API reference¶
The server (agentic_security server) has test endpoints. Use them to test a spec before you scan a real target.
POST /v1/self-probe¶
A mock LLM for text. It refuses approximately 20% of prompts at random.
Request body:
The response has the OpenAI chat-completion format.
POST /v1/self-probe-image¶
A mock LLM for images. It ignores the request body and always gives the same response.
POST /v1/self-probe-file¶
A mock audio transcription endpoint. It accepts a multipart upload.
| Part | Description |
|---|---|
file |
An .m4a, .mp3, or .wav file. |
model |
Optional. The default is whisper-large-v3. |
The request must have the header Authorization: Bearer <any value>. The endpoint does not examine the key. The response is {"text": "...", "model": "..."}.
Other endpoints¶
| Method | Path | Description |
|---|---|---|
| GET | /health |
Returns {"status": "ok"}. |
| GET | /v1/llm-specs |
Returns the built-in spec templates. |
| GET | /v1/data-config |
Returns the dataset registry. The response is very large. |
| POST | /verify |
Sends a test request with a spec. |
| POST | /scan |
Starts a scan. The response is a stream of JSON lines. |
| POST | /stop |
Stops the current scan. |
| POST | /scan-csv |
Starts a scan with an uploaded CSV file. |
| GET | /failures |
Returns the failures of the last scan. |