Skip to main content

Near AI Extensions

After we added Near AI agent support in Bettear Bot, we realized that the UX can be improved by adding a few more features. These extensions are backward compatible with the existing Near AI interface at app.near.ai, so implementing them won't break existing user interactions.

1. Welcome buttons​

Add buttons field in the welcome object in metadata.json:

{
  "category": "agent",
  "name": "inteartest",
  "version": "0.0.1",
  "description": "",
  "tags": [],
  "details": {
    "agent": {
      "defaults": {
        "model": "llama-v3p1-70b-instruct",
        "model_provider": "fireworks",
        "model_temperature": 1,
        "model_max_tokens": 16384
      },
      "welcome": {
        "title": "Test Agent",
        "description": "This is a test.",
        "buttons": [
          [
            "Gm",
            "Still Row 1"
          ],
          [
            "Row 2"
          ]
        ]
      },
      "framework": "web-agent"
    }
  },
  "show_entry": true
}

How it looks:

Welcome buttons

2. Buttons in messages​

Add suggestions of how the user can reply to the agent's message. For example, in a staking agent, you can add buttons like "Stake more", "Stake 10 NEAR (amount that the user has)", "Unstake" etc.

Here's how to add it in the code:

env.add_reply("""
              {
                "buttons": [
                  ["Simple Text", "Column 2 on row 1"],
                  [{"LongText":{"text_on_button":"Short","text_message":"Full message that will be sent to the agent. Can even be a JSON payload"}}],
                  [{"Url":{"text":"Claim free 1 NEAR","url":"https://youtube.com/watch?v=dQw4w9WgXcQ"}}]
                ]
              }
              """, message_type = "system:buttons")
env.add_reply("Welcome!")

You should add a reply with message_type = "system:buttons" before the message itself.

How it looks:

Buttons in messages

3. Run Status​

If the run takes long time to process, you can send multiple messages to inform the user about the status. But even better is to add "ephemeral" messages that edit previous progress messages, and are removed when the agent is done. To do this, you can use env.add_reply with message_type = "system:status":

env.add_reply("Welcome!")
time.sleep(1)
env.add_reply("""
              {
                "message": "Loading...",
                "progress": 0.5
              }
              """, message_type = "system:status")
time.sleep(1)
env.add_reply("""
              {
                "message": "Almost done...",
                "progress": 0.8
              }
              """, message_type = "system:status")
time.sleep(1)
env.add_reply("Done")

The progress field is optional, and ranges from 0 (0%) to 1 (100%), so that the frontends can display a progress bar.

4. Runner Environment​

You can check os.environ["RUNNING_IN_BETTEAR"] environment variable to see if the agent is running in Bettear Bot, and communicate in more Telegram-friendly way.