# Deterministic Resolution Engine

*/Software/Deterministic_Resolution_Engine*

## Solution Overview

The Deterministic Resolution Engine provides an API layer that enforces rigid structural and logical constraints on probabilistic AI outputs. It evaluates raw unstructured LLM responses against hard-coded compliance rules, mathematical proofs, and exact schema requirements. The engine outputs a strictly verified cryptographically signed payload that guarantees absolute adherence to defined constraints while stripping out hallucinations or formatting anomalies.

FinTech compliance teams and medical billing automation providers integrate this API to eliminate the risk of deploying non-deterministic models in highly regulated environments. Instead of manually reviewing AI-generated claims or transaction flags for subtle probabilistic errors, these buyers rely on the engine to automatically reject or correct outputs that violate statutory logic like Medicare coding dependencies or SEC reporting limits.

Operating as a headless SaaS primitive, the engine consumes raw text streams from foundation models and proprietary LLMs. It functions as a strict gateway for downstream autonomous agents and services-as-software platforms ensuring they only execute transactions based on verified data. When a logic conflict occurs that cannot be mapped deterministically, the engine halts the workflow and routes the anomalous payload to a human compliance officer for a mandatory override decision before releasing it to the execution layer.

## Headless Saas Data Model

**Entities**:
- Name: Resolution · Description: The evaluation task processing unstructured LLM outputs against rigid constraints
- Name: Workspace · Description: Tenant boundary for a specific FinTech or medical billing provider
- Name: ConstraintSet · Description: Group of statutory logic schemas and mathematical proofs defining valid outputs
- Name: Constraint · Description: Individual hard-coded compliance rule or schema requirement
- Name: Anomaly · Description: Logic conflict or hallucination detected that violates a statutory constraint
- Name: Override · Description: Mandatory human compliance decision applied to a halted anomalous workflow
**Relations**:
- To: Resolution · From: Workspace · Label: processes · Cardinality: one-to-many
- To: ConstraintSet · From: Workspace · Label: defines · Cardinality: one-to-many
- To: Resolution · From: ConstraintSet · Label: validates · Cardinality: one-to-many
- To: Constraint · From: ConstraintSet · Label: contains · Cardinality: one-to-many
- To: Anomaly · From: Resolution · Label: detects · Cardinality: one-to-many
- To: Anomaly · From: Constraint · Label: triggers · Cardinality: one-to-many
- To: Override · From: Anomaly · Label: resolved by · Cardinality: one-to-one
**Tenant Anchor**: Workspace
**Primary Resource**: Resolution

## Api Definition

**Protocols**:
- REST
- SDK
- MCP
- Webhooks
**Consumed By**:
- [Medical Claims Agent](/Agents/Medical_Claims_Agent)
- [Trade Execution Agent](/Agents/Trade_Execution_Agent)
- [KYC Verification Agent](/Agents/KYC_Verification_Agent)
- [SEC Reporting Agent](/Agents/SEC_Reporting_Agent)
**Integrations**:
- [OpenAI](/Products/OpenAI)
- [Anthropic](/Products/Anthropic)
- [AWS Bedrock](/Products/AWS_Bedrock)
- [Okta](/Products/Okta)
- [Datadog](/Products/Datadog)
**Consumption Model**: An agent registers the MCP server as a validation tool, routing every LLM generation through the Evaluate Probabilistic Payload call and awaiting a webhook trigger if an anomaly requires human override.
**Workflow Wrappers**:
- Name: Evaluate Probabilistic Payload · Wraps: Evaluates raw text against constraint schemas, halts on anomalies, and signs valid outputs.
- Name: Commit Human Override · Wraps: Applies an override decision, injects the corrected payload, and resumes the halted resolution.
- Name: Publish Constraint Set · Wraps: Validates logic expressions across all constraints, bumps the version, and activates the set.

