool/ai/classifier.py - Project type classifier (keyword scoring + AI confirmation)
  • src/bidding_tool/references/ - Domain knowledge modules (glossary, anti_hollow, project_types)
  • src/bidding_tool/models/proposal.py - PROPOSAL_OUTLINE (13 chapters definition)
  • src/bidding_tool/generator/assembler.py - DOCX assembly orchestrator
  • src/bidding_tool/generator/styles.py - Word document style definitions
  • src/bidding_tool/generator/sections.py - Markdown-to-DOCX renderer
  • src/bidding_tool/generator/toc.py - TOC field injection + headers/footers

Common Troubleshooting

Python not found: Use the full Python path:

"C:/Users/lenovo/AppData/Local/Programs/Python/Python312/python.exe"

Encoding issues: The tool auto-detects UTF-8 and reconfigures stdout. If you see garbled characters, run:

set PYTHONIOENCODING=utf-8

API errors: Check .env has valid ANTHROPIC_API_KEY. Verify model name matches available models.

Config validation: Run python -m bidding_tool.main config-validate to check company.yaml and projects.yaml syntax.