Integrator API Reference
The Integrator API provides programmatic access to the AgentV engine, enabling seamless integration with CI/CD pipelines, custom dashboards, and automated forensic workflows.
🔐 Authentication
Section titled “🔐 Authentication”All Integrator API requests must include the X-Api-Key header.
X-Api-Key: {SERVICE_API_KEY}🚀 Priority Roadmap Endpoints (/api/v1)
Section titled “🚀 Priority Roadmap Endpoints (/api/v1)”POST /api/v1/mutate
Section titled “POST /api/v1/mutate”Priority: P0 Programmatic scenario mutation for variance and robustness testing.
- Body:
type(string): Mutation type (e.g.,typo,jumble,synonym).path(string): Path to a scenario JSON file.raw_json(object, optional): Raw scenario JSON to mutate directly.
- Support: Accepts Raw Content for cloud-native integration.
GET /api/v1/metrics
Section titled “GET /api/v1/metrics”Priority: P0 Lists all registered evaluation metrics available for dynamic hydration.
GET /api/v1/taxonomy
Section titled “GET /api/v1/taxonomy”Priority: P0 Returns the failure taxonomy for categorical diagnostics.
POST /api/v1/spec-to-eval
Section titled “POST /api/v1/spec-to-eval”Priority: P1 Converts structured Markdown PRDs into validated scenario JSON stubs.
- Body:
markdown(string): Raw Markdown content.input_path(string): Path to a Markdown file.
- Support: Accepts Raw Content.
GET /api/v1/doctor
Section titled “GET /api/v1/doctor”Priority: P1 Environmental health audit. Checks project readiness, plugin loading, and catalog integrity.
GET /api/v1/explain/<run_id>
Section titled “GET /api/v1/explain/<run_id>”Priority: P2 Forensic Root Cause Analysis (RCA). Returns the primary failure trigger and causal chain.
🛠️ Troubleshooting: Spec-to-Eval
Section titled “🛠️ Troubleshooting: Spec-to-Eval”The spec-to-eval service uses a combination of hierarchical section splitting and LLM tasks synthesis. If parsing fails, verify the following Markdown requirements:
1. Structure Requirements
Section titled “1. Structure Requirements”- Title: Must start with an H1 header (
# PRD: ...). - Sections: Use H2 headers (
## Tasks) to demarcate logic blocks.
2. Common Parsing Pitfalls
Section titled “2. Common Parsing Pitfalls”- Missing Tasks: If no
## Tasksor## Test Casessection is found, the engine will attempt LLM synthesis. Ensure this section contains either a bulleted list or H3 sub-headers. - Malformed Metadata: Ensure
**Industry:**and**Use Case:**segments are correctly labeled in the Overview. - Topology Failures: Agent topology requires the pattern
**AgentName:** writes to [X], reads from [Y].
3. LLM Fallback Triggers
Section titled “3. LLM Fallback Triggers”If the heuristic parser cannot identify structured nodes, it triggers Gemini-powered Synthesis. Ensure GOOGLE_API_KEY is configured in your environment for high-fidelity fallback.
📂 Core Management APIs
Section titled “📂 Core Management APIs”GET /api/scenarios
Section titled “GET /api/scenarios”Faceted search across the global scenario catalog.
GET /api/info
Section titled “GET /api/info”Consolidated system health, engine versioning, and telemetry.
GET /api/v1/verify/<run_id> (Public)
Section titled “GET /api/v1/verify/<run_id> (Public)”Public integrity audit for signed traces.