How to Make Parallel Requests for the Same Text Using Different Variables

[luanans]

Checked other resources

• I added a very descriptive title to this question.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.

Commit to Help

• I commit to help with one of those options 👆

Example Code

chains = {}
inputs = {}
for category, content in prompts.get("medical_queries", {}).items():
    try:
        prompt_content = content["prompt"]
        prompt_examples = content["examples"]

        prompt_template = ChatPromptTemplate.from_template(prompt_content)

        chains[category] = LLMChain(prompt=prompt_template, llm=llm)

        inputs[category] = {"input": text, "example": content["examples"]}
    except Exception as e:
        print(f"{category}: {e}")

pipeline = RunnableParallel(chains)

responses = pipeline.invoke(inputs)

Description

I’m a beginner with LangChain, and I’m working on a project that requires making parallel requests to process the same input text across multiple categories. Here’s the challenge: each category needs a different set of examples to guide the output generation.

Here’s what I’m trying to achieve:

• I have a single input text that needs to be processed.
• For each category (e.g., "medications," "family history," "exams"), I have a unique prompt and a specific set of examples.
• I want to execute these requests in parallel to improve efficiency.
• I’m using RunnableParallel, but I’m struggling to figure out the best way to handle the examples dynamically for each category.

What I’ve tried so far:

• Passing examples dynamically during the request.
• Formatting prompts by embedding examples beforehand. However, I either encounter issues with missing inputs variables.

I’m new to LangChain, so any help or suggestions (even simple ones!) would be highly appreciated.

System Info

langchain==0.3.4
langchain-core==0.3.12
langchain-google-genai==2.0.1
langchain-google-vertexai==2.0.5
langchain-text-splitters==0.3.0
Note: you may need to restart the kernel to use updated packages.

[feijoes]

@luanans Another approach, instead of using RunnableParallel, could be to try threads, but your current approach seems fine. Could you share an example of a "set of examples" for one of your categories? Also, does each category have different input variables?