Core Concepts
Provon turns evidence from real agent runs into reviewable artifacts that support repair and evaluation.
The Improvement Loop#
The repair workflow is the core loop: evidence → Findings → repair handoff → verification against the next traces.
Every derived artifact keeps a path back to the evidence that motivated it. This makes improvements reviewable instead of treating model output or automated diagnosis as unquestioned truth.
Traces#
A Trace records one causally connected execution. For an agent, that can include model calls, tool calls, retrieval, handoffs, latency, token usage, cost, errors, and the final response.
Multiple traces can belong to one conversation. Stable conversation identity lets Provon evaluate a multi-turn outcome instead of judging isolated requests.
See Tracing and the Trace model.
Findings#
A Finding is a supported, actionable problem distilled from one or more conversation occurrences. It combines a cause assessment, severity, confidence, remediation direction, and evidence references.
Findings are intended for review and repair. They can be confirmed, dismissed, resolved, or handed off to an issue tracker.
See Findings.
AI Gateway#
The AI Gateway is an OpenAI-compatible model boundary. It resolves model targets, applies routing and governance policy, enforces request/response guardrails in Monitor or Enforce mode, records provider attempts, and emits trace evidence. Guardrails protect the model-call boundary; passing a guardrail is not proof that the agent completed the user's goal, and application authorization and tool execution remain your responsibility.
The Gateway is optional. Applications can continue calling providers directly and export OpenTelemetry instead.
See AI Gateway.
Success Check#
After your first setup, confirm that:
- Traces shows a causally ordered execution for one agent run;
- one conversation groups related multi-turn traces;
- Findings review states and evidence links resolve back to the originating traces;
- Gateway traffic, when used, records provider attempts alongside the spans.
Troubleshooting#
Conceptual mismatches usually map to one of the concrete failure stages below:
- No trace reaches the project: capture path guidance and Troubleshooting.
- Traces exist but produce no Findings: repair loop and Troubleshooting.
- Guardrail mode confusion: modes are
off,monitor, andenforce; see AI Gateway.
Next Steps#
- Capture evidence: Choose a capture path.
- Walk the end-to-end flow: Quickstart.
- Turn Findings into repairs: Repair loop.