ABAO77's picture
Upload 60 files
172064c verified
raw
history blