# Advanced Memory Manager in Prompt Widget

Many creators use context to manage LLM chat memory, which can be complicated and redundant.

We've introduced a new, more convenient method for managing memory in the prompt widget that offers enhanced capabilities.

Here's an example of chatting with auto-memory using Claude 3 Haiku:

```json
{
  "id": "prompt_widget_auto_memory",
  "initial": "init_state",
  "states": {
    "init_state": {
      "render": {
        "text": "Hello, I am an example agent. I'm here to demonstrate auto-memory in the widget. Let's chat."
      },
      "transitions": {
        "CHAT": "home_state"
      }
    },
    "home_state": {
      "inputs": {
        "text": {
          "type": "text",
          "source": "IM",
          "user_input": true
        }
      },
      "tasks": [
        {
          "name": "llm_widget_example_task",
          "module_type": "AnyWidgetModule",
          "module_config": {
            "widget_id": "1744218088699596809", // Claude 3 Haiku
            "user_prompt": "{{text}}",
            "system_prompt": "Be a nice agent.", // Optional field. You can input system prompt of agent.
            "top_p": 0.5, // Optional field. Default value is 0.5
            "temperature": 0.5, // Optional field. Default value is 0.5
            "frequency_penalty": 0, // Optional field. Default value is 0
            "presence_penalty": 0, // Optional field. Default value is 0
            "max_tokens": 1024, // Optional field. Default value is 1024
            "memory": "auto"
            "output_name": "result"
          }
        }
      ],
      "render": {
        "text": "{{result}}"
      },
      "transitions": {
        "CHAT": "home_state"
      }
    }
  }
}
```

The key to this configuration is using `auto` in the `memory` field within the module config:

```json
 "module_config": {
           ...
            "memory": "auto",
          }
```

We automatically manage your chat memory in our store, eliminating the need to store it again in context. In more detail, memory management treats each LLM task as the smallest unit. For example, if you use auto-memory in different states or different tasks, their memory management is separate and distinct—tasks do not share memory. Additionally, if you change the state name to which a task belongs or alter the order of the tasks array, memory may be lost after saving.

### Additional Prompt

For users who want to use the fulfill mechanism or add prefixes and suffixes to user input, automatic memory management now includes a configuration to achieve this function.

```json
 "module_config": {
            "user_prompt": "{{text}}",
            "prompt_addition": {
              "user_prompt_prefix": "[User Input Prefix]",
              "user_prompt_suffix": "[User Input Suffix]",
              "pre_messages": [
                {"role": "user", "content": "Hello, I am a user message."},
                {"role": "assistant", "content": "Hello, I am an agent message."}
              ],
              "post_messages": [
                {"role": "user", "content": "Hello, I am a user message."},
                {"role": "assistant", "content": "Hello, I am an agent message."}
              ]
            },
            "memory": "auto",
            ...
         }
```

The key is `prompt_addition`. It has four fields that allow you to create a more complex LLM chat system:

| Name                 | Type        | Description                                  |
| -------------------- | ----------- | -------------------------------------------- |
| user\_prompt\_prefix | string      | Adds prefix to user's prompt                 |
| user\_prompt\_suffix | string      | Adds suffix to user's prompt                 |
| pre\_messages        | array\<map> | Adds messages before memory                  |
| post\_messages       | array\<map> | Adds messages after memory and user's prompt |

If you set all of these fields, the request messages order will be:

```json
[
pre_messages...,
memory...,
user_prompt_prefix + user_prompt + user_prompt_suffix,
post_messages...
]
```

### Message Editing

With using auto memory, you can delete or edit single message whether it is sent by the user or the agent to change its corresponding memory. No need to clear all context anymore.


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.myshell.ai/create/pro-config-mode/advanced/advanced-memory-manager-in-prompt-widget.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.