# aevyra

## Docs

- [Adapters & on-ramps](https://docs.aevyra.ai/origin/adapters.md): Three ways to get a trace into Origin — live pipeline, existing logs, or raw AgentTrace.
- [Attribution API](https://docs.aevyra.ai/origin/api/attribution.md): Attribution, NodeAttribution, and PromptAttribution reference.
- [diagnose_pipeline & Origin](https://docs.aevyra.ai/origin/api/diagnose.md): API reference for diagnose_pipeline(), Origin, and judge_from_verdict().
- [CLI reference](https://docs.aevyra.ai/origin/cli.md): All commands and flags for the aevyra-origin CLI.
- [Introduction](https://docs.aevyra.ai/origin/introduction.md): Failure attribution for agent pipelines — find where it went wrong and what kind of fix it needs.
- [Attribution methods](https://docs.aevyra.ai/origin/methods.md): Critic, decomposition, and ablation — how each works and when to use it.
- [Providers & Models](https://docs.aevyra.ai/origin/providers.md): Choose which LLM powers Origin's attribution engine.
- [Quick start](https://docs.aevyra.ai/origin/quickstart.md): Diagnose your first pipeline failure in under 5 minutes.
- [Tutorial: Bring your own trace](https://docs.aevyra.ai/origin/tutorial-bring-your-own-trace.md): You already have logs from Langfuse, OpenTelemetry, or your own observability stack. Convert them once and run Origin against your existing traces — no instrumentation required.
- [Tutorial: Coding agent](https://docs.aevyra.ai/origin/tutorial-coding-agent.md): A coding agent ships broken code twice in a row. Origin traces the failure back to a single planner prompt — and the coder's first draft was actually correct.
- [Tutorial: Support triage agent](https://docs.aevyra.ai/origin/tutorial-support-triage.md): A customer-support agent denies a valid refund. Origin finds the culprit span, classifies the fix type, and tells you exactly what to change.
- [Analysis output](https://docs.aevyra.ai/reflex/analysis.md): Understanding the optimization report — trajectory, strategy breakdown, prompt diff, and before/after examples.
- [PromptOptimizer](https://docs.aevyra.ai/reflex/api/optimizer.md): API reference for PromptOptimizer, OptimizerConfig, and strategy registration.
- [OptimizationResult](https://docs.aevyra.ai/reflex/api/result.md): API reference for OptimizationResult and supporting dataclasses.
- [CLI reference](https://docs.aevyra.ai/reflex/cli.md): All commands and flags for the aevyra-reflex CLI.
- [Configuration](https://docs.aevyra.ai/reflex/configuration.md): Tuning iterations, thresholds, parallelism, and strategy-specific parameters.
- [Dashboard](https://docs.aevyra.ai/reflex/dashboard.md): The local web UI for browsing runs, inspecting iteration traces, reading prompt diffs, and branching new experiments from any iteration.
- [Explainability & auditability](https://docs.aevyra.ai/reflex/explainability.md): Every decision reflex makes is recorded, reasoned, and queryable — so you can understand, audit, and reproduce any optimization run.
- [Integrations](https://docs.aevyra.ai/reflex/integrations.md): Log reflex runs to MLflow, Weights & Biases, or any custom experiment tracker.
- [Introduction](https://docs.aevyra.ai/reflex/introduction.md): Agentic prompt optimization built for production workloads — crash-safe, resumable, and fully auditable.
- [Optimization modes](https://docs.aevyra.ai/reflex/modes.md): Standard mode for dataset-driven optimization and pipeline mode for agentic systems — when to use each and how they work.
- [Providers & Models](https://docs.aevyra.ai/reflex/providers.md): Choose which models power your optimization runs — reasoning agent and target model.
- [Quick start](https://docs.aevyra.ai/reflex/quickstart.md): Optimize your first prompt in under 5 minutes. One install, one command.
- [Strategies](https://docs.aevyra.ai/reflex/strategies.md): How each optimization axis works and when to use it.
- [Tutorial: optimizing a tool-calling developer assistant](https://docs.aevyra.ai/reflex/tutorial-dev-assistant.md): Teach an agent to reliably call search_docs, calculate, and get_date before answering — a complete pipeline-mode walkthrough with real run logs.
- [Tutorial: format-compliance optimization](https://docs.aevyra.ai/reflex/tutorial-security-incidents.md): Teach a model to produce strict 3-sentence executive briefs from security incident reports — a complete walkthrough with real run logs.
- [Dataset](https://docs.aevyra.ai/verdict/api/dataset.md): API reference for the Dataset and Conversation classes.
- [EvalResults](https://docs.aevyra.ai/verdict/api/results.md): API reference for EvalResults and ModelResult.
- [EvalRunner](https://docs.aevyra.ai/verdict/api/runner.md): API reference for EvalRunner and RunConfig.
- [CLI reference](https://docs.aevyra.ai/verdict/cli.md): All commands and flags for the aevyra-verdict CLI.
- [Configuration](https://docs.aevyra.ai/verdict/configuration.md): Tuning concurrency, retries, and rate-limit handling.
- [Dataset](https://docs.aevyra.ai/verdict/dataset.md): Loading and working with eval datasets.
- [Introduction](https://docs.aevyra.ai/verdict/introduction.md): A framework for evaluating and comparing LLM outputs across models and providers.
- [Metrics](https://docs.aevyra.ai/verdict/metrics.md): Scoring model responses with built-in and custom metrics.
- [Providers](https://docs.aevyra.ai/verdict/providers.md): Configuring models and API providers.
- [Quick start](https://docs.aevyra.ai/verdict/quickstart.md): Run your first eval in under 5 minutes.
- [Adapters](https://docs.aevyra.ai/witness/adapters.md): Import existing agent logs into AgentTrace — without changing any agent code.
- [Trace schema](https://docs.aevyra.ai/witness/api/trace.md): AgentTrace and TraceNode field reference.
- [Interceptors](https://docs.aevyra.ai/witness/interceptors.md): Capture live agent traffic as AgentTrace spans — no decorators required.
- [Introduction](https://docs.aevyra.ai/witness/introduction.md): The shared agent trace primitive for the Aevyra stack — capture what happened, in any framework, from any language.
- [Quick start](https://docs.aevyra.ai/witness/quickstart.md): Instrument a pipeline and get your first AgentTrace in under 2 minutes.

## OpenAPI Specs

- [openapi](https://docs.aevyra.ai/api-reference/openapi.json)