Input Variable Error When using Example Selector

[komi786]

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

def extract_information(input_text, examples, model):
    try:
        active_model = LLMManager.get_instance(model=model)
        print(f"Examples={examples}")
        embedding = FastEmbedEmbeddings()
        example_selector = SemanticSimilarityExampleSelector.from_examples(
            examples,
            embedding,  
            FAISS,
            k=4
        )
        few_shot_prompt = FewShotChatMessagePromptTemplate(
            example_selector=example_selector,
            example_prompt=ChatPromptTemplate.from_messages(
            [("human", "{input}"), ("ai", "{output}")]
            )
        )
        print(f"loaded model")
        base_prompt=f"""Role: You are a helpful assistant expert in information extraction.....\n
                Input Text: "{input_text}"\n,
                output: """

        final_prompt = (
                SystemMessagePromptTemplate.from_template(
                    base_prompt
                )
                + few_shot_prompt
                + HumanMessagePromptTemplate.from_template("{input_text}")
            )
        print(f"final prompt={final_prompt.format()}")
        chain = final_prompt | active_model
        result = chain.invoke({"input_text": input_text})
        print(f"result={result}")
    except Exception as e:
        print(f"Error in prompt:{e}")
        return None

Description

Hello, I'm new to using LangChain and currently experimenting with the SemanticSimilarityExampleSelector to improve efficiency by avoiding unnecessary examples in my setup. However, I've encountered an issue when trying to integrate an ExampleSelector with a ChatPromptTemplate.

Here’s what I’m trying to achieve:
I want to use ExampleSelector within ChatPromptTemplate to dynamically select relevant examples based on the input.

The implementation works fine without the ExampleSelector, but when I add the selector, I encounter the following error:

Error Message:
Error in prompt: 'input_text'. It probably throws error before final prompt as It does not get printed.
`
def extract_information(input_text, examples, model):
try:
active_model = LLMManager.get_instance(model=model)
print(f"Examples={examples}")
embedding = FastEmbedEmbeddings()
example_selector = SemanticSimilarityExampleSelector.from_examples(
examples,
embedding,
FAISS,
k=4
)
few_shot_prompt = FewShotChatMessagePromptTemplate(
example_selector=example_selector,
example_prompt=ChatPromptTemplate.from_messages(
[("human", "{input}"), ("ai", "{output}")]
)
)
print(f"loaded model")
base_prompt=f"""Role: You are a helpful assistant
Input Text: "{input_text}"\n,
output: """

    final_prompt = (
            SystemMessagePromptTemplate.from_template(
                base_prompt
            )
            + few_shot_prompt
            + HumanMessagePromptTemplate.from_template("{input_text}")
        )
    print(f"final prompt={final_prompt.format()}")
    chain = final_prompt | active_model
    result = chain.invoke({"input_text": input_text})
    print(f"result={result}")
except Exception as e:
    print(f"Error in prompt:{e}")
    return None`

System Info

linux
langchain-python
latest langchain package

[dosubot[bot]]

Hey there, @komi786! I'm a bot here to help you with any issues you're experiencing. Let's work together to solve this problem!

To resolve the error Error in prompt: 'input_text' when integrating an ExampleSelector with a ChatPromptTemplate, ensure that you are providing the correct input variables required by the FewShotChatMessagePromptTemplate. The error typically occurs when the input variables expected by the template are not provided or are incorrectly named.

Here is an example of how to correctly set up and use the FewShotChatMessagePromptTemplate:

import { FewShotChatMessagePromptTemplate } from 'path-to-few_shot';
import { PromptTemplate } from 'path-to-prompt';
import { SemanticSimilarityExampleSelector } from 'path-to-example_selector';
import { OpenAIEmbeddings } from 'path-to-openai-embeddings';
import { HNSWLib } from 'path-to-hnswlib';

// Define the example prompt template
const examplePrompt = PromptTemplate.fromTemplate(
  "Input: {input}\nOutput: {output}"
);

// Create an example selector
const exampleSelector = await SemanticSimilarityExampleSelector.fromExamples(
  [
    { input: "happy", output: "sad" },
    { input: "tall", output: "short" },
    { input: "energetic", output: "lethargic" },
    { input: "sunny", output: "gloomy" },
    { input: "windy", output: "calm" },
  ],
  new OpenAIEmbeddings(),
  HNSWLib,
  { k: 1 }
);

// Create the FewShotChatMessagePromptTemplate
const dynamicPrompt = new FewShotChatMessagePromptTemplate({
  exampleSelector,
  examplePrompt,
  prefix: "Give the antonym of every input",
  suffix: "Input: {adjective}\nOutput:",
  inputVariables: ["adjective"],
});

// Format the dynamic prompt with the input 'rainy'
console.log(await dynamicPrompt.format({ adjective: "rainy" }));

Ensure that the inputVariables array in the FewShotChatMessagePromptTemplate includes all the variables you intend to use in the prompt, and that you provide these variables when calling the format method. In this example, the input variable adjective is correctly provided when formatting the prompt ^[1].

To continue talking to Dosu, mention @dosu.