SuperAPI AI Directory

uspto-database

Data & analytics

View repository

Updated 10-15 18:45

About this skill

Source: https://github.com/aipoch/medical-research-skills

When to Use

  • Find relevant US patents by keyword, inventor, assignee, CPC/IPC class, or date range for prior-art screening.
  • Retrieve patent application/prosecution history (PEDS) to understand office actions, events, and timeline status.
  • Check trademark status and documents via TSDR for clearance, monitoring, or portfolio review.
  • Validate patent lifecycle signals such as maintenance-fee-related status or expiration indicators as part of diligence.
  • Pull assignment/ownership-related information to support chain-of-title checks.

Key Features

  • Patent search client for querying USPTO/PatentsView-style endpoints and returning matching patent identifiers/records.
  • PEDS client for fetching examination history and event data for applications.
  • TSDR access for trademark status and document retrieval.
  • Script-oriented workflow designed for automation (batch queries, pipelines, and integration into internal tools).

Dependencies

  • Python >=3.9
  • requests >=2.31.0
  • uspto-opendata-python >=0.3.0

Example Usage

# Example: search patents by abstract text and print results
# Prerequisite: set required API keys in environment variables as expected by your scripts/clients.
# For example (names may vary by implementation):
#   export USPTO_API_KEY="..."
#   export PATENTSVIEW_API_KEY="..."

from scripts.patent_search import PatentSearchClient

def main():
    client = PatentSearchClient()

    # Query example: search for patents whose abstract contains "AI"
    query = {"_text_all": {"patent_abstract": "AI"}}

    results = client.search_patents(query)
    print(results)

if __name__ == "__main__":
    main()

Implementation Details

  • Workflow

    1. Search: Use scripts/patent_search.py to locate relevant patent numbers/records from a text or fielded query.
    2. History: Use scripts/peds_client.py to retrieve prosecution/examination history (events, dates, and status signals).
    3. Status checks: Use retrieved metadata to assess maintenance-fee/expiration-related status where available.
  • Query model

    • Patent search is driven by a JSON query object (e.g., full-text constraints such as "_text_all" over fields like patent_abstract).
    • Returned results are intended to be fed into downstream steps (e.g., PEDS lookups by application/patent identifiers).
  • Authentication

    • API keys for USPTO and/or PatentsView are required and should be provided via environment variables or configuration used by the clients.
  • Networking

    • HTTP requests are performed via requests; ensure timeouts/retries are configured in the underlying clients if running large batches.