Formatting

Both your messages and the assistant's responses support rich formatting. You can use Markdown syntax in your messages, and the assistant uses the same formatting in its responses to present information clearly.

Markdown support

Standard Markdown syntax works in messages and responses.

Text formatting

Headers

Use headers to organize longer messages:

## Main section
### Subsection
#### Smaller heading

The assistant uses headers to structure longer responses, especially when presenting multi-step findings or comparing options.

Lists

Unordered and ordered lists render as expected:

- First item
- Second item
  - Nested item

1. Step one
2. Step two
3. Step three

Links

Links render as clickable text:

[Link text](https://example.com)

When the assistant references documentation or external resources, it formats them as links you can click directly.

Blockquotes

Use blockquotes to highlight important information or quote log messages:

> This is a quoted block

Code blocks

Code blocks are essential for sharing error messages, configuration snippets, and sample code with the assistant.

Syntax highlighting

Specify a language after the opening fence for syntax highlighting:

```python
def calculate_error_rate(errors, total):
    return (errors / total) * 100
```

The assistant automatically applies syntax highlighting when it shares code snippets, query examples, or configuration files.

Supported languages

Common languages with syntax highlighting include:

Inline code

Use backticks for inline code, field names, and technical terms:

The `user_id` field in the `checkout_events` table contains the value.

This helps distinguish technical identifiers from regular text.

Mermaid diagrams

The assistant can create diagrams using Mermaid syntax. These diagrams render as interactive visualizations in the conversation.

Requesting diagrams

Ask the assistant to create diagrams for:

• System architecture
• Request flows and sequences
• Error propagation paths
• State machines and workflows
• Timeline visualizations

Example prompts:

Draw a sequence diagram showing the checkout flow between services.

Create a flowchart showing how requests are routed through the API gateway.

Diagram types

The assistant supports several Mermaid diagram types:

Flowcharts for decision trees and process flows:

flowchart TD
    A[Request] --> B{Auth valid?}
    B -->|Yes| C[Process]
    B -->|No| D[Reject]

Sequence diagrams for service interactions:

sequenceDiagram
    Client->>API: POST /checkout
    API->>PaymentService: processPayment()
    PaymentService-->>API: success
    API-->>Client: 200 OK

State diagrams for status transitions:

stateDiagram-v2
    [*] --> Pending
    Pending --> Processing
    Processing --> Completed
    Processing --> Failed
    Failed --> Pending

Interacting with diagrams

Diagrams in the dashboard support:

• Zoom: Scroll or pinch to zoom in and out
• Pan: Click and drag to move around larger diagrams
• Reset: Double-click to reset the view

Tables

Tables organize structured data for easy comparison and reference.

Creating tables

Use standard Markdown table syntax:

| Service | Error Count | Error Rate |
|---------|-------------|------------|
| api     | 142         | 2.3%       |
| auth    | 89          | 1.1%       |
| payment | 12          | 0.4%       |

Assistant-generated tables

When you ask for comparisons or aggregations, the assistant often presents results as tables:

Show me error counts by service for the last hour.

The assistant formats the response as a table with sortable columns. You can request specific columns:

Show me the top 10 slowest endpoints with their p50, p95, and p99 latencies.

Table artifacts

For larger datasets, the assistant may create a table artifact instead of inline Markdown. Table artifacts support:

• Column sorting
• Filtering
• Pagination for large datasets
• Export to CSV

Export and copy options

You can export and copy formatted content from conversations.

Copying text

To copy text from a response:

1. Select the text you want to copy
2. Use Cmd/Ctrl + C or right-click and select Copy

Copying code blocks

Code blocks include a copy button in the top-right corner. Click it to copy the entire code block to your clipboard without selecting text.

Exporting artifacts

Artifacts like charts, tables, and diagrams can be exported:

To export an artifact:

1. Click the artifact to expand it
2. Click the export button (download icon)
3. Select the format

Copying diagrams

Mermaid diagrams can be copied as:

• Image: PNG for sharing in documents or presentations
• Source: The Mermaid code to recreate the diagram elsewhere

Sharing formatted content

When sharing thread content with teammates:

• Links to threads preserve all formatting
• Shared threads include artifacts and diagrams
• Export individual artifacts for use outside Sazabi

Tips for effective formatting

When to use code blocks

Use code blocks for:

• Error messages and stack traces
• Log snippets you want the assistant to analyze
• Configuration files
• Sample code or queries

I'm seeing this error:

```python
TypeError: 'NoneType' object is not subscriptable
  File "checkout.py", line 42, in process_order
    item_id = order['items'][0]['id']

What could cause this?

### When to request diagrams

Request diagrams when you need to:

- Visualize complex service interactions
- Understand error propagation paths
- Document system architecture
- Present findings in incident reviews

### Keeping messages clear

- Break complex questions into separate messages
- Use headers to organize multi-part questions
- Include relevant code or logs in code blocks
- Specify what format you want for the response

## Next steps

<Cards>
  <Card title="Understanding Responses" href="/chat/threads/responses">
    Learn how to interpret tool calls, artifacts, and evidence.
  </Card>
  <Card title="Effective Prompting" href="/chat/prompting">
    Write better prompts that get more useful results.
  </Card>
</Cards>