> For the complete documentation index, see [llms.txt](https://docs.getdot.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.getdot.ai/train-dot/evaluation.md).
# Evaluation
Check Dot's answers against trusted numbers.
Evaluations check whether Dot answers your business questions correctly. Start with a few questions that matter—revenue, paying customers, conversion—and pair each with a trusted answer from a dashboard, reviewed SQL, or a governed metric.
## Create and run
1. Open **Model → Evaluation → New evaluation** and follow the setup conversation.
2. Give each question a clear metric, time period, and relevant filters. Use the same data snapshot for the expected answer and Dot's query.
3. Run the evaluation. It tests the active environment, so use the [environment switcher](/train-dot/environments.md) to choose a saved version first.
4. Open **Needs attention** to inspect failures and their linked conversations.
Dot compares the observed and expected values. Numeric tolerance defaults to **3%**; set it to **0** for an exact check. A completed run can still contain failures or errors. Evaluations use Economy mode and may make one follow-up attempt when the first answer needs clarification.
Inspect the expected and observed answers, then open the conversation to understand a failure.
Fix missing context or incorrect metric definitions, then rerun. Change an expected answer only when its business definition or source data changes.
## Versions and environments
Evaluations are saved as `evaluations/.yaml` in the model repository. Each save creates a version. Changes stay in their environment until merged to Production and participate in [GitHub](/train-dot/version-control/github.md) or [GitLab](/train-dot/version-control/gitlab.md) sync.
Use **History** to inspect or restore an older version. Archiving preserves the file, history, and previous runs. CLI/API runs keep their submitted questions and results; correcting expected answers or tolerances can regrade eligible manual runs. A frozen training keeps grading its pinned evaluation version.
## Run in CI
Choose **Run in CI** to export the saved questions and copy terminal commands or a GitHub Actions workflow. CLI suites use JSON or YAML and are separate from the model repository's file format.
Follow [Evaluations in CI](/developers/api/evaluations-in-ci.md) for setup and examples.
---
# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.
## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.
Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:
```
GET https://docs.getdot.ai/train-dot/evaluation.md?ask=&goal=
```
`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
