Google Patents

# Google Patents

Search and retrieve patent data via SerpApi. Requires SERPAPI_API_KEY env var (free: 100/month at serpapi.com).

## 5 Commands

```bash
bash scripts/patents.sh search "keywords" [options]     # Search patents
bash scripts/patents.sh detail "US11734097B1"            # Basic info + claims
bash scripts/patents.sh fulltext "US11734097B1"          # Description full text
bash scripts/patents.sh full "US11734097B1"              # ALL data in one call
bash scripts/patents.sh pdf "US11734097B1" output.pdf    # Download PDF
```

Patent ID: short `US11734097B1` or full `patent/US11734097B1/en`. Supports all countries: CN, US, EP, JP, KR, WO, DE, etc.

## Search Options

```
--country US,CN,JP,WO,EP,KR    --status GRANT|APPLICATION
--type PATENT|DESIGN            --assignee "Company"
--inventor "Name"               --sort relevance|new|old
--after publication:20230101    --before publication:20251231
--num 10-100                    --page N
--language ENGLISH|CHINESE      --litigation YES|NO
--scholar                       --clustered
```

Boolean: `"(massage) AND (glove OR mitt)"` | Multi-term + CPC: `"(pet grooming);(A01K13)"`

## What Each Command Returns

**search**: patent_id, title, snippet, assignee, inventor, dates, pdf, country_status
**detail**: title, abstract, claims[], inventors[], assignees[], classifications[], legal_events[], citations, similar_documents[], images[], pdf, family_id, worldwide_applications
**fulltext**: description full text (FIELD OF INVENTION, BACKGROUND, SUMMARY, DETAILED DESCRIPTION)
**full**: Everything from detail + description_full combined
**pdf**: Downloads PDF file to specified path

## E-commerce Scenarios

```bash
# Infringement risk check (pre-listing must-do)
bash scripts/patents.sh search "product" --type DESIGN --country US --status GRANT

# Competitor patents
bash scripts/patents.sh search "category" --assignee "Company" --num 50

# Read claims to assess real risk
bash scripts/patents.sh full "USD975937S1"

# Download patent PDF for reference
bash scripts/patents.sh pdf "USD975937S1" ./patent.pdf

# Expired patents (free to use)
bash scripts/patents.sh search "tech" --before "filing:20040101"

# Latest trends
bash scripts/patents.sh search "tech" --sort new --after "publication:20240101"

# Litigation-prone patents
bash scripts/patents.sh search "product" --litigation YES --country US
```

## Error Handling

All errors return JSON with `error` and `code` fields. No exceptions thrown.

| Code | Meaning |
|---|---|
| PATENT_NOT_FOUND | Patent ID doesn't exist (404) |
| AUTH_ERROR | Invalid/expired API key (401/403) |
| MAX_RETRIES_EXCEEDED | Network failure after 3 retries |
| NO_DESCRIPTION | Patent has no description text |
| PARSE_ERROR | HTML parsing failed |
| NO_PDF | No PDF available |
| DOWNLOAD_ERROR | PDF download failed |
| MISSING_QUERY | No search query provided |
| MISSING_ID | No patent ID provided |

Auto-retry: 3 attempts with exponential backoff (2s, 4s, 8s) on 429/5xx errors.
Rate limit: 1 second between requests to avoid triggering anti-scraping.
Timeouts: 10s connect, 30s max per request, 60s for PDF downloads.

## FAQ

**Q: Why are some fields empty?**
A: Different countries have different patent page formats. Some patents may have incomplete data, or the description may not be digitized.

**Q: Can I batch-fetch patents?**
A: Yes, loop through IDs. Respect the 1s rate limit. Free tier = 100 calls/month. Cached results (same query within 1h) are free.

**Q: How to get PDF?**
A: `bash scripts/patents.sh pdf "US11734097B1" output.pdf`

**Q: Chinese vs English patents?**
A: Chinese pages (patent/CNxxxxxx/zh) have native Chinese content. English pages have machine-translated content. Use `--language CHINESE` for search.

**Q: Patent ID formats?**
A: Country code + number + type suffix. Examples: CN106484775A, US20180232442A1, EP2264377A2, USD975937S1 (design), JP2020123456A.