Taiwan Stock Coverage Database

A structured equity research database covering 1,735 Taiwan-listed companies (TWSE + OTC) across 99 industry sectors. Each report contains a business overview, supply chain mapping, customer/supplier relationships, and financial data — all cross-referenced through 4,900+ wikilinks that form a searchable knowledge graph.

Why This Exists

Taiwan's stock market has 1,800+ listed companies, many of which are critical nodes in global supply chains (semiconductors, electronics, automotive, textiles). Public information is fragmented across Chinese-language filings, investor presentations, and industry reports. This database consolidates that research into a consistent, searchable format.

The wikilink graph is the core feature. Searching [[Apple]] reveals 207 Taiwanese companies in Apple's supply chain. Searching [[CoWoS]] shows every company involved in TSMC's advanced packaging. Searching [[光阻液]] (photoresist) maps every supplier and consumer of that material.

Quick Start

Prerequisites

pip install yfinance pandas tabulate

Browse Reports

Reports are markdown files organized by industry:

Pilot_Reports/
├── Semiconductors/           (155 tickers)
│   ├── 2330_台積電.md        # TSMC
│   ├── 2454_聯發科.md        # MediaTek
│   └── ...
├── Electronic Components/    (267 tickers)
├── Computer Hardware/        (114 tickers)
└── ... (99 sector folders)

Each report follows a consistent structure:

# 2330 - [[台積電]]

## 業務簡介
**板塊:** Technology
**產業:** Semiconductors
**市值:** 47,326,857 百萬台幣
**企業價值:** 44,978,990 百萬台幣

[Traditional Chinese business description with [[wikilinks]]...]

## 供應鏈位置
**上游:** [[ASML]], [[Applied Materials]], [[SUMCO]]...
**中游:** **台積電** (晶圓代工)
**下游:** [[Apple]], [[NVIDIA]], [[AMD]], [[Broadcom]]...

## 主要客戶及供應商
### 主要客戶
- [[Apple]], [[NVIDIA]], [[AMD]], [[Qualcomm]]...
### 主要供應商
- [[ASML]], [[Tokyo Electron]], [[Shin-Etsu]]...

## 財務概況
### 估值指標
| P/E (TTM) | Forward P/E | P/S (TTM) | P/B | EV/EBITDA |
[Valuation multiples from yfinance]

### 年度/季度財務數據
[Annual (3yr) and Quarterly (4Q) financial tables with 14 metrics]

Add a New Ticker

python scripts/add_ticker.py 2330 台積電
python scripts/add_ticker.py 2330 台積電 --sector Semiconductors

Update Financial Data

python scripts/update_financials.py 2330                        # Single ticker
python scripts/update_financials.py 2330 2454 3034              # Multiple tickers
python scripts/update_financials.py --batch 101                 # By batch
python scripts/update_financials.py --sector Semiconductors     # By sector
python scripts/update_financials.py                             # ALL tickers

Update Valuation Only (Fast)

Refreshes only the 估值指標 table (P/E, Forward P/E, P/S, P/B, EV/EBITDA, stock price) without re-fetching full financial statements. ~3x faster than update_financials.

python scripts/update_valuation.py 2330                         # Single ticker
python scripts/update_valuation.py --batch 101                  # By batch
python scripts/update_valuation.py --sector Semiconductors      # By sector
python scripts/update_valuation.py                              # ALL tickers

Update Enrichment Content

Prepare a JSON file with enrichment data, then apply:

python scripts/update_enrichment.py --data enrichment.json 2330
python scripts/update_enrichment.py --data enrichment.json --batch 101
python scripts/update_enrichment.py --data enrichment.json --sector Semiconductors

JSON format:

{
  "2330": {
    "desc": "台積電為全球最大晶圓代工廠，專注於 [[CoWoS]]、[[3奈米]] 先進製程...",
    "supply_chain": "**上游:**\n- [[ASML]]...\n**中游:**\n- **台積電**...\n**下游:**\n- [[Apple]]...",
    "cust": "### 主要客戶\n- [[Apple]]...\n\n### 主要供應商\n- [[ASML]]..."
  }
}

Audit Quality

python scripts/audit_batch.py 101 -v      # Single batch
python scripts/audit_batch.py --all -v    # All batches

The audit checks: minimum 8 wikilinks, no generic terms in brackets, no placeholders, no English text, metadata completeness, and section depth.

Rebuild Wikilink Index

python scripts/build_wikilink_index.py

Regenerates WIKILINKS.md — a browsable index of all 4,900+ wikilinks categorized by type (Technologies, Materials, Applications, Companies). Run after any enrichment update.

Discover Companies by Buzzword

Hear a buzzword on the news? Find every related Taiwan-listed company instantly.

python scripts/discover.py "液冷散熱"                    # Search all sectors
python scripts/discover.py "液冷散熱" --smart            # Auto-detect relevant sectors
python scripts/discover.py "液冷散熱" --apply            # Tag [[wikilinks]] in reports
python scripts/discover.py "液冷散熱" --apply --rebuild  # Also rebuild themes + network
python scripts/discover.py "液冷散熱" --sector Semiconductors  # Limit to specific sector

Results show companies grouped by relationship type (core business, supply chain, customer/supplier) with context snippets. Use --smart to auto-filter irrelevant sectors (tech buzzwords skip banks/insurance/real estate).

