IMPORTANT
AI Assist Note (Knowledge Heritage): This document is part of the "Sovereign Reality" documentation.
- @docs ARCHITECTURE:Documentation
- Failure Path: Information drift, legacy terminology, or documentation mismatch.
- Telemetry Link: Cross-reference with
execution/parity_guard.pyresults.
AI Assist Note
Automated governance and architectural tracking.
🔍 Debugging & Observability
Traceability via parity_guard.py.
📄 API Contract: Tadpole OS Engine
@docs ARCHITECTURE:ApiContract
This is the minimalist REST contract for the Tadpole OS Engine. For high-level architecture, see the Architecture Overview.
⚡ Quick Start (3 Steps)
- Set Token:
export NEURAL_TOKEN=your_secure_token - Verify Engine:
curl -H "Authorization: Bearer $NEURAL_TOKEN" http://localhost:8000/v1/engine/health - List Agents:
curl -H "Authorization: Bearer $NEURAL_TOKEN" http://localhost:8000/v1/agents
🛰️ Standard Headers
Every response from the engine includes these observability headers:
| Header | Description |
|---|---|
X-Request-Id | Unique UUID for end-to-end request tracing. |
X-RateLimit-Limit | Maximum requests allowed per 60s window. |
X-RateLimit-Remaining | Remaining requests in the current window. |
🔌 REST Endpoints (v1)
Health & Control
| Method | Path | Description |
|---|---|---|
GET | /v1/engine/health | Engine health & uptime pulse. |
POST | /v1/engine/deploy | Trigger external deployment logic. |
POST | /v1/engine/kill | Emergency shutdown of all active agent tasks. |
POST | /v1/engine/speak | Text-to-Speech (TTS) synthesis. |
Agents & Swarms
| Method | Path | Description |
|---|---|---|
GET | /v1/agents | List all registered agents (Paginated). |
POST | /v1/agents | Register a new agent node. |
GET | /v1/agents/{id} | Retrieve a specific agent's schematic. |
PUT | /v1/agents/{id} | Update agent configuration. |
POST | /v1/agents/{id}/tasks | Execute: Dispatch a task to the agent runner. |
DELETE | /v1/agents/{id} | De-register an agent node. |
Oversight & Governance
| Method | Path | Description |
|---|---|---|
GET | /v1/oversight/pending | List all actions awaiting human approval. |
POST | /v1/oversight/{id}/decide | Approve or Reject a tool execution. |
GET | /v1/oversight/ledger | Review historical governance decisions. |
GET | /v1/oversight/security/integrity | Verify the Merkle Audit Chain. |
📄 Pagination
All list endpoints support cursor-free pagination:
| Parameter | Default | Max | Description |
|---|---|---|---|
page | 1 | — | 1-indexed page number. |
per_page | 25 | 100 | Items per page (Clamped 1-100). |
Response Envelope:
json
{
"data": [...],
"total": 42,
"_links": {
"self": { "href": "/v1/agents?page=1", "method": "GET" },
"next": { "href": "/v1/agents?page=2", "method": "GET" }
}
}⚠️ Error Protocol (RFC 9457)
Errors return application/problem+json for machine-readable remediation.
| Status | Code | Meaning |
|---|---|---|
401 | UNAUTHORIZED | Invalid or missing NEURAL_TOKEN. |
429 | RATE_LIMITED | Engine-level token bucket exhausted. |
507 | OOM_QUANT | Provider VRAM OOM; suggestive of quantization fallback. |
TIP
Versioning Strategy: All business logic is nested under /v1/. Root-level paths are legacy and will be deprecated. Always use the /v1 prefix for new integrations.
