National Bidding AI Search Assistant

The Core Problem It Solves

Querying public procurement and bidding information presents three major pain points for business, sales, and bidding teams: fragmented data sources (requiring traversal of multiple government procurement sites), complex query syntax (demanding understanding of specialized parameters like regional codes, industry classifications, and announcement types), and the natural language-to-API gap (spoken requirements cannot directly translate into executable API queries). For instance, a request like “find me recent medical equipment procurement opportunities in Guangdong” necessitates manual decomposition of keywords, supplementation of time ranges, selection of the correct endpoint, and parsing of raw results—an inefficient and error-prone process.

How the Skill Works

This skill functions as an AI intermediary layer whose core capability is converting a user’s natural language intent into precise calls to the Shibo Technology data APIs. Its workflow centers on these key steps:

1. Intent Parsing & Structured Condition Extraction:

   • First, it identifies the core intent of the user’s query (e.g., searching announcements, querying project details, analyzing contracts).
   • It automatically extracts explicit conditions from the natural language sentence, such as keywords, region, procurement unit, and budget amount range.
   • For missing necessary conditions, it applies conservative completion rules: default time is the last 90 days, default region is nationwide, and default announcement types cover common stages from procurement intention to contract notices. The output explicitly lists both “identified conditions” and “auto-completed conditions.”

2. Intelligent Condition Generation & API Routing:

   • For complex or ambiguous descriptions, it prioritizes calling the aiSearchSubmitPolling interface, where an AI model generates a set of executable search conditions.
   • When industry terms are involved, it calls industryReasoning to infer industry codes or directions.
   • Based on user intent and condition complexity, it intelligently routes API calls: simple queries might directly use SearchProjectForAI; more advanced filtering, pagination, or statistics would route to searchProjectApi. Specific needs for contracts or upcoming projects would separately invoke searchProjectContactApi and searchNZJProjectApi.

3. Data Retrieval, Parsing, and Structured Output:

   • After fetching raw data from the selected API, it filters out irrelevant content like recruitment ads or news.
   • Results are presented in a clear structured format: key fields (project name, region, amount, procurement unit, source link, etc.) are displayed in tables.
   • As needed, it can make follow-up calls to detail endpoints (e.g., getZTBStructreDetail, getNZJProjectDetail) to retrieve project text, attachment lists, or original announcement links, and provide summaries of commercial opportunities or contract essentials.

Scope of Applicability and Important Notes

This skill is a data retrieval and query enhancement tool with well-defined boundaries:
* Strictly Uses Designated APIs: It exclusively uses the 13 standard interfaces provided by Shibo Technology and should not or cannot add or call data sources from other providers.
* Output is Fact-Based: All returned information (including amounts, contacts, attachments) comes directly from API responses. Undisclosed information is marked as “not disclosed,” and fabrication is prohibited.
* Focuses on the Bidding Domain: It excels at handling queries related to bids, tenders, awards, contracts, and upcoming project opportunities. Non-bidding content (e.g., pure recruitment, news) is automatically identified and excluded.
* Requires Valid API Credentials: Operation is contingent on a valid BBIAO_API_KEY environment variable. It will explicitly prompt for the key in case of absence or insufficient quota, and will not bypass authentication.
* Acts as a Component in Agent Pipelines: It is typically triggered when a user initiates a procurement-related query. Its output (structured conditions, project lists) can serve as input for subsequent CRM data entry, sales lead analysis, or RAG document retrieval.