core/input_router.py

from pathlib import Path
from urllib.parse import urlparse

from .file_analyzer import analyze_file, get_supported_extensions

MAX_ROUTED_TEXT_CHARACTERS = 400000


def process_input(input_data, multimodal_manager):
    raw_input = str(input_data).strip()
    if not raw_input:
        raise ValueError("Please enter some input.")

    if is_youtube_url(raw_input):
        transcript = multimodal_manager.extract_youtube_transcript(raw_input)
        return format_routed_text(
            "youtube",
            transcript,
            raw_input,
        )

    if looks_like_file_path(raw_input):
        file_path = Path(raw_input)
        file_data = analyze_file(file_path, multimodal_manager)
        return format_routed_text(
            file_data["kind"],
            file_data["content"],
            file_data["name"],
        )

    return raw_input


def looks_like_file_path(raw_input):
    file_path = Path(raw_input)
    if file_path.exists():
        return True
    if file_path.suffix.lower() in get_supported_extensions():
        return True
    if "\\" in raw_input or "/" in raw_input:
        return True
    if len(raw_input) > 2 and raw_input[1] == ":":
        return True
    return False


def is_youtube_url(raw_input):
    parsed_url = urlparse(raw_input)
    host = parsed_url.netloc.lower()
    return host in {
        "youtube.com",
        "www.youtube.com",
        "m.youtube.com",
        "youtu.be",
        "www.youtu.be",
    }


def format_routed_text(input_kind, content, source_name):
    clean_content = content.strip()
    if not clean_content:
        raise ValueError("No readable content was found for this input.")
    if len(clean_content) > MAX_ROUTED_TEXT_CHARACTERS:
        clean_content = clean_content[:MAX_ROUTED_TEXT_CHARACTERS].strip()

    if input_kind == "pdf":
        return f"Source: PDF file ({source_name})\n\n{clean_content}"
    if input_kind == "image":
        return f"Source: Image file ({source_name})\n\n{clean_content}"
    if input_kind == "document":
        return f"Source: Document file ({source_name})\n\n{clean_content}"
    if input_kind == "media":
        return f"Source: Audio or video file ({source_name})\n\n{clean_content}"
    if input_kind == "youtube":
        return f"Source: YouTube video ({source_name})\n\n{clean_content}"
    return clean_content