## Api Function Cascade

**Ai Role**: This headless API primitive runs straight-through constraint evaluations using Code and Generative validation logic, exposing anomalies via webhooks for exception-based human overrides controlled by the consuming application.
**Cascade**:
- Kind: Code · Note: Accepts raw LLM output from the calling agent via MCP. · Step: Ingest Probabilistic Payload · Verb: ingest · Realizes: Receive System Data · Oversight: none
- Kind: Code · Note: Pulls the current versioned constraint schemas. · Step: Fetch Active Constraints · Verb: fetch · Realizes: Retrieve System Configurations · Oversight: none
- Kind: Generative · Note: Uses supported LLMs to check non-deterministic or fuzzy constraints. · Step: Evaluate Semantic Bounds · Verb: inspect · Realizes: Evaluate Information Content · Oversight: none
- Kind: Code · Note: Applies strict schema constraints to the generated payload. · Step: Validate Deterministic Logic · Verb: validate · Realizes: Validate Data Compliance · Oversight: none
- Kind: Code · Note: Emits webhook and halts execution if an anomaly is detected. · Step: Halt And Webhook Exceptions · Verb: route · Realizes: Route Exceptions · Oversight: review-on-exception
- Kind: Code · Note: Receives human decision via API and injects the corrected payload. · Step: Apply Override And Resume · Verb: update · Realizes: Apply Corrections · Oversight: none
- Kind: Code · Note: Cryptographically signs the valid payload and returns it to the agent. · Step: Certify Resolved Output · Verb: certify · Realizes: Authorize Release · Oversight: none
**Optimizes**:
- [Validation Latency](/Metrics/Validation_Latency)
- [First-Pass Yield](/Metrics/First-Pass_Yield)
- [Anomaly Detection Rate](/Metrics/Anomaly_Detection_Rate)
- [Constraint Adherence Rate](/Metrics/Constraint_Adherence_Rate)

## Headless Saas Representative Offer

**Warranty**: 99.99% API uptime for payload evaluation, providing proportional service credits for validation latency exceeding 100ms.
**Price Band**: ~$0.005 to $0.02 per probabilistic payload evaluated, scaling with the complexity of the constraint set.
**Pricing Kind**: UsageMeter
**Deliverables**:
- Cryptographically signed validation payloads
- Anomaly detection webhook events
- Version-controlled constraint schemas
- Human override state management API access
**Delivery Mode**: Clients provision API credentials or register the MCP server instantly, routing generations through the evaluation endpoints on a self-serve, pay-as-you-go basis.
**Business Function**: ProvideService
**Agent Checkout Support**:
- agentic-commerce-protocol
- stored-credential

## Headless Saas Crud Surface

**Auth Model**: mTLS
**Endpoints**:
- GET /resolutions — list resolutions
- POST /resolutions — create a resolution task
- GET /resolutions/{id} — fetch a specific resolution
- PATCH /resolutions/{id} — update a resolution
- POST /resolutions/{id}/evaluate — evaluate the raw payload against constraints
- GET /workspaces — list workspaces
- POST /workspaces — create a new workspace
- GET /workspaces/{id} — fetch a specific workspace
- PATCH /workspaces/{id} — update a workspace
- GET /workspaces/{workspaceId}/constraint-sets — list constraint sets for a workspace
- POST /workspaces/{workspaceId}/constraint-sets — create a new constraint set
- GET /constraint-sets/{id} — fetch a specific constraint set
- PATCH /constraint-sets/{id} — update a constraint set
- GET /constraint-sets/{constraintSetId}/constraints — list constraints within a set
- POST /constraint-sets/{constraintSetId}/constraints — create a new constraint
- GET /constraints/{id} — fetch a specific constraint
- PATCH /constraints/{id} — update a constraint
- GET /resolutions/{resolutionId}/anomalies — list anomalies detected in a resolution
- POST /resolutions/{resolutionId}/anomalies — log an anomaly
- GET /anomalies/{id} — fetch a specific anomaly
- PATCH /anomalies/{id} — update an anomaly
- GET /anomalies/{anomalyId}/overrides — list overrides applied to an anomaly
- POST /anomalies/{anomalyId}/overrides — apply a human compliance override
- GET /overrides/{id} — fetch a specific override
- PATCH /overrides/{id} — update an override justification
**Multitenancy**: Schema-per-tenant
**Webhook Events**:
- resolution.evaluated
- resolution.anomalous
- resolution.verified
- anomaly.detected
- override.applied

