metagpt/utils/text.py

from typing import Generator, Sequence

from metagpt.utils.token_counter import TOKEN_MAX, count_output_tokens


def reduce_message_length(
    msgs: Generator[str, None, None],
    model_name: str,
    system_text: str,
    reserved: int = 0,
) -> str:
    """Reduce the length of concatenated message segments to fit within the maximum token size.

    Args:
        msgs: A generator of strings representing progressively shorter valid prompts.
        model_name: The name of the encoding to use. (e.g., "gpt-3.5-turbo")
        system_text: The system prompts.
        reserved: The number of reserved tokens.

    Returns:
        The concatenated message segments reduced to fit within the maximum token size.

    Raises:
        RuntimeError: If it fails to reduce the concatenated message length.
    """
    max_token = TOKEN_MAX.get(model_name, 2048) - count_output_tokens(system_text, model_name) - reserved
    for msg in msgs:
        if count_output_tokens(msg, model_name) < max_token or model_name not in TOKEN_MAX:
            return msg

    raise RuntimeError("fail to reduce message length")


def generate_prompt_chunk(
    text: str,
    prompt_template: str,
    model_name: str,
    system_text: str,
    reserved: int = 0,
) -> Generator[str, None, None]:
    """Split the text into chunks of a maximum token size.

    Args:
        text: The text to split.
        prompt_template: The template for the prompt, containing a single `{}` placeholder. For example, "### Reference\n{}".
        model_name: The name of the encoding to use. (e.g., "gpt-3.5-turbo")
        system_text: The system prompts.
        reserved: The number of reserved tokens.

    Yields:
        The chunk of text.
    """
    paragraphs = text.splitlines(keepends=True)
    current_token = 0
    current_lines = []

    reserved = reserved + count_output_tokens(prompt_template + system_text, model_name)
    # 100 is a magic number to ensure the maximum context length is not exceeded
    max_token = TOKEN_MAX.get(model_name, 2048) - reserved - 100

    while paragraphs:
        paragraph = paragraphs.pop(0)
        token = count_output_tokens(paragraph, model_name)
        if current_token + token <= max_token:
            current_lines.append(paragraph)
            current_token += token
        elif token > max_token:
            paragraphs = split_paragraph(paragraph) + paragraphs
            continue
        else:
            yield prompt_template.format("".join(current_lines))
            current_lines = [paragraph]
            current_token = token

    if current_lines:
        yield prompt_template.format("".join(current_lines))


def split_paragraph(paragraph: str, sep: str = ".,", count: int = 2) -> list[str]:
    """Split a paragraph into multiple parts.

    Args:
        paragraph: The paragraph to split.
        sep: The separator character.
        count: The number of parts to split the paragraph into.

    Returns:
        A list of split parts of the paragraph.
    """
    for i in sep:
        sentences = list(_split_text_with_ends(paragraph, i))
        if len(sentences) <= 1:
            continue
        ret = ["".join(j) for j in _split_by_count(sentences, count)]
        return ret
    return list(_split_by_count(paragraph, count))


def decode_unicode_escape(text: str) -> str:
    """Decode a text with unicode escape sequences.

    Args:
        text: The text to decode.

    Returns:
        The decoded text.
    """
    return text.encode("utf-8").decode("unicode_escape", "ignore")


def _split_by_count(lst: Sequence, count: int):
    avg = len(lst) // count
    remainder = len(lst) % count
    start = 0
    for i in range(count):
        end = start + avg + (1 if i < remainder else 0)
        yield lst[start:end]
        start = end


def _split_text_with_ends(text: str, sep: str = "."):
    parts = []
    for i in text:
        parts.append(i)
        if i == sep:
            yield "".join(parts)
            parts = []
    if parts:
        yield "".join(parts)