KeyError-FewshotpromptTemplate

[arun1980in]

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

from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder, FewShotChatMessagePromptTemplate, PromptTemplate,FewShotPromptTemplate

Question_4 = " Write Python code for App "
with open('**Path to my file/test.py', 'r') as file:
    # Read the content of the file
    Answer_4 = file.read()
examples = [
    {"Question": Question_4, "Answer": Answer_4},
]

example_prompt = PromptTemplate(
    input_variables=["Question","Answer"],
    template="Question: {Question}\nAnswer: {Answer}",

prompt_with_examples = FewShotPromptTemplate(
    examples=examples,
    example_prompt=example_prompt,
    suffix="Question:{Input}",
    input_variables=["Input"],
)

print(prompt_with_examples.format(Input="ABC"))

Description

I am trying to read a simple .py file and make that as an example answer in the fewshotprompt template. It is giving a keyError for this one example input, rest seems to be working fine. Below is a content of the test.py file i am trying to read above. the Key Error starts at "Cell_ID" and propagates further down. I am not sure what is causing the issue in the method of .format. All I intend to do is be able to do a fewshot prompt with Q/A format when the answer is the content of the .py file as is.. and then be able to use example selector after this..

      for counter_name, per_counter_data in self.incoming_data_dict.items():
                         cm_updates.append({
                        "cell_id": str(cell_id),
                        "parameter_name": f"{counter_name}_avg",
                        "parameter_value": avg,
                    })

                self.state_data_dict[counter_name][cell_id] = avg
                # Take one item to slide the window
                per_cell_data.pop(0)

   
    if cm_updates:
        cm_data = {
            "values": cm_updates
        }

        job_data = {
            "data_type": rapp.cmwrite_info_type,
            "job_definition": cm_data,
        }
        self.job_request(job_data)

System Info

I am on Windows machine, using WSL and jupyter lab
langchain==0.2.11
langchain-anthropic==0.1.3
langchain-cli==0.0.21
langchain-community==0.0.28
langchain-core==0.2.26
langchain-experimental==0.0.50
langchain-openai==0.0.8
langchain-text-splitters==0.2.2
langchainhub==0.1.20

Python 3.11.6

[dosubot[bot]]

To resolve the KeyError in FewShotPromptTemplate when using the .format method, ensure that the keys in your examples match the input_variables defined in the example_prompt. The KeyError typically occurs when the keys in the examples dictionary do not align with the expected input variables.

Here's a step-by-step approach to resolve this issue:

1. Check the input_variables in example_prompt: Ensure that the keys in your examples dictionary match the input_variables defined in the example_prompt.

2. Format the examples correctly: When formatting the examples, make sure that each example dictionary contains all the keys specified in the input_variables.

Here is an example to illustrate this:

from langchain_core.prompts import FewShotPromptTemplate, PromptTemplate

# Define the example prompt template
example_prompt = PromptTemplate(
    input_variables=["Question", "Answer"],
    template="Question: {Question}\nAnswer: {Answer}"
)

# Define the few-shot prompt template
few_shot_prompt = FewShotPromptTemplate(
    examples=[
        {"Question": "Write Python code for App", "Answer": "def app():\n    pass"}
    ],
    example_prompt=example_prompt,
    suffix="Question: {Input}",
    input_variables=["Input"]
)

# Format the prompt
formatted_prompt = few_shot_prompt.format(Input="ABC")
print(formatted_prompt)

In this example, the examples list contains dictionaries with keys "Question" and "Answer", which match the input_variables in the example_prompt. This alignment ensures that the .format method can correctly substitute the values without encountering a KeyError ^[1].

Additionally, if your .py file contains curly brackets {}, you might need to escape them before formatting. You can use the following helper functions to escape curly brackets in your examples:

def escape_examples(examples):
    return [{k: escape_f_string(v) for k, v in example.items()} for example in examples]

def escape_f_string(text):
    return text.replace('{', '{{').replace('}', '}}')

# Read the content of the file
with open('**Path to my file/test.py', 'r') as file:
    Answer_4 = file.read()

# Escape curly brackets in the examples
examples = escape_examples([
    {"Question": "Write Python code for App", "Answer": Answer_4}
])

# Define the few-shot prompt template with escaped examples
few_shot_prompt = FewShotPromptTemplate(
    examples=examples,
    example_prompt=example_prompt,
    suffix="Question: {Input}",
    input_variables=["Input"]
)

# Format the prompt
formatted_prompt = few_shot_prompt.format(Input="ABC")
print(formatted_prompt)

This workaround ensures that curly brackets are properly escaped before f-string formatting, preventing errors ^[2].

To continue talking to Dosu, mention @dosu.

[arun1980in]

Thanks, it worked. Escaping the curly brackets in the example resolved this issue. What causes this issue? it is format method taking curly brackets as arguments? Are there other such aspects that we need to consider and escape to use this method for future examples

[dosubot[bot]]

The issue with the format method taking curly brackets as arguments is that the method interprets curly brackets {} as placeholders for variables. If these placeholders are not properly named or escaped, it can lead to errors like KeyError or ValueError.

For future examples, consider the following guidelines:

1. Named Placeholders: Ensure all placeholders in the template are named (e.g., {foo}) rather than positional (e.g., {}).
2. Escaping Curly Brackets: If you need literal curly brackets in your output, escape them by doubling them (e.g., {{ and }}).

Here is an example test case demonstrating the issue:

def test_does_not_allow_args() -> None:
    """Test formatting raises error when args are provided."""
    template = "This is a {} test."
    with pytest.raises(ValueError):
        formatter.format(template, "good")

This test confirms that the formatter.format method does not support positional arguments and will raise an error if they are used ^[1].

Additionally, when working with structured formats like JSON or YAML, ensure to escape curly brackets properly. For example:

STRUCTURED_FORMAT_INSTRUCTIONS = """The output should be a markdown code snippet formatted in the following schema, including the leading and trailing "```json" and "```":

```json
{{
{format}
}}
```"""

By following these guidelines, you can avoid common pitfalls with the format method and ensure your templates are correctly formatted ^[2].

To continue talking to Dosu, mention @dosu.