## Headless Saas Erd

```mermaid
erDiagram
    Resolution {
        UUID id PK
        UUID workspaceId FK
        UUID constraintSetId FK
        TEXT rawPayload
        JSONB verifiedPayload
        VARCHAR cryptographicSignature
        VARCHAR status
    }
    Workspace {
        UUID id PK "TenantKey"
        VARCHAR name
        VARCHAR industry
        TIMESTAMP createdAt
    }
    ConstraintSet {
        UUID id PK
        UUID workspaceId FK
        VARCHAR name
        VARCHAR version
        BOOLEAN isActive
    }
    Constraint {
        UUID id PK
        UUID constraintSetId FK
        VARCHAR ruleType
        VARCHAR logicExpression
        VARCHAR errorMessage
    }
    Anomaly {
        UUID id PK
        UUID resolutionId FK
        UUID constraintId FK
        VARCHAR detectedValue
        VARCHAR severity
    }
    Override {
        UUID id PK
        UUID anomalyId FK
        VARCHAR reviewerId
        VARCHAR decision
        JSONB correctedPayload
        TEXT justification
    }
    Workspace ||--o{ Resolution : processes
    Workspace ||--o{ ConstraintSet : defines
    ConstraintSet ||--o{ Resolution : validates
    ConstraintSet ||--o{ Constraint : contains
    Resolution ||--o{ Anomaly : detects
    Constraint ||--o{ Anomaly : triggers
    Anomaly ||--|| Override : "resolved by"
```

## Neighborhood

### Composed into

- [Discrepancyridge](/Startups/Discrepancyridge) — composes · Startups

### Optimizes

- [Constraint Adherence Rate](/Metrics/Constraint_Adherence_Rate) — optimizes · Metrics
- [First-Pass Yield](/Metrics/First-Pass_Yield) — optimizes · Metrics
- [Validation Latency](/Metrics/Validation_Latency) — optimizes · Metrics
- [Anomaly Detection Rate](/Metrics/Anomaly_Detection_Rate) — optimizes · Metrics

### What it uses

- [AWS Bedrock](/Products/AWS_Bedrock) — uses · Products
- [Anthropic](/Products/Anthropic) — uses · Products
- [OpenAI](/Products/OpenAI) — uses · Products
- [Datadog](/Software/Datadog) — uses · Software
- [Okta](/Software/Okta) — uses · Software

### Who consumes this

- [KYC Verification Agent](/Agents/KYC_Verification_Agent) — consumed by · Agents
- [Medical Claims Agent](/Agents/Medical_Claims_Agent) — consumed by · Agents
- [SEC Reporting Agent](/Agents/SEC_Reporting_Agent) — consumed by · Agents
- [Trade Execution Agent](/Agents/Trade_Execution_Agent) — consumed by · Agents

### Similar Software

- [RegTech Solutions](/Metrics/Compliance_Review_Cycle_Time/Software/RegTech_Solutions) — similar · Software
- [Requirement Tracking Software](/Metrics/Mission_Development_Cycle_Time/Software/Requirement_Tracking_Software) — similar · Software
- [Fintech Core Ledgers](/Resources/Regulatory_code_databases/Software/Fintech_Core_Ledgers) — similar · Software
- [Compliance Validation Engine](/Software/Compliance_Validation_Engine) — similar · Software
- [Cryptographic Ledger Engine](/Software/Cryptographic_Ledger_Engine) — similar · Software
- [Compliance Management Software](/Metrics/Time_To_Sign-Off/Software/Compliance_Management_Software) — similar · Software
- [Vendor Resolution Engine](/Software/Vendor_Resolution_Engine) — similar · Software
- [Sanctions Screening API](/Software/Sanctions_Screening_API) — similar · Software
- [Identity Management Systems](/Metrics/Policy_Compliance_Rate/Software/Identity_Management_Systems) — similar · Software
- [GRC Platforms](/Metrics/Compliance_Review_Cycle_Time/Software/GRC_Platforms) — similar · Software
- [Invoice Validation](/Software/Invoice_Validation) — similar · Software
- [Requirements Management Systems](/Metrics/Requirements_Traceability_Index/Software/Requirements_Management_Systems) — similar · Software
- [Contract Rules Engine](/Software/Contract_Rules_Engine) — similar · Software
- [Cryptographic Validation Engine](/Software/Cryptographic_Validation_Engine) — similar · Software

### Similar Agents

- [Deterministic Flagging Agent](/Agents/Deterministic_Flagging_Agent) — similar · Agents

### Similar Startups

- [Difficultylane](/Startups/Difficultylane) — similar · Startups
- [Farosses](/Startups/Farosses) — similar · Startups
- [Bankient](/Startups/Bankient) — similar · Startups
- [Vellumworks](/Startups/Vellumworks) — similar · Startups
- [Abought](/Problems/State_Machine_Generation/Startups/Abought) — similar · Startups