Tendering and Bidding Business Opportunities AI Search Assistant

Problem to Solve

Business teams in tendering and bidding often face a gap between natural language queries and the precise parameters required by APIs. For instance, a salesperson might ask, "Find school air conditioning procurement tenders in Guangdong for the last three months," but directly calling a tendering data API requires breaking this down into fields like region, time, keywords, and announcement type. Manual configuration is time-consuming and error-prone, especially for vague, cross-industry requests.

How the Skill Works

This skill uses an AI pipeline to convert user colloquial questions into executable searches. Core capabilities include:

1. Intent Parsing and Condition Extraction: First, it identifies whether the user intent is to search announcements, query details, or reason about industries. Then it extracts explicit conditions from the text, such as keywords, region, time range, and announcement type. For example, "last three months" is automatically mapped to a date interval.
2. Conservative Completion and AI Enhancement: For missing conditions, it applies sensible defaults—defaulting to the last 90 days for time and nationwide for region if unspecified. For complex sentences, it calls the aiSearchSubmitPolling interface to generate structured search conditions; for industry phrases like "government cloud," it uses industryReasoning to assist in identifying industry direction.
3. API Call Chain and Result Processing: Based on user intent, it selects interfaces—prioritizing SearchProjectForAI for natural language searches and making secondary calls to getZTBStructreDetail for structured fields when details are needed. Before output, it filters irrelevant content (e.g., recruitment posts), presents project lists in tabular format, and explicitly lists "identified conditions" and "auto-completed conditions" to ensure transparency.

Application Boundaries and Considerations

This skill only reuses 13 specified interfaces from Shibo Technology (e.g., SearchProjectForAI, getZTBStructreDetail), without adding dedicated APIs, thus covering tendering, bidding, contract, and upcoming project data. Data is updated annually with over 66 million records via the Shibo Technology API, but missing fields are marked as "undisclosed" rather than fabricated.

Key limitations include:
- API Dependency: Requires configuration of the BBIAO_API_KEY environment variable; otherwise, calls fail. When quota is insufficient, manual purchase via a designated page is needed—no automatic retries.
- Data Accuracy: Outputs are based on API returns and do not guarantee 100% coverage of all projects, but original links are provided for traceability.
- Error Handling: For empty results, it expands synonyms or relaxes conditions; for parameter errors, it prompts checks on date formats, amount ranges, etc.
- Content Boundaries: Strictly distinguishes tendering opportunities from non-relevant content like recruitment or training, and does not use competitor APIs or fabricate information.

This skill suits business teams seeking quick tendering opportunities and R&D personnel integrating data into Agent or RAG systems, but requires basic API integration knowledge.