AIRS Embodied Bidding Tracker

Problem

When researching embodied intelligence companies, bidding and award records are scattered across Tianyancha. Manually searching each company, copying project names, amounts, and dates is slow and error-prone, especially when short names, full names, or missing fields cause mismatches.

How it works

The skill turns Tianyancha lookups into a repeatable CLI workflow:

• status checks Node.js, Chrome connection, and dependencies
• search confirms full company names and links from a company list, preferring a local enterprise database
• download batch-collects bidding records using --start-date, --end-date, and --min-amount
• query supports fuzzy matching for a single company, candidate selection, and interactive filtering

Typical outputs are data/company_list.csv and data/bidding_records.csv, including company name, project name, notice type, procurer, award amount, publication date, and detail-page URL.

Boundaries

It relies on a logged-in Chrome remote debugging window, and security checks may require manual captcha or slider handling. Keep batch runs under 200 companies and time ranges under one year; high request frequency can trigger rate controls, so pause and retry if needed.