Skip to main content

Add your description here

Project description

langchain-chain-of-verification

Based off https://github.com/ritun16/chain-of-verification , packaged (with uv) and updated for newer langchain versions for easier consumption.

Can be used as CLI or library.

CoVe: https://arxiv.org/pdf/2309.11495

Enhanced by DuckDuckGo search (by ritun16)

Usage

CLI

# uvx --from langchain-chain-of-verification cove --help
$ cove --help

usage: cove [-h] --question QUESTION [--llm-name LLM_NAME] [--temperature TEMPERATURE] [--max-tokens MAX_TOKENS] [--show-intermediate-steps SHOW_INTERMEDIATE_STEPS]

Chain of Verification (CoVE) parser.

options:
  -h, --help            show this help message and exit
  --question QUESTION   The original question user wants to ask
  --llm-name LLM_NAME   The openai llm name
  --temperature TEMPERATURE
                        The temperature of the llm
  --max-tokens MAX_TOKENS
                        The max_tokens of the llm
  --show-intermediate-steps SHOW_INTERMEDIATE_STEPS
                        The max_tokens of the llm

Library

def create_cove_chain(
    original_query: str,
    llm_name="gpt-4o",
    temperature=0.1,
    router_max_tokens=500,
    show_intermediate_steps=True,
) -> str:
    """
    Creates a Chain of Verification (CoVE) using specified language models.

    Args:
        original_query (str): The original question to be processed.
        llm_name (str, optional): The name of the language model to use. Defaults to "gpt-4o".
        temperature (float, optional): The temperature setting for the language model. Defaults to 0.1.
        router_max_tokens (int, optional): The maximum number of tokens for the language model. Defaults to 500.
        show_intermediate_steps (bool, optional): Whether to show intermediate steps. Defaults to True.

    Returns:
        str: The result (final answer) of the CoVE chain processing.

    Example:
        >>> result = create_cove_chain("What is the capital of France?")
        >>> print(result)
    """
    ...

Examples

cove --question 'name athletes born in raleigh'
Chain selected: WIKI_CHAIN

################################################################################

{'baseline_response': '1. Chasity Melvin\n'
                      '2. Ryan Jeffers\n'
                      "3. Devonte' Graham\n"
                      '4. Trea Turner',
 'final_answer': 'Based on the verification questions and answers, the refined '
                 'answer should only include athletes who were confirmed to be '
                 'born in Raleigh. Therefore, the final refined answer is:\n'
                 '\n'
                 '1. Ryan Jeffers\n'
                 "2. Devonte' Graham",
 'original_question': 'name athletes born in raleigh',
 'verification_answers': 'Question: 1. Was Chasity Melvin born in Raleigh? '
                         'Answer: No, Chasity Melvin was not born in Raleigh. '
                         'She was born in Roseboro, North Carolina.\n'
                         'Question: 2. Was Ryan Jeffers born in Raleigh? '
                         'Answer: Yes, Ryan Jeffers was born in Raleigh, North '
                         'Carolina.\n'
                         "Question: 3. Was Devonte' Graham born in Raleigh? "
                         "Answer: Yes, Devonte' Graham was born in Raleigh, "
                         'North Carolina.\n'
                         'Question: 4. Was Trea Turner born in Raleigh? '
                         'Answer: No, Trea Turner was not born in Raleigh. '
                         'According to the provided context, Trea Turner was '
                         'born on June 30, 1993, in Boynton Beach, Florida.\n',
 'verification_question_template': 'Was [athlete] born in [Raleigh]?',
 'verification_questions': '1. Was Chasity Melvin born in Raleigh?\n'
                           '2. Was Ryan Jeffers born in Raleigh?\n'
                           "3. Was Devonte' Graham born in Raleigh?\n"
                           '4. Was Trea Turner born in Raleigh?'}

################################################################################

Final Answer: Based on the verification questions and answers, the refined answer should only include athletes who were confirmed to be born in Raleigh. Therefore, the final refined answer is:

1. Ryan Jeffers
2. Devonte' Graham

Installation

To run without installing with uv, try uvx --from langchain-chain-of-verification cove --help.

pipx

This is the recommended installation method.

$ pipx install langchain-chain-of-verification

pip

$ pip install langchain-chain-of-verification

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

langchain_chain_of_verification-0.1.6.tar.gz (69.7 kB view details)

Uploaded Source

Built Distribution

File details

Details for the file langchain_chain_of_verification-0.1.6.tar.gz.

File metadata

File hashes

Hashes for langchain_chain_of_verification-0.1.6.tar.gz
Algorithm Hash digest
SHA256 bf6630e1fc2dd4044f3edf54501a1cd61d5ee2833e341124c16996b2c1d382a3
MD5 22d10ebfda55518006cbf60255c9f77a
BLAKE2b-256 2d2e0888d8e5424deaca2c33ac2ef9d2ae3558a003a175406a019a1418193ed1

See more details on using hashes here.

File details

Details for the file langchain_chain_of_verification-0.1.6-py3-none-any.whl.

File metadata

File hashes

Hashes for langchain_chain_of_verification-0.1.6-py3-none-any.whl
Algorithm Hash digest
SHA256 24bd45cf67075eef46917b50235b1303f86da794d0d8c3a101c94756e3adc674
MD5 2d91acef75384c02d65a50f44b0496cc
BLAKE2b-256 a87da808f8195bed0c53375b955c7ab30c00f890eddadfc693193cfdc7a6bca5

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page