CLARIXO Behavior Evidence Export API

Track what happened

Capture the event identity, runtime path, model, timestamps, and user-facing summaries needed to reconstruct an AI event.

Preserve record integrity

Include a stable evidence hash generated from normalized core fields so exported records are more than simple log dumps.

Support review and retention

Use evidence records for internal accountability, customer review, post-incident reconstruction, and audit-oriented export workflows.

Not a routing API

It does not decide provider selection, fallback order, or execution-path control between enterprises and external AI providers.

Not an execution-control API

It does not sit inline in the runtime path and does not act as the control plane for model execution decisions.

Not an approval or enforcement API

It does not approve, deny, block, or escalate execution. Its role is evidence export and downstream audit support.

CLARIXO Behavior Evidence Export API Integration Protocol

Defines the formal evidence export contract, including role, scope, boundary, minimum evidence record, and integrity rule. Use this document when you need to understand what this API is, what it is not, how evidence_hash is defined, and why this API remains separate from CLARIXO Core Runtime API.

Contract, Role, and Boundary

CLARIXO Behavior Evidence Export API Integration Manual

Explains how to implement the minimum evidence export path in a real system. Use this document when you need to define an evidence event, map fields, normalize a payload, generate evidence_hash, submit records, retrieve records, and verify that the exported result is stable, readable, and operationally useful.

Primary entry for implementation, verification, and rollout

POST /v1/evidence/events

Write one or more evidence events after a user-facing AI behavior has been completed and normalized into an evidence payload.

GET /v1/evidence/events

Query or export evidence events by project_id, session_id, event_id, final_status, record_origin, application_id, route_path, recorded_from, recorded_to, and limit.

Complete the user-facing AI behavior

Let the enterprise system finish the user-visible response or result before exporting an evidence record.

Normalize into an evidence payload

Map the event into stable fields such as project scope, summaries, provider, route path, final status, and metadata.

Generate evidence_hash

Generate a SHA-256 evidence hash from normalized core fields for basic record-integrity verification.

Send to CLARIXO

Write the evidence event to CLARIXO for later query, export, dispute review, audit, and post-incident reconstruction.

Customer dispute review

Reconstruct what the AI system showed or returned to a user when support, trust, or escalation teams need a stable event record.

Internal accountability and audit export

Retain normalized evidence events that can be queried or exported during internal review, compliance preparation, or post-incident analysis.

Post-incident reconstruction

Preserve the route path, summaries, final status, and integrity marker needed to reconstruct what happened after a user-facing AI event.

Low-friction first integration

Start with evidence export when a team is not yet ready to move runtime control, routing, or approval behavior into a deeper middleware layer.

Teams shipping user-facing AI features

Useful for teams that need traceable records of what users experienced without redesigning the execution stack first.

Teams handling review and dispute workflows

Useful for support, trust, operations, and escalation teams that need stable evidence when user-visible AI behavior is questioned later.

Teams moving toward stronger AI governance

Useful for organizations that want to start with evidence export now, then add runtime control, approval, or execution governance later.

Minimal exported record

{
  "schema_version": "v1",
  "project_id": "trial_proj_b17e8105174f862a",
  "application_id": "tgtracing:assistant:customer_chat",
  "event_id": "evt_20260408_0001",
  "session_id": "sess_7c1f3b9d",
  "recorded_at": "2026-04-08T17:58:11Z",
  "user_ref": "user_anon_1024",
  "input_summary": "User asked whether a shared tracking link could reveal location without consent.",
  "output_summary": "Assistant explained that consent-based interaction is required and did not claim passive tracking.",
  "provider": "enterprise_app",
  "model": "gpt-4.1",
  "route_path": "direct",
  "fallback_used": false,
  "guard_flags": ["policy_boundary_explained"],
  "confidence_signals": {
    "answer_mode": "policy-aligned",
    "uncertainty_level": "low"
  },
  "operator_action": "none",
  "final_status": "delivered",
  "metadata": {
    "locale": "en",
    "record_origin": "manual_post"
  },
  "evidence_hash": "sha256:examplehexvalue"
}

Why this record remains reviewable

event_id uniquely identifies the exported evidence event within the evidence stream.

input_summary and output_summary preserve the user-facing behavior in a reviewable form without requiring a full raw transcript in the minimum version.

route_path, fallback_used, and guard_flags preserve the normalized path and behavior context needed for later reconstruction.

final_status captures the final externally relevant outcome that the user effectively received.

evidence_hash provides a normalized integrity marker so the record can be checked as evidence rather than treated as a simple log entry.