For AI agents: a documentation index is available at the root level at /llms.txt and /llms-full.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.
Copy to LLMGithubGo to App
DocumentationIntegrationsAgent OptimizationSelf-hosting OpikSDK & API referenceOpik University
DocumentationIntegrationsAgent OptimizationSelf-hosting OpikSDK & API referenceOpik University
  • Getting Started
    • Opik Agent Optimizer
    • Optimization Studio
    • Quickstart
    • Quickstart notebook
    • FAQ
    • Changelog
    • Known Issues
  • Optimization
    • Concepts
    • Configure LLM Providers
    • Define datasets
    • Define metrics
    • Optimize prompts
    • Optimize tools (MCP)
    • Optimize agents
    • Optimize multimodal
    • Dashboard results
  • Optimization Algorithms
    • Overview
    • Benchmarks
    • MetaPrompt
    • HRPO
    • Few-Shot Bayesian
    • Evolutionary
    • GEPA
    • Parameter
    • Tool Optimization
  • Cookbooks & Tutorials
    • Optimizer introduction
    • Synthetic data optimizer
    • ARC-AGI tutorial
    • Multimodal agent tutorial
  • Advanced Topics
    • Extending optimizers
    • Custom metrics
    • Custom optimizer prompts
    • Sampling controls
    • Multiple completions (n)
    • Chaining optimizers
    • API Reference
LogoLogo
Copy to LLMGithubGo to App
On this page
  • Navigate to your run
  • Key panels
  • Reuse results
  • Next steps
Optimization

Dashboard results

Was this page helpful?
Previous

Optimization algorithms overview

Next
Built with

After each optimization run, visit the Opik dashboard to understand what changed and decide whether to ship the new prompt.

Navigate to your run

  1. Open https://www.comet.com/opik.
  2. In the left nav, click Optimization runs under Evaluation.
  3. Select the run you care about (grouped by dataset + optimizer). The detail view shows charts, trials, prompts, and per-sample traces.

Key panels

Optimization progress chart

Plots every trial score in chronological order and highlights the current best prompt. Hover to read exact values and see percentage improvements.

Trial table

Lists each trial, the optimizer used, the prompt JSON, and per-trial scores. Click a trial row to expand dataset items and attached traces.

Examples & traces

When you expand a trial, you can inspect every dataset item that ran during that trial plus the corresponding trace tree (tool calls, attachments, etc.).

Failure modes (reflective runs only)

HRPO runs add a panel that clusters similar failures. Expand a cluster to read metric reasons and sample traces.

Dataset coverage

Confirms how many dataset rows were sampled per trial so you can judge statistical significance.

Reuse results

While the UI currently focuses on analysis, you can always pull prompts and history directly from the SDK after the run finishes:

1# result is the OptimizationResult returned by your optimizer call
2optimized_prompt = result.prompt
3history = result.history

Use optimized_prompt to update your application and history to build custom reports or attach evidence to pull requests.

Next steps

  • Feed the exported prompt back into your application.
  • Attach dashboards or screenshots to your PR so reviewers understand the improvement.
  • Use Optimization Studio for UI-driven runs and comparisons.