Skip to main content

5.5 Hands-On Example: HTTP Summary with LLM Switch

info

Please note that this example is provided as a prototypical implementation, NOT as a production-ready integration. Therefore certain datapoints may require manual adaptation, especially when referencing external resources.

In this hands-on, we reuse the Hello World regulation example from this Chapter: How to Start - Hello World and turn it into a workflow.

Instead of calling a connector directly, the orchestration logic is expressed as a workflow that:

  • Fetches content from a URL using an HTTP GET datapoint
  • Stores the fetched content using a set task
  • Dynamically selects between two LLM summarization datapoints
  • Returns a summarized version of the content

What we learn​

  • How step-input chaining works in workflows
  • How to use the set task to store intermediate results
  • How to use a switch task to dynamically select between datapoints
  • How existing datapoints can be reused unchanged inside workflows

Workflow definition​

This workflow reuses the following datapoints from the Hello World scenario:

  • helloworld-dp1-hello-world-http-get
    Fetches content from a user-provided URL via HTTP GET
  • helloworld-dp3-hello-world-prompt-chatgpt
    Summarizes content using ChatGPT
  • helloworld-dp4-hello-world-prompt-mistral
    Summarizes content using Mistral

The workflow performs the following steps:

  1. Call helloworld-dp1-hello-world-http-get with the URL provided as workflow input

  2. Store the returned response content in a workflow variable named httpContent

  3. Use a switch task on root.model to select the LLM datapoint:

    • chatgpt → ChatGPT summarization datapoint
    • mistral → Mistral summarization datapoint
  4. Call the selected summarization datapoint with httpContent

  5. Return the generated summary as the workflow result

Workflow configuration​

Example Workflow

Example Workflow

{
"id": "helloworld-wf2-http-summary-llm-switch",
"name": "helloworld-wf2-http-summary-llm-switch",
"type": "workflow",
"version": "0.0.1",
"displayName": "helloworld-wf2-http-summary-llm-switch",
"tags": [
"example",
"simple-example",
"scenario-helloworld",
"scenario-example-helloworld"
],
"description": "Fetch content via an HTTP GET datapoint, store it in a workflow variable, and summarize it using either ChatGPT or Mistral based on workflow input.",
"workflow": {
"type": "temporal",
"definition": {
"do": [
{
"fetchContent": {
"call": "olive",
"with": {
"datapoint": "helloworld-dp1-hello-world-http-get",
"parameters": {
"url": "${root.url}"
}
}
}
},
{
"storeContent": {
"set": {
"httpContent": "${input}"
}
}
},
{
"chooseModel": {
"switch": [
{
"useChatGPT": {
"when": "${root.model = \"chatgpt\"}",
"then": "summarizeWithChatGPT"
}
},
{
"useMistral": {
"when": "${root.model = \"mistral\"}",
"then": "summarizeWithMistral"
}
},
{
"fallback": {
"when": "${true}",
"then": "summarizeWithChatGPT"
}
}
]
}
},
{
"summarizeWithChatGPT": {
"call": "olive",
"with": {
"datapoint": "helloworld-dp3-hello-world-prompt-chatgpt",
"parameters": {
"content": "${httpContent}"
}
},
"then": "end"
}
},
{
"summarizeWithMistral": {
"call": "olive",
"with": {
"datapoint": "helloworld-dp4-hello-world-prompt-mistral",
"parameters": {
"content": "${httpContent}"
}
},
"then": "end"
}
}
]
}
}
}

Example trigger via data-orchestrator:

Request Input

{
"id": "helloworld-wf2-http-summary-llm-switch",
"input": {
"url": "<your-url>",
"model": "chatgpt"
}
}