Generate Wikilink Network Graph

Interactive D3.js force-directed graph showing wikilink co-occurrences across all tickers. Hover to highlight neighbors, search by name, adjust edge weight threshold.

python scripts/build_network.py                    # Default: min 5 co-occurrences
python scripts/build_network.py --min-weight 10    # Fewer edges, cleaner view
python scripts/build_network.py --top 200          # Only top 200 nodes

Open network/index.html in your browser. Node colors: red = Taiwan company, blue = international, green = technology, orange = material, purple = application.

Generate Thematic Investment Screens

python scripts/build_themes.py               # Build all 20 themes
python scripts/build_themes.py "CoWoS"       # Single theme
python scripts/build_themes.py --list        # List available themes

Generates themes/ — supply chain maps for key investment themes. Each page shows companies grouped by upstream/midstream/downstream role. See themes/README.md for the full index.

Token Usage & Cost Guide

Tools fall into two categories: Python scripts (free, run locally) and Claude Code skills (consume API tokens for AI research).

Free — Python Scripts (No Tokens)

These run 100% locally with Python + yfinance. No AI, no API cost.

Script	Command	What it does
Update Financials	`python scripts/update_financials.py [scope]`	Refresh financial tables from yfinance
Update Valuation	`python scripts/update_valuation.py [scope]`	Refresh P/E, P/B, EV/EBITDA only (fast)
Update Enrichment	`python scripts/update_enrichment.py --data <json> [scope]`	Apply pre-prepared enrichment data
Audit	`python scripts/audit_batch.py <batch> -v`	Quality check reports
Discover (search)	`python scripts/discover.py "<buzzword>"`	Scan reports for keyword matches
Build Themes	`python scripts/build_themes.py`	Generate thematic supply chain pages
Build Network	`python scripts/build_network.py`	Generate interactive D3.js graph
Build Wikilink Index	`python scripts/build_wikilink_index.py`	Rebuild WIKILINKS.md

Consumes Tokens — Claude Code Skills (Requires AI)

These use Claude AI for web research, content generation, and intelligent enrichment. They require Claude Code and consume API tokens.

Slash Command	Token Usage	What it does
`/add-ticker 2330 台積電`	Medium	Generate .md + fetch financials + AI researches business desc, supply chain, customers
`/update-enrichment 2330`	Medium	AI re-researches and rewrites business content (preserves financials)
`/discover 液冷散熱`	Low-High	Scans database (free) → if no results, AI researches online and enriches reports

Token cost drivers:

/add-ticker: ~1 web search + content generation per ticker
/update-enrichment: ~3-5 web searches + content synthesis per ticker
/discover with results: zero tokens (Python scan only)
/discover without results: varies by research depth (web searches + file edits)

Tip: For bulk operations, use Python scripts directly. Use slash commands for individual tickers or when AI research is needed.

Wikilink Graph

Browse the full index: WIKILINKS.md

The database contains 4,900+ unique wikilinks across three categories:

Category	Examples	Purpose
Companies	`[[台積電]]`, `[[Apple]]`, `[[Bosch]]`	Map supply chain relationships
Technologies	`[[CoWoS]]`, `[[HBM]]`, `[[矽光子]]`, `[[EUV]]`	Find all companies in a technology ecosystem
Materials	`[[光阻液]]`, `[[碳化矽]]`, `[[ABF 載板]]`	Track material suppliers and consumers

Top Referenced Entities

Entity	Mentions	What it reveals
`[[台積電]]`	469	Taiwan's semiconductor ecosystem revolves around TSMC
`[[NVIDIA]]`	277	AI supply chain — who makes NVIDIA's components
`[[Apple]]`	207	Apple's Taiwanese supplier network
`[[AI 伺服器]]`	237	AI server supply chain mapping
`[[電動車]]`	223	EV component suppliers
`[[5G]]`	232	5G infrastructure companies
`[[PCB]]`	263	Printed circuit board ecosystem

