Langchain YouTube audio loader duplicating transcripts

Kubululo commented 1 month ago

Checked other resources

[X] I added a very descriptive title to this issue.
[X] I searched the LangChain documentation with the integrated search.
[X] I used the GitHub search to find a similar question and didn't find it.
[X] I am sure that this is a bug in LangChain rather than my code.
[X] The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package).

Example Code

youtube_url = 'https://youtu.be/RXQ5AtjUMAw'

loader = GenericLoader(
  YoutubeAudioLoader(
      [youtube_url],
      './videos'
  ),
  OpenAIWhisperParser(
      api_key=key,
      language='en'
  )
)

loader.load()

Error Message and Stack Trace (if applicable)

Transcribing part 1!
Transcribing part 2!
Transcribing part 1!
Transcribing part 1!
Transcribing part 3!
Transcribing part 3!

Description

I'm using Langchain to generate transcripts of YouTube videos, but I've noticed that the usage on my api_key is high. After closer examination, I discovered that the OpenAIWhisperParser is Transcribing the same part multiple times

Sometimes it goes through parts like 1,2,3 and then returns to 1 and repeats
I've noticed that even with specified language, the first chunk is always in the original language as if the parameter is not passed to the first request
I've tried not using language argument but the issue was still there

System Info

System info: Python 3.11.9 inside PyCharm venv

langchain==0.2.1 langchain-community==0.2.1 langchain-core==0.2.1 langchain-openai==0.1.7 langchain-text-splitters==0.2.0 langgraph==0.0.55 langsmith==0.1.63

pow3rpi commented 1 month ago

Hi there,

I've reproduced your code with the same version of python and the same versions of libraries in PyCharm. Actually, I haven't faced any issues like you have.

Here is my full code:

from langchain_community.document_loaders.blob_loaders.youtube_audio import (
    YoutubeAudioLoader,
)
from langchain_community.document_loaders.generic import GenericLoader
from langchain_community.document_loaders.parsers import OpenAIWhisperParser

OPENAI_API_KEY = '...'

youtube_url = 'https://youtu.be/RXQ5AtjUMAw'

loader = GenericLoader(
    YoutubeAudioLoader(
        [youtube_url],
        './videos'
    ),
    OpenAIWhisperParser(
        api_key=OPENAI_API_KEY,
        language='fr'
    )
)

docs = loader.load()
print(docs)

Here is my full list of dependencies as we additionally need to install pydub and yt_dlp:

aiohttp==3.9.5
aiosignal==1.3.1
annotated-types==0.7.0
anyio==4.4.0
attrs==23.2.0
Brotli==1.1.0
certifi==2024.6.2
charset-normalizer==3.3.2
dataclasses-json==0.6.7
distro==1.9.0
frozenlist==1.4.1
greenlet==3.0.3
h11==0.14.0
httpcore==1.0.5
httpx==0.27.0
idna==3.7
jsonpatch==1.33
jsonpointer==3.0.0
langchain==0.2.1
langchain-community==0.2.1
langchain-core==0.2.1
langchain-openai==0.1.7
langchain-text-splitters==0.2.0
langgraph==0.0.55
langsmith==0.1.63
marshmallow==3.21.3
multidict==6.0.5
mutagen==1.47.0
mypy-extensions==1.0.0
numpy==1.26.4
openai==1.34.0
orjson==3.10.5
packaging==23.2
pycryptodomex==3.20.0
pydantic==2.7.4
pydantic_core==2.18.4
pydub==0.25.1
PyYAML==6.0.1
regex==2024.5.15
requests==2.32.3
sniffio==1.3.1
SQLAlchemy==2.0.30
tenacity==8.3.0
tiktoken==0.7.0
tqdm==4.66.4
typing-inspect==0.9.0
typing_extensions==4.12.2
urllib3==2.2.1
uuid6==2024.1.12
websockets==12.0
yarl==1.9.4
yt-dlp==2024.5.27

Also, I'd like to mention that I tried to run the code several times and use different values for language parameter, so every time I get the same output: output

Kubululo commented 4 weeks ago

What I've observed is that this only happens with videos longer than 60 minutes so maybe that can help with diagnosis

langchain-ai / langchain