Explorer
Syncing—

Model-based Prompt Injection Detection in Text - language-forensics APIService

model-based-prompt-injection-detection

Model-based prompt-injection and jailbreak detection for text (a trained classifier, not a rule-based pattern scanner), served through the language-forensics /v1/detect API. The expected endpoints are POST /v1/detect and GET /health; suppliers MAY expose more (GET /v1/models, Lakera Guard and Azure Prompt Shields compatibility layers) but consumers MUST NOT depend on it. Responses are injection probabilities, not verdicts. This service is model-agnostic: suppliers choose their own classifier, so scores, calibration, and detection quality vary and are not guaranteed by the protocol; each response carries a pipeline_fingerprint identifying the model and window policy.

Active Suppliers
1
of 1 ever configured
Service Info
Service ID
model-based-prompt-injection-detection
Compute Units / Relay
15,000
Owner
pokt1r…2jkc
Active Applications
1
Relay Mining EMA
1 relays
Service Metadata Cardsource: on-chain metadata (LCD)
{8 items
"schema":"pocket-service-card/v1"
"description":"Model-based prompt-injection and jailbreak detection for text (a trained classifier, not a rule-based pattern scanner), served through the language-forensics /v..."
"rpc_types":[1 item
0:{4 items
"type":"REST"
"intent":"expected"
"backend_hint":"language-forensics gateway on :8080 (language-forensics-service); mount at /"
"notes":"Only POST /v1/detect and GET /health are expected. Request body is JSON with exactly one of text or inputs[{text, role?, id?}], plus an optional threshold. Long..."
}
]
"apis":[1 item
0:"language-forensics-detect"
]
"specs":[2 items
0:{3 items
"kind":"markdown"
"url":"https://raw.githubusercontent.com/pnyxai/pocket-network-services/6f3912ee520cb75a98288aaf333d6ba1ef729788/mainnet/docs_model-based-prompt-injection-detection.md"
"sha256":"538e4ae5776b7f3a649e63726a3923f5ba1a18bab68824b792ca8131aa90aafc"
}
1:{3 items
"kind":"openapi"
"url":"https://raw.githubusercontent.com/RawthiL/language-forensics-service/9d42a70d0355532a1a3bd70fc20b8c0d2d685a0b/docs/openapi.json"
"sha256":"0650937afab6147f97422a7265e350391627e938e7590dff4ac630d87cacadff"
}
]
"access":"public"
"results":"variable"
"serving":{3 items
"backend":"Any server exposing POST /v1/detect and GET /health with the request/response shape of the OpenAPI document in specs[1]. Recommended: language-forensics-service..."
"healthcheck":[1 item
0:{3 items
"rpc_type":"REST"
"request":{2 items
"path":"/health"
"method":"GET"
}
"expect":{2 items
"json_path":"$.status"
"matches":"^healthy$"
}
}
]
"notes":"Use a server with dynamic batching so one backend serves many concurrent requests. Recommended model: Horizon-Labs prompt-injection-guard base (Apache-2.0). Sup..."
}
}