Managing Connections

In the Admin Console, click Connections to view the list of existing connections.

1. Click the Add Connection (+) icon to create a new connection.
2. After a connection is created, you can edit, disable/enable, or delete it. In the future, observability metrics related to connections will also be provided to facilitate issue diagnosis.
3. On the left tree-structured page, hover over a connection name, and three dots will appear on the right. Click them to reveal quick action buttons for the corresponding connection.

Connections are categorized into two types: TDengine TSDB and AI. TDengine TSDB connections associate a TDengine TSDB instance with TDengine IDMP, and AI connections integrate LLMs with TDengine IDMP for AI chat and recommendations.

TDengine TSDB Connections

TDengine TSDB connections are used for business data management. After creating a connection, you can import data from it into the IDMP platform for management. It supports two authentication methods: username/password or token. In the form, for the Explorer Address field, please enter the Explorer address corresponding to the current TDengine TSDB connection, which is usually http://[host]:6060.

AI Connections

TDengine IDMP includes a default connection to the OpenAI API with a 3-day trial period. To avoid affecting AI-related functionality, add your own API key before the trial period ends.

Procedure

Obtain an OpenAPI AI Key

1. Log in to the OpenAPI platform.

2. In the main menu on the left, select API Keys.

3. Click Create new secret key.

4. Enter a name for the key and select any project.

5. Set Permissions to All.

6. Click Create secret key.

7. Click Copy to obtain your key and save it in a secure location.

   important

   Do not click Done until you have saved your key. You cannot view the key again, and if you lose it, you will be required to create a new one.

   For more information about OpenAI API keys, see the OpenAI documentation.

Next, add your new key to TDengine IDMP.

Add Your API Key in TDengine IDMP

1. Click your profile in the top right and select Admin Console.
2. In the main menu on the left, select Connections > OpenAI.
3. Click the edit (pencil) icon.
4. In the API Key field, delete the existing value. Then enter the key that you obtained in the preceding procedure.
5. Click Check to verify that the key is valid.
6. Click Save.

TDengine IDMP now uses the API key from your OpenAI account.

Advanced Options

When editing a connection form and selecting the type as AI, you can refer to the following instructions:

1. Models: The dropdown lists several commonly used models. You can select one or enter directly. If entering manually, it is recommended to provide two models separated by English commas—the first for general chat and the second for deep thinking. If only one model is provided, the deep thinking feature in AI chat will not have an additional effect. If more than two are provided, only the first two will be used, and the rest will be discarded.
2. Url: The dropdown lists several commonly used AI service urls. You can select one or enter directly. When selecting, ensure the chosen url is compatible with the entered models. When entering manually, make sure the url follows the OpenAI API conventions.
3. Auth Type: By default, API Key is used. However, locally deployed AI services usually do not require a key, so you can select None.

If the Models does not use an item from the dropdown, or the Url does not use an item from the dropdown, or the Models and Url do not follow the recommended correspondence, the configuration may be incorrect. Please use the following Python code for validation, or use other similar code:

from openai import OpenAI

if __name__ == "__main__":
  """
  Test the LLM interface
  
  base_url: Service url
  api_key: API key
  model: Model name
  """
  openai_client = OpenAI(
      base_url="[Service url]",
      api_key="[API key]",
  )
  messages = [
    {"role": "system", "content": "System prompt"},
    {"role": "user", "content": "User prompt"}
  ]
  response = openai_client.chat.completions.create(
      model="[Model name]",
      messages=messages,
      temperature=0,
      stream=False,
      timeout=300,
  )
  content = response.choices[0].message.content
  print(content)