File size: 3,032 Bytes
2df8973
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
import os
from youtube_transcript_api import YouTubeTranscriptApi, TranscriptsDisabled
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain_community.vectorstores import FAISS
from langchain_core.prompts import PromptTemplate
from langchain_huggingface import HuggingFaceEmbeddings
from langchain_groq import ChatGroq
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnableParallel, RunnablePassthrough, RunnableLambda

VECTORSTORE_DIR = "vectorstores"
EMBEDDING_MODEL_NAME = os.environ["EMBEDDING_MODEL"]

_embedding_model = HuggingFaceEmbeddings(model_name=EMBEDDING_MODEL_NAME)


def load_transcript(video_id: str) -> str:
    """Fetch and flatten a YouTube video's transcript."""
    try:
        yt_api = YouTubeTranscriptApi()
        transcript_list = yt_api.fetch(video_id, languages=["en"])
        transcript = " ".join(chunk.text for chunk in transcript_list)
        return transcript
    except TranscriptsDisabled:
        raise ValueError("No captions available for this video.")


def split_transcript(transcript: str):
    splitter = RecursiveCharacterTextSplitter(
        chunk_size=1000,
        chunk_overlap=200,
    )
    return splitter.create_documents([transcript])


def get_or_build_vectorstore(video_id: str):
    """Load a cached FAISS index for this video, or build + save one if missing."""
    path = os.path.join(VECTORSTORE_DIR, video_id)

    if os.path.exists(path):
        return FAISS.load_local(
            path,
            _embedding_model,
            allow_dangerous_deserialization=True,
        )

    transcript = load_transcript(video_id)
    chunks = split_transcript(transcript)
    vectorstore = FAISS.from_documents(chunks, _embedding_model)

    os.makedirs(VECTORSTORE_DIR, exist_ok=True)
    vectorstore.save_local(path)

    return vectorstore


def format_docs(retrieved_docs):
    return "\n\n".join(doc.page_content for doc in retrieved_docs)


def get_rag_chain(video_id: str):
    """Builds and returns the full RAG chain for a given YouTube video."""
    vectorstore = get_or_build_vectorstore(video_id)
    retriever = vectorstore.as_retriever(search_kwargs={"k": 2})

    llm = ChatGroq(
        model="llama-3.1-8b-instant",
        temperature=0,
        api_key=os.environ["GROQ_API_KEY"],
    )

    prompt = PromptTemplate(
        template="""

          You are a helpful assistant.

          Answer ONLY from the provided transcript context.

          If the context is insufficient, just say you don't know.



          {context}

          Question: {question}

        """,
        input_variables=["context", "question"],
    )

    parser = StrOutputParser()

    parallel_chain = RunnableParallel({
        "context": retriever | RunnableLambda(format_docs),
        "question": RunnablePassthrough(),
    })

    main_chain = parallel_chain | prompt | llm | parser
    return main_chain