Structured fields
Issuer teams enter fixed financial and continuity values directly.
IPO Diagnostic architecture
Extract facts with AI. Compare thresholds with rules.
Documents become resolved facts before deterministic listing tests run. Missing values stay missing instead of being guessed.
Uploaded materials
Agent1 and an LLM extract facts from documents, financials, and tables.
CompanyProfile
A shared resolved-value object. Missing values become MISSING_INPUT.
One fact model connects two input paths.
Manual fields and extracted documents resolve into CompanyProfile. The same versioned rule engine then produces an auditable gap report.
Hard rule engine
Numeric thresholds are deterministic, repeatable, and versioned.
Soft signal queue
Qualitative issues remain queued for retrieval and professional review.
Gap report
The result shows evidence, shortfalls, missing inputs, and review items without a listing verdict.
Hard thresholds deliberately avoid model judgment.
Profit, revenue, market capitalisation, continuity, and foreign-exchange gates are comparisons, not language tasks. The engine never invents a number to complete a test.
A status model built for review
PASS
A resolved value meets the modeled threshold.
SHORTFALL
A resolved value falls below the threshold.
MISSING_INPUT
A required value was not supplied or resolved.
INDETERMINATE
A comparison lacks date or exchange-rate context.
REVIEW_REQUIRED
Professional judgment is needed.
DEFERRED_REVIEW
A modeled qualitative area is tracked but not scored.
Modeled pathways
HKEX Main Board Rule 8.05
Chapter 8A WVR quantitative subset
Chapter 18C specialist technology
Chapter 18A and CSRC filing placeholders
Seven qualitative substance signals