Xiao Xiao Bid Document Generator

Problem It Addresses

Tender workflows have concrete pain points: tender documents are long and clause-heavy, so disqualification lines, scoring points, and package-specific requirements are easy to miss; after drafting, it is hard to check the bid file line by line against the tender; and multi-package projects also require choosing which package to pursue. This skill connects three steps to the Bailian® Bid Writing open API: interpret the tender document, generate the bid document, and run a compliance review. It suits engineers and bid teams who need to check whether to bid, how to bid, and what risks remain.

How It Works

• Smart interpretation: accepts .pdf, .doc, or .docx tender files, usually up to 50 MB, from a local path or cloud URL. It outputs project info, qualification items, disqualification lines, scoring rules, commercial terms, pricing requirements, procurement background, and bid-control insights, with optional HTML / Word reports.
• Bid drafting: works from an interpreted project and selected package, with total_pages up to 500. It produces a .docx bid file with table of contents, chapters, standard templates, and to-be-filled bookmarks. Uncertain fields are kept as placeholders instead of being invented.
• Compliance review: uses the interpreted tender plus one or more .doc / .docx bid files, each up to 1024 MB. It returns a risk list with high-risk, needs-review, and note items, including tender basis, bid evidence, and revision suggestions, and supports multi-file similarity checks.
• Async execution: tasks run through the App Key via a submit, poll, fetch model; bid generation may take more than 10 minutes and reports progress.

Boundaries And Notes

• Interpretation, package selection, and compliance review do not consume credits, while bid generation does; API / Skill entry points still require a positive balance.
• Users mainly provide files, state requirements, and paste the App Key; background commands are run by the assistant and should not be shown to users.
• Reports must provide full absolute file paths; if semantic review only returns partial results, it should not be described as a complete review.
• Experience capture and issue feedback are consent-based and do not upload file contents or personal information automatically.