:: Integrity API for Semantic Integrity and Meaning Preservation

HOME MEANINGSYSTEM
KEYS 🤖


: Integrity API

When AI systems and agents change information, especially across multi-hop steps, how do we know original intent has survived the journey?


For when maintaining meaning must remain traceable from origin → outcome

: The Problem

Modern AI systems constantly rewrite, summarise, and restructure information. But most tools cannot detect when a transformation has altered intrinsic meaning. Such transformations happen constantly:
  • Summarising documents.
  • Translating text.
  • Restructuring data.
  • Chaining agent operations.

Small changes can quietly distort original meaning:

As systems become more automated — especially with multi-agent workflows — these small shifts accumulate. Eventually the result may no longer reflect original intent or fall outside acceptable tolerance. Without verification, there is no reliable way to detect this.


➤ Recent studies by Microsoft [ LLMs Corrupt Your Documents When You Delegate ] by authors: Philippe Laban, Tobias Schnabel, Jennifer Neville] found that integrity degrades across runs, even for short assemble-then-disassemble requests.


No one needs their agents acting like some game of ‘Chinese whispers’ as they step through multi-step, multi-tool pathways. So this Maenen "Integrity" tool could be viewed almost like an electrician's circuit tester – measuring intent preservation through an AI/agent pipeline 'circuit wire'.

: The Integrity Solution

The Integrity API provides deterministic meaning shift analysis along pipelines. It compares the original and final content across several dimensions and produces a structured integrity result. It analyses the relationship between two pieces of information, and the output classifies the transformation into four states:
  • Preserved – meaning intact
  • Minor tension – slight variation
  • Semantic drift – meaning partially changed
  • Semantic break – meaning lost

This allows you and your agent to verify transformations instead of trusting end results blindly.

Integrity API can evaluate the chain and verify whether the final result still reflects the user’s original request. This allows systems to:

  • Detect meaning meandering across long chains.
  • Monitor agent workflows.
  • Audit pipeline integrity.
  • Verify transformations before results are accepted.

: Why Integrity Matters

Integrity checking becomes essential as systems become more autonomous and automated. The API allows developers and agents to:
  • Verify that a transformation preserved intentions.
  • Detect 'story drift' across complex workflows.
  • Audit multi-agent processes.
  • Track transformation quality over time.
  • Anchor results to a ledger for later review.

This makes automated systems and agent workflows trustworthy at scale.

: Not Drift Detection

Integrity analysis is not the same as drift detection. Drift typically measures behavioural deviation over time.
The Integrity API instead verifies whether transformation preserved the original meaning, even across complex multi-step pipelines. It is designed for situations such as:
  • Multi-agent workflows.
  • Document transformations.
  • Translation chains.
  • Summarisation pipelines.
  • Maenen Ledger branch execution.

: Compare Integrity API

embedding to token transport api
Technique Smaller Reusable Deterministic Auditable
 
Raw Prompting ✅ ❌ ❌ ❌
Summarise ✅ ❌ ❌ ❌
Heuristic Structuring ⚠️ ⚠️ ⚠️ ⚠️
WaveInform™ ✅ ✅ ✅ ✅

: Example Use Cases

AI summarisation
 Verify that a summary still reflects the original document. Also see: This≠That API
Translation pipelines
 Detect when meaning shifts during translation.
Agent workflows
 Monitor whether a chain of agents preserved the user's original request.
Plan analysis
 Evaluate the integrity of a multi-step workflow across many operations.

: Ledger or Agent Workflow Example

➤ Integrity tool is especially useful in multi-hop agent pipelines such as ledger branches:


User request
➤➤➤ reference measure
↓
Agent A – research
↓
Agent B – summarise
↓
Agent C – generate proposal
↓
Agent D – final output
↓
➤➤➤ integrity measure


Endpoints

POST /v1/integrity/analyze

Plan Integrity is internal and has no public standalone endpoint.


: Which API?

Integrity API
Verify that meaning survived the workflow.
Value API
Measure whether the result improved the situation.
This≠That API
Compare meaning between documents.

: Support

For technical issues or questions, [ contact support ] via the site. Note that during the beta trial, full support and any SLA is not offered, though your query or comment will be carefully considered. Please include:

  • A short description of the issue.
  • The receipt from the relevant request.
Receipts significantly reduce back-and-forth and speed up diagnosis.


: Beta Notes

This API is currently in limited beta.

  • Supported modes and limits may evolve.
  • No uptime or performance guarantees are implied.
  • Breaking changes will be communicated ahead of time where possible.

: Versioning

This documentation applies to:


API version: v0.1 (beta)

FOR-THE-DREAM-IS-WITH-INTEGRITY-FOR-THE-FREEDOM-OF-INTELLIGENCE