Models and tools
Record actual model information, inference inputs and outputs, tool activity, evaluation and policy details.
Product
AI transaction traces and verifiable records
Sigillum Finans connects observable application steps into a transaction trace. Teams can inspect the recorded information, verify its integrity and compare selected fields with institutional source records.
01 / Traceability
Review the execution behind an outcome
An AI-enabled workflow can include model calls, document retrieval, service requests, business rules and human intervention. Sigillum relates the observations received through supported integrations using the actual execution context.
Inspect the inputs, outputs, references and explanations supplied by the application. The record shows what was observed; it does not reconstruct hidden model reasoning.
Documents and references
Retain supplied document identifiers, versions, retrieval results and original content relevant to the transaction.
Application outcomes
Add the application’s recorded result, explanation and business references through the shared SDK interfaces.
HOW IT WORKS
Review a response in its transaction context.
Model, document, service and human-review records provide different evidence for one transaction.
THE INSTITUTION’S SERVERS AND NETWORK
- 01Input and response
Model and prompt
Recorded interaction, model and prompt references.
- 02Retrieval / RAG
Knowledge sources
Reported document references, contents and versions.
- 03Call relationships
Tools and services
Inputs, outputs and outcomes of dependent services.
- 04Decision evidence
Human review and outcome
Reported review, intervention and application outcome.
02 / Review
Navigate from a transaction to its individual records
Use application, date and status filters to locate a transaction. While it is being prepared, expand its observed progress. Once ready, open the summary or the complete record stream.
Inspect failures alongside the affected service, time and available explanation. A failed business operation can still produce a complete, sealed trace.
Observed progress
Received records and the current step and preparation states, without inferred completion percentages.
Technical details
Original content and identifiers remain available without overwhelming the summary.
Integrity result
A separate verification result describes the consistency of recorded evidence with its seals.
TRANSACTION RECORD EXAMPLEIllustrative example
Records linked to one transaction
CONTEXT & REFERENCES↗
Documents and policies used
Review the document, policy and service responses recorded for the transaction, with their version references.
- Document
- Credit assessment policy
- Reference
- Version and content hash
- Record origin
- Service declaration
Retained references show which version was recorded at the time of the decision.
An illustrative example of how it works: review a request, its supporting records, model response and outcome within the institution’s configured recording scope.
03 / Source comparison
Compare selected fields with the institution’s record
Define comparison fields for the application and configure a read-only HTTP JSON or PostgreSQL JSON source. Use a recorded reference to retrieve the corresponding business record.
The comparison definition attached at transaction completion remains associated with that transaction. Review matches, differences, missing values and retrieval failures separately.
04 / Verification
Verify evidence in the portal or offline
Verify recorded content and relationships against the expected seals and trusted keys. Export the evidence and required verification material for the desktop verifier.
Integrity verification does not certify the accuracy of an AI output, the appropriateness of a business decision or compliance with every applicable regulation.
05 / Operations
Manage integration and review within the institution
Configure environments, applications, entry points and service connections in the portal. Reuse an installed SDK in a shared service while preserving application-specific execution context.
Product health, delivery status and security findings support technical operations. Retention, recovery and access arrangements must be defined for the institution’s actual deployment.
PRODUCT INFORMATION
Frequently asked questions
Which teams use Sigillum Finans?
Application engineering, operations, risk and audit teams use transaction records and verification results to examine AI-enabled workflows.
Which languages are supported?
.NET, Java and Python SDKs are available, with separate javax and Jakarta Servlet adapters. The integration guide lists supported runtime and HTTP boundaries.
Does a verified record prove that an AI decision was correct?
No. Verification checks recorded content and its cryptographic integrity. Business accuracy, fairness and legal suitability require their own assessment.
Can evidence be checked outside the portal?
Yes. Use the offline verifier with the corresponding evidence package and verification keys whose trust has been established through an authorised channel.
DISCUSS YOUR REQUIREMENTS
Plan an evaluation for your institution.
Bring your workflow, integration requirements and review questions. We’ll discuss how Sigillum Finans fits your environment.