AI-Powered Bid and Tender Intelligent Search

The screening of bidding and tender information typically faces two major challenges: fragmented information sources requiring manual queries across platforms, and the inflexibility of traditional keyword searches, which struggle to precisely match complex needs (e.g., 'recent medical equipment purchases by hospitals in a specific region') and often return a large volume of irrelevant results. This skill aims to solve this problem by using natural language understanding to translate vague requirements into structured search instructions and connecting to professional bidding data sources for retrieval.

Core Capabilities and Workflow

The skill's core is an automated pipeline, with key steps as follows:

1. Natural Language Intent Parsing: The user inputs an original query (e.g., 'hospital bed procurement projects in Wuhan over the last month'). The skill calls the /outer-gateway/bbid/aiSearchSubmitPolling endpoint to rewrite the natural language into structured search conditions, including keyword synonyms, industry codes (industryCodes), area codes (areaCode), and party information (buyer/seller).
2. Key and Service Management: The skill reads configuration from environment variables BBIAO_API_KEY and BBIAO_SERVER_URL (default: https://gate.gov-bid.com). Key management follows a security-first principle: it guides users to obtain a key if not configured and persists it for reuse in subsequent tasks if configured.
3. Structured Search Execution: The conditions obtained from the previous step are mapped to specialized search parameters and passed to the /outer-gateway/bid/SearchProjectForAI endpoint. Search results are ranked by relevance and timeliness, containing key fields like title, summary, amount, buyer/seller, and deadline.
4. Result Explanation and Dynamic Adjustment: The skill explains why each result matches the original requirement. When results are zero or excessive, it performs single-dimension adjustments based on predefined rules: for zero results, it sequentially removes exclusion words, relaxes the region, and reduces required words; for excessive results, it adds required words, shortens the date range, or restricts company names.
5. Detail Supplementation: If needed by the user, more detailed structured data (e.g., budget, winning bid amount, contact person) can be retrieved by calling the /outer-gateway/bbid/getZTBStructreDetail endpoint, passing the id and publishTime.

Applicable Boundaries and Important Notes

This skill is not omnipotent; usage requires attention to the following:
* Dependency on Data and Services: All search capabilities are entirely dependent on the backend APIs provided by Shibo Technology (世舶科技). Service interruptions or API changes will directly impact functionality. Callers must ensure the BBIAO_API_KEY is valid and has sufficient quota.
* Auxiliary Nature of AI Reasoning: The synonyms, company names, area codes, etc., generated by the 'AI Rewrite' process are search auxiliary conditions, not verified facts. The final search condition summary presented to users must clearly indicate which parts are AI-generated and which are user-confirmed.
* Search Depth Limitation: To balance performance, the pageNumber for the AI-specific search does not exceed 100. For complex analysis tasks requiring deep traversal of massive datasets, batch queries may be necessary.
* Security and Compliance: The API key (BBIAO_API_KEY) is the credential for accessing the service, and the skill is designed not to leak it in outputs or logs. However, it is crucial to store it securely in environment variables or protected credential files to avoid plaintext exposure.