> ## Documentation Index
> Fetch the complete documentation index at: https://docs.getunblocked.com/llms.txt
> Use this file to discover all available pages before exploring further.

# PR Chat

> Use PR Chat to ask questions, request focused reviews, get PR summaries, update pull request descriptions, and generate diagrams directly in your pull requests

## What is PR Chat?

PR Chat lets you interact with Unblocked directly in your pull request by mentioning `@unblocked` in a comment. Beyond requesting standard reviews and summaries, you can have a conversation with Unblocked to get deeper insight into the code, request targeted re-reviews, or generate visual diagrams.

PR Chat understands natural language — not just keywords or fixed commands. A short prompt like `@unblocked review` works, and so does `@unblocked can you review just the python code`. You can include arbitrary guidance to shape the response.

Unblocked reacts to your comment with 👀 to confirm it's working, then posts its response when complete.

## Request a focused review

After an initial review, you may want Unblocked to look at your PR again with a different focus. Mention `@unblocked` with specific instructions to guide the review.

**Quick review**

```text theme={null}
@unblocked review
```

**Provide additional context**

Give Unblocked more information about intent or constraints so it can refine its feedback:

```text theme={null}
@unblocked review this again with more focus on
input validation and error handling.
```

**Focus on a specific area**

Ask Unblocked to concentrate on a particular aspect of the changes:

```text theme={null}
@unblocked can you review just the python code
```

**Re-review after new commits**

When you've pushed additional commits to address earlier feedback, ask for a fresh look at specific files:

```text theme={null}
@unblocked review just src/auth/session.ts again.
```

## Ask questions

Use PR Chat to ask Unblocked questions about the code in your pull request, the broader system, or the intent behind the changes. Unblocked draws on the context connected to your workspace to provide informed answers.

**Understand the changes**

```text theme={null}
@unblocked what does this PR change about retry logic?
```

**Explore the broader system**

```text theme={null}
@unblocked how does the auth middleware work in this repo?
```

**Clarify intent or design decisions**

```text theme={null}
@unblocked why was this service split into two modules?
```

Because Unblocked has access to your connected context — including past pull requests, documentation, and team discussions — answers go beyond what's visible in the diff.

## Get a PR summary

Ask Unblocked to summarize a pull request and it will post a concise explanation of what changed and why, directly in the PR comment thread. This is useful when you want a quick overview before reviewing, or need to catch up on a PR you've been tagged on.

```text theme={null}
@unblocked summarize this PR
```

You can also narrow the scope:

```text theme={null}
@unblocked create a summary of the changes to the infra module
```

## Write or update the PR description

Ask Unblocked to write or update the pull request description field itself — not just post a summary in chat. Unblocked generates a description based on the diff and any connected context, then edits the PR description in place.

```text theme={null}
@unblocked update PR description
```

This is useful when you've opened a PR without a description, or want to replace a placeholder with something more informative before requesting review.

## Request diagrams

Ask Unblocked to generate diagrams that visualize data flows, architecture, or relationships in the codebase. Diagrams are rendered directly in the PR comment thread.

```text theme={null}
@unblocked draw the data flow for order processing in this PR.
```

```text theme={null}
@unblocked illustrate how the API gateway interacts with
the downstream services touched here.
```

Diagrams are useful for onboarding reviewers, documenting complex flows, and confirming that your mental model matches the actual implementation.
