> ## Documentation Index
> Fetch the complete documentation index at: https://www.doc-reviewer.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Quickstart: run your first Doc Reviewer evaluation

> From a fresh install to your first LLM-powered evaluation in minutes. Connect a model, add a project, specify evaluation criteria, upload a document and review results.

This guide walks you through a complete first evaluation in Doc Reviewer — from connecting an LLM to reviewing results and exporting them. By the end, you'll have a working evaluation workflow you can repeat on any document.

<Steps>
  <Step title="Launch Doc Reviewer">
    **Using the .exe:** Double-click `doc-reviewer.exe`. The app starts a local backend and opens the UI in your browser automatically. If your browser does not open, go to `http://localhost:8000`.

    **From source:** Run the following command from the project root:

    ```bash theme={null}
    python run_dev.py
    ```

    Then open `http://localhost:5173` in your browser.

    <Note>
      The first launch creates a `data\db.sqlite` database file next to the `.exe` and seeds it with default LLM models and evaluation criteria. This takes a few seconds.
    </Note>
  </Step>

  <Step title="Connect an LLM">
    Doc Reviewer needs an LLM to evaluate instructions. Go to **Settings → Models**.

    You'll see a list of pre-configured model entries. To activate one:

    1. Click the model you want to use (for example, `gpt-4o` or `claude-3-5-sonnet`).
    2. Enter your API key in the **API key** field.
    3. Click **Save**.
    4. Click **Set as active** to make this the model used for evaluations.

    <CodeGroup>
      ```text OpenAI theme={null}
      Base URL: https://api.openai.com/v1
      API key: sk-...
      ```

      ```text Anthropic theme={null}
      Base URL: https://api.anthropic.com/v1
      API key: sk-ant-...
      ```

      ```text Ollama (local, no key required) theme={null}
      Base URL: http://localhost:11434/v1
      API key: (leave blank)
      ```
    </CodeGroup>

    <Tip>
      If you're using Ollama, make sure the Ollama server is running locally before you run an evaluation. No API key is required.
    </Tip>
  </Step>

  <Step title="Create a project">
    Projects group documents from the same product so Doc Reviewer can use shared context to improve evaluation accuracy.

    1. Click **Projects** in the left sidebar.
    2. Click **New project**.
    3. Enter a name that identifies the product (for example, `Firewall Manager` or `API Gateway`).
    4. Click **Create**.

    You'll be taken to the project page. You can add a product context later — it helps the LLM understand your product's terminology, audience, and components when evaluating instructions.
  </Step>

  <Step title="Upload a document">
    From the project page, upload a file:

    1. Click **Add document** or drag and drop a file onto the upload area.
    2. Doc Reviewer accepts PDF, DOCX, Markdown (`.md`), and plain text (`.txt`) files.
    3. Wait for the upload and parsing to complete. The document structure appears in the tree on the left.

    Alternatively, to load a web page:

    1. Go to **Evaluate → By URL**.
    2. Paste the page URL and click **Load**.
    3. The page is fetched through a headless Chromium browser, so JavaScript-rendered pages load correctly.

    <Note>
      Loading web pages requires Chromium to be installed. See [Install Doc Reviewer](/en/installation) if you haven't set it up yet.
    </Note>
  </Step>

  <Step title="Review detected instructions">
    After parsing, Doc Reviewer automatically classifies each section of your document. Sections identified as instructions or possible instructions appear highlighted in the document tree.

    Browse the tree to confirm the detected sections look correct. If a section has been incorrectly flagged as an instruction — or missed — you can adjust its classification manually by clicking the section and changing its type. Sections marked **non-instruction** are excluded from LLM evaluation.
  </Step>

  <Step title="Run evaluation">
    Click **Evaluate** in the top bar of the document view.

    Doc Reviewer sends each instruction section to the active LLM with your evaluation criteria and streams results back in real time. A progress indicator shows how many sections have been evaluated. Depending on the number of instructions and the LLM response time, this typically takes 20–90 seconds for a mid-sized document.

    <Tip>
      If you have a product context set on the project, it's automatically included in each evaluation prompt — no extra steps needed.
    </Tip>
  </Step>

  <Step title="Review results">
    When evaluation finishes, each instruction in the document tree shows a color-coded indicator:

    | Color  | Meaning                                       |
    | ------ | --------------------------------------------- |
    | Green  | The instruction meets all criteria            |
    | Yellow | Minor issues — easy to fix                    |
    | Orange | Problems that may affect usability            |
    | Red    | Critical issues that need immediate attention |

    Click any instruction to see the full breakdown: which criteria passed, which failed, and the LLM's specific recommendations for improvement.

    If you believe a result is incorrect — for example, the LLM flagged an issue that doesn't apply to your product — click **Mark as false positive** to exclude that result from the summary score. False positive overrides are preserved across re-evaluations.

    To understand the scoring in more detail, click **How does evaluation work? ↗** in the summary bar above the results.
  </Step>

  <Step title="Export results">
    When you're ready to share results or log them for tracking, click **Export** in the top bar.

    Doc Reviewer downloads an `.xlsx` file with one row per instruction, including the section title, evaluation color, criteria results, and recommendations.

    To save a point-in-time snapshot of results — useful for comparing quality across document versions — click **Save snapshot** and give it a name. You can compare snapshots on the **Snapshots** page.
  </Step>
</Steps>

## Next steps

Now that you've completed your first evaluation, here are useful things to explore:

<CardGroup cols={2}>
  <Card title="Add product context" icon="brain" href="/en/concepts/projects">
    Generate a product context for your project to improve evaluation accuracy on domain-specific terminology.
  </Card>

  <Card title="Customize criteria" icon="list-check" href="/en/configuration/criteria-sets">
    Edit the default evaluation criteria or create a new criteria set tailored to your documentation standards.
  </Card>

  <Card title="Evaluate web pages" icon="globe" href="/en/workflows/evaluate-web-page">
    Load documentation pages directly by URL using the built-in headless browser.
  </Card>

  <Card title="Compare snapshots" icon="code-compare" href="/en/workflows/snapshots">
    Save evaluation results as snapshots and track quality improvements across document versions.
  </Card>
</CardGroup>