> ## Documentation Index
> Fetch the complete documentation index at: https://docs.4minds.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# How SYMI works

> What happens between sending a message and seeing the answer in SYMI.

SYMI runs each turn in a private agent runtime that belongs to your account. This page explains what happens between sending a message and seeing the answer.

## One private runtime per account

SYMI starts your runtime when you need it, and starts preparing it as soon as you begin signing in. Your conversations and files live there. Other accounts cannot reach it. Your runtime has no direct internet access. Requests leave only through the 4Minds platform, which blocks internal addresses, and SYMI can reach public websites this way. SYMI may stop a runtime that is not being used and starts it again when you need it. A response that is still running keeps the runtime alive.

SYMI keeps your session fresh across browser tabs. If you log out of 4Minds, including with **Log out** in **Settings**, you are signed out of SYMI too.

## What happens when you send a message

1. SYMI checks that the message comes from your signed-in session.
2. If SYMI is already responding, your message steers the current response or waits for the next one. You can set your default in **Settings**.
3. SYMI prepares the request: your message, the conversation so far, your memory when it applies, and the tools available.
4. The model answers. When it needs a tool, such as reading a file or calling a connection, SYMI runs the tool and sends the result back to the model.
5. SYMI repeats steps 3 and 4 until the task is done, then shows the answer and saves it to your conversation.

> [!NOTE]
>
> You can stop a response at any time with the stop button. Stopping also ends the helpers SYMI started for that response.

## Models and reasoning

SYMI runs on its own model by default and can also answer from any 4Minds model you have trained. With SYMI's default model, use the reasoning control in the chat box to choose **High** (default) or **Max**. SYMI saves the level for each conversation.

## Limits

| Limit | Value |
| - | - |
| Message size, including attached files | 65,536 tokens (about 65K) |
| Messages waiting while SYMI responds | 5 |
| Context window | Up to about 524K tokens |

SYMI in the browser estimates the size of a message before it sends it: the number of characters divided by 4, plus the number of words divided by 4. This rough estimate is usually higher than the true token count. If a message is over the limit, SYMI does not send it and tells you about how many tokens it has. Shorten the message, remove a file, or split it into parts.

If a conversation is too long, SYMI tells you to use `/compact` or start a new chat.

## SYMI in the browser and the 4Minds Symi CLI

SYMI in the browser runs in the 4Minds cloud. The 4Minds Symi CLI runs the same agent engine on your own computer and links to your 4Minds account. Replies come from SYMI's default model, so the CLI needs an internet connection, and once linked it can search the web through 4Minds. To work with files on your own computer, use the 4Minds Symi CLI. If you start it before linking, it asks you to run `symi link`.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.