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

# Conversations (Chat)

> The heart of Symphony: write questions and instructions and receive responses from the AI or an agent, with attachments and advanced features.

## What it is

The **conversation (chat)** is the heart of Symphony. It's where you write questions and instructions and receive responses from the AI or an **agent**. You can enrich the conversation with **attachments**, **knowledge bases**, **prompts**, **tools**, and features like **web search**, **image generation**, and **code interpretation**.

## How to access

* When you enter the platform, you are already on the conversation screen (**Home**).
* To start a new conversation at any time, click **New conversation** in the sidebar or use the **New** button in the top right.
* Keyboard shortcut: **Ctrl/⌘ + Shift + O**.

## How to use

### 1. Choose the model or agent

At the top of the conversation area is the **model selector**. Click it to choose which **AI model** or **agent** you want to chat with.

* If your company allows **multiple models**, you can select more than one and **compare responses** side by side.
* A model must be selected to send messages. If none is chosen, the system displays **"Choose a model"**.

### 2. Write and send a message

<Steps>
  <Step title="Click the text box at the bottom of the screen" />

  <Step title="Type your message" />

  <Step title="Press Enter to send">
    Use **Shift + Enter** to add a line break without sending.
  </Step>

  <Step title="Stop, if needed">
    To stop an ongoing response, use the **stop** button (which replaces the send button while the AI is responding).
  </Step>
</Steps>

<Note>
  While an attached file is still **uploading** or **processing**, sending is blocked — wait for it to finish so the AI receives the file content.
</Note>

### 3. Attach content (+ button)

The **+** (Attach) button, to the left of the bar, opens the attachments menu. Options appear according to company permissions:

| Option               | What it does                                                                            |
| -------------------- | --------------------------------------------------------------------------------------- |
| **Upload files**     | Uploads files from your computer to the conversation.                                   |
| **Attach knowledge** | Inserts a **knowledge base** (command shortcut: `#`).                                   |
| **Attach prompt**    | Inserts a saved **prompt** (command shortcut: `/`).                                     |
| **Attach tool**      | Activates a **tool** for the message (command shortcut: `$`).                           |
| **Capture**          | Takes a photo via camera (on mobile opens the camera; on desktop captures from webcam). |

<Tip>
  You can also **drag and drop** files onto the screen, or **paste** an image directly into the text box.
</Tip>

### 4. Quick commands with `@ / # $`

Typing a special character in the message box opens a selection list:

| Type | For                                                   |
| ---- | ----------------------------------------------------- |
| `@`  | Talk to a specific **model/agent** within the message |
| `/`  | Insert a saved **prompt**                             |
| `#`  | Attach a **knowledge base**                           |
| `$`  | Attach a **tool**                                     |

<Note>
  You can attach a **maximum of 2 tools per message**. When trying a third, the system warns you.
</Note>

### 5. Extra features (more tools)

The **more tools** button (component icon) opens features you can enable for the message, when allowed:

| Feature            | What it does                                                     |
| ------------------ | ---------------------------------------------------------------- |
| **Search the web** | Allows the AI to search for updated information on the internet. |
| **Create image**   | Generates images from your description.                          |
| **Interpret code** | Executes/interprets code to solve the task.                      |

When you activate a feature, it appears as a "tag" in the bar; click **X** to remove it.

<Tip>
  **Automatic image detection:** if you type a message with a clear intent to generate an image and the **Create image** feature is available, it may be activated automatically upon sending.
</Tip>

### 6. Voice input (microphone)

When the company enables **speech transcription (STT)**, a **microphone** button appears. Click to start recording; speech is converted to text and added to the message. During recording you see a timer and buttons to **cancel** or **finish**.

### 7. Temporary chat

The **temporary chat** button (dashed speech bubble icon) activates a mode where the conversation **is not saved** in the history. Useful for specific and confidential questions. Click again to deactivate.

## Actions on each message

When hovering over a response (on mobile they are always visible), an action bar appears. Available options depend on permissions:

| Action                   | Description                                                                                      |
| ------------------------ | ------------------------------------------------------------------------------------------------ |
| **Edit**                 | Edits the message; when editing your question, a new **version** of the conversation is created. |
| **Copy**                 | Copies the response content to the clipboard.                                                    |
| **Read aloud**           | Converts the response to audio (TTS) and plays it.                                               |
| **Generate image**       | Generates an image based on the response content.                                                |
| **Information (tokens)** | Shows usage/consumption details of the response.                                                 |
| **Good / bad response**  | Sends a rating (👍/👎) that helps improve quality.                                               |
| **Continue response**    | Asks the AI to continue from where it stopped (only on the last response).                       |
| **Regenerate**           | Generates a new response for the same question.                                                  |

### Message versions

When you **edit** a question or **regenerate** a response, Symphony saves the variations as **versions**. Use the navigation arrows that appear in the message to switch between them (e.g., "1/2", "2/2").

### Advanced features in responses

* **Citations/Sources:** when the response uses documents or web search, clickable references with sources appear.
* **Web search status:** during the search, an indicator shows progress.
* **Artifacts:** rich content (such as code or visualizations) may open in a split side panel, next to the conversation.
* **Export message:** responses can be exported in formats like **DOCX**, **XLSX**, and **PDF**, when applicable.

## Best practices

* **Be specific**: explain the goal, audience, and desired response format.
* **Attach context**: use knowledge bases and files for more precise answers.
* **Rate responses** (👍/👎): feedback helps calibrate quality.
* Use **temporary chat** for topics that should not remain in history.
* For recurring tasks, save a **prompt** (see [Prompts](/en/guia/prompts)).

## Frequently asked questions

<AccordionGroup>
  <Accordion title="Why is the send button disabled?">
    Because no model is selected, the message is empty (no text and no attachments), or files are still uploading/processing.
  </Accordion>

  <Accordion title="Can I compare two models on the same question?">
    Yes, if the **multiple models** permission is active: select more than one model and the responses appear for comparison.
  </Accordion>

  <Accordion title="Did the AI search the internet?">
    Only if you activated **Search the web** (and the feature is allowed). When active, the response includes sources.
  </Accordion>

  <Accordion title="Where are the images I generated?">
    They appear in the conversation itself, alongside the corresponding message, and can be included when exporting the conversation.
  </Accordion>
</AccordionGroup>

## Known limitations

* Features like **web search**, **generate image**, **interpret code**, **voice (STT/TTS)**, **attachments**, and **temporary chat** depend on company and profile permissions.
* Maximum of **2 tools per message**.
* Voice input uses the browser's speech recognition; browsers without support display a warning and the feature doesn't work.
* Sending is blocked while attachments are uploading or processing.