RFC: Support for bidirectional communication (prompting the user)

[Pwuts]

The protocol currently supports making requests to an agent service. However, some agents may need to be able to communicate with the user in order to function optimally. For example:

1. User: please buy me a new set of cutting boards
2. AI: would you like wooden or plastic cutting boards?
3. User: I like wood
4. AI: searches for wooden cutting boards and places an order on Amazon

Adding a way for agents to prompt the user would greatly increase the versatility of the protocol imo.

Proposal

Two primary options:

1. Extension of the protocol with a status awaiting_input, and a way to resolve this status with additional input for an existing task or step

2. Extension of the task endpoint with a callback (or similar) attribute through which a client can specify a callback URL which may be polled with prompts for the user until they are resolved. Example:

   1. Giving the agent a task

   POST /agent/tasks
   {
     "input": "Please find a nice olive wood cutting board on Amazon and order it for me.",
     "callback_url": "https://my-service.url/agents/203820/callbacks"
   }
   

   2. The agent wants more info

   POST https://my-service.url/agents/203820/callbacks
   {
     "prompt": "What is your budget for this purchase?"
   }
   

   {
     "prompt_id": 123,
     "status": "pending",
     "created": "2023-08-23T13:49:51.141Z",
     "last_updated": "2023-08-23T13:49:51.141Z"
   }
   

   3. The agent polls the client until the prompt is resolved

   GET https://my-service.url/agents/203820/callbacks/123
   

   Responses:

   {
     "prompt_id": 123,
     "status": "pending",
     "created": "2023-08-23T13:49:51.141Z",
     "last_updated": "2023-08-23T13:49:51.141Z"
   }
   

   {
     "prompt_id": 123,
     "status": "resolved",
     "answer": "I don't want to spend more than €40 on this purchase",
     "created": "2023-08-23T13:49:51.141Z",
     "last_updated": "2023-08-23T13:53:12.634Z",
   }
   

   {
     "prompt_id": 123,
     "status": "rejected",
     "created": "2023-08-23T13:49:51.141Z",
     "last_updated": "2023-08-23T13:53:12.634Z",
   }
   

Alternatives

• Extending the protocol with full chatting capabilities:

  • GET /agent/tasks/<task_id>/chats List chats regarding task <task_id>

  • POST /agent/tasks/<task_id>/chats Start a new chat regarding task <task_id>

  • POST /agent/tasks/<task_id>/chats/<chat_id>/messages Post a new message in an existing chat

  • GET /agent/tasks/<task_id>/chats/<chat_id>/messages Get all messages in a chat

  • POST /agent/tasks/<task_id>/chats/<chat_id>/close Close/resolve a chat

[hackgoofer]

Interesting, I feel like chats are better implemented with websockets - What do you think? @jzanecook

[jzanecook]

This might be one of the perfect examples for a plugin referencing #71 since it's something that not all agents might have.

[ntindle]

Can you attend a meeting the Agent protocol meeting on the 12th to discuss this?