agentantibody docs

Enable reasoning

Point agentantibody at a model endpoint so it can investigate graph changes and write findings.

Reasoning is off until you set a model endpoint. Without one, the graph and telemetry still materialize and the explorer works, but no findings are produced and there is nothing to remediate.

Set the endpoint

export AGENTANTIBODY_MODEL_ENDPOINT='<your OpenAI-compatible endpoint>'
make run

The service sends requests to <endpoint>/v1/responses using the OpenAI Responses API shape. It sends no credentials of its own, so point it at a gateway or proxy that attaches them.

Choose a model

FlagEnvironment variableDefaultMeaning
--model-endpointAGENTANTIBODY_MODEL_ENDPOINTemptyEndpoint to call. Empty turns reasoning off.
--modelAGENTANTIBODY_MODELgpt-6-astraModel name sent with each request.
--reasoning-effortAGENTANTIBODY_REASONING_EFFORTemptyReasoning effort passed through to the Responses API.
--reasoning-retrievalAGENTANTIBODY_RETRIEVAL_MODEmodel_selectedHow the investigator gets graph context.
--reasoning-intervalnone1sHow often the service polls for graph changes.
--reasoning-consumernonereasoning/defaultStable ID of the durable change-feed consumer.

Model quality decides how much an investigator finds. Run the evals against the model you plan to use. make eval-matrix compares several.

Retrieval modes

  • model_selected lets the model choose which of the six read-only tools to call, and with what arguments.
  • deterministic has the service fetch the baseline, timeline, a default graph path, and governance posture first. The model can still make more calls.

What happens next

Each graph change that matters to reasoning is written to a transactional outbox. A detector checks whether it is new. Only changes with a signal reach the model. A model response becomes a finding only if it passes validation. See Reasoning.

Findings appear in the explorer with their evidence. Fixes you can apply appear next to them. See Remediation.