Project Structure

├── CLAUDE.md                  # Project rules and quality standards
├── WIKILINKS.md               # Browsable wikilink index (auto-generated)
├── task.md                    # Batch definitions and progress tracking
├── requirements.txt           # Python dependencies
├── README.md
├── scripts/
│   ├── utils.py               # Shared utilities (file discovery, wikilink normalization)
│   ├── add_ticker.py          # Generate new ticker reports
│   ├── update_financials.py   # Refresh financial tables + valuation multiples
│   ├── update_enrichment.py   # Update business descriptions from JSON
│   ├── audit_batch.py         # Quality auditing
│   ├── update_valuation.py     # Refresh valuation multiples only (fast)
│   ├── discover.py            # Reverse search: buzzword → related companies
│   ├── build_wikilink_index.py # Rebuild WIKILINKS.md index
│   ├── build_themes.py        # Generate thematic investment screens
│   ├── build_network.py       # Generate interactive network graph
│   └── generators/            # Historical base report generators
├── Pilot_Reports/             # 1,735 ticker reports across 99 sectors
│   ├── Semiconductors/
│   ├── Electronic Components/
│   └── ... (99 folders)
├── network/                   # Interactive wikilink network graph (auto-generated)
│   ├── index.html             # D3.js visualization (open in browser)
│   └── graph_data.json        # Raw graph data (339 nodes, 1,452 edges)
├── themes/                    # Thematic investment screens (auto-generated)
│   ├── README.md              # Theme index
│   ├── CoWoS.md               # 39 companies in CoWoS supply chain
│   ├── AI_伺服器.md            # 148 companies in AI server ecosystem
│   ├── NVIDIA.md              # 104 companies in NVIDIA supply chain
│   └── ... (20 themes)
└── .claude/
    └── skills/                # Claude Code skill definitions

Quality Standards

Every report is validated against 8 quality rules (defined in CLAUDE.md):

Wikilinks must be specific proper nouns — no generic terms like 供應商 or 大廠
Ticker-company identity verification — filename is ground truth
Minimum 8 wikilinks per report
Financial tables preserved — never modified during enrichment
All content in Traditional Chinese
No placeholders in completed reports
Complete metadata (sector, industry, market cap, enterprise value)
Segmented supply chain — upstream/midstream/downstream by category

Current audit score: 1,733/1,733 (100%) pass all quality checks.

Data Sources

Financial data: yfinance (Yahoo Finance Taiwan)
Business content: Company IR pages, MOPS filings (公開資訊觀測站), investor conference transcripts (法說會), annual reports (年報)
Supply chain data: Industry reports, news sources, company disclosures

Limitations

Financial data depends on yfinance availability — some OTC stocks may have gaps
Business descriptions reflect research as of the enrichment date — they don't auto-update
Wikilinks are manually curated — new technologies or companies need manual addition
Content is in Traditional Chinese — English speakers will need translation

Wikilink Naming Convention

Category	Canonical form	Examples
Taiwan companies	Chinese	`[[台積電]]`, `[[鴻海]]`, `[[聯發科]]`
Foreign companies	English	`[[NVIDIA]]`, `[[Samsung]]`, `[[Micron]]`
Materials/substrates	Chinese	`[[碳化矽]]`, `[[氮化鎵]]`, `[[電動車]]`
Industry acronyms	Acronym	`[[PCB]]`, `[[CPO]]`, `[[HBM]]`, `[[CoWoS]]`

Wikilink normalization is built into the enrichment pipeline — aliases are automatically merged to canonical form on every write.

Contributing

Contributions are welcome. When adding or updating ticker reports:

Follow the quality rules in CLAUDE.md
Run python scripts/audit_batch.py --all -v before submitting
Ensure every [[wikilink]] is a specific proper noun
Verify the company name matches the ticker number

License

MIT License. See LICENSE for details.

Financial data sourced from Yahoo Finance via yfinance. Business descriptions are original research.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

Taiwan Stock Coverage Database

Why This Exists

Quick Start

Prerequisites

Browse Reports

Add a New Ticker

Update Financial Data

Update Valuation Only (Fast)

Update Enrichment Content

Audit Quality

Rebuild Wikilink Index

Discover Companies by Buzzword

Generate Wikilink Network Graph

Generate Thematic Investment Screens

Token Usage & Cost Guide

Free — Python Scripts (No Tokens)

Consumes Tokens — Claude Code Skills (Requires AI)

Wikilink Graph

Top Referenced Entities

Project Structure

Quality Standards

Data Sources

Limitations

Wikilink Naming Convention

Contributing

License

About

Uh oh!

Releases

Packages

Uh oh!

Contributors

Uh oh!

Languages

Name		Name	Last commit message	Last commit date
Latest commit History 15 Commits
.claude/skills		.claude/skills
Pilot_Reports		Pilot_Reports
network		network
scripts		scripts
test		test
themes		themes
.gitignore		.gitignore
CLAUDE.md		CLAUDE.md
LICENSE		LICENSE
README.md		README.md
WIKILINKS.md		WIKILINKS.md
requirements.txt		requirements.txt
task.md		task.md
tw_coverage_app.py		tw_coverage_app.py

Folders and files

Latest commit

History

Repository files navigation

Taiwan Stock Coverage Database

Why This Exists

Quick Start

Prerequisites

Browse Reports

Add a New Ticker

Update Financial Data

Update Valuation Only (Fast)

Update Enrichment Content

Audit Quality

Rebuild Wikilink Index

Discover Companies by Buzzword

Generate Wikilink Network Graph

Generate Thematic Investment Screens

Token Usage & Cost Guide

Free — Python Scripts (No Tokens)

Consumes Tokens — Claude Code Skills (Requires AI)

Wikilink Graph

Top Referenced Entities

Project Structure

Quality Standards

Data Sources

Limitations

Wikilink Naming Convention

Contributing

License

About

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages