Skip to main content

Using Otto-m8: Simple Chat

This tutorial is demonstration on how to use otto-m8, by building a simple chatbot powered by OpenAI API.

1. Name your Workflow​

The first step would be to name your workflow. Make sure its a unique name for each workflow you create, as the name will form the basis of the Docker image that will be created.

workflowname

2. Select the Open AI Chat Block​

Now that you've named your Workflow, you'll need to select the OpenAI Chat Completion Block from the Add Block button. Once your block appears on the screen, configure the block. For OpenAI, you'll need an OpenAI API key.

step2

3. Connect the blocks​

When you connect a block's output to another block's input, you're essentially allowing the blocks to pass data between each other. Therefore, without passing the input block to the chat block, you won't see the block's name show up. But once you do, you can then use that in the prompt template.

connect

In this example, we're simply gonna pass the user input as is via {user_input} but you could essentially use the Prompt Template as a way to add custom prompts.

4. Add a Chat Output Block​

Adding a Chat Output Block is optional. Do so if you plan to use the Chat UI, or else, the Output Block will be enough if you plan to use it as an API.

deploy

5. Interact​

Once deployed, you can use the Chat UI to chat with your workflow.

interact

To use this workflow as an API, you'll need to know the names of the Input Block. This could be found when you click on the Input Block, within the Block Config side bar.


The last thing you'd need to know is the endpoint for your workflow, which can be found on the Templates page on the left nav bar. If it's your first workflow, then it should look something like this: http://localhost:8001/workflow_run. With that, write the following code, and run:

import requests
import json

deployment_url = "http://localhost:8001/workflow_run"
payload = {
   # Based on the name you found, replace spaces with an underscore.
   "Input_Block": "Write a poem, haiku style, about otto-m8, a low code automation platform"
}

request = requests.post(
    deployment_url, 
    json={"data": payload}
)
response = request.json()['message']

response = json.loads(response)
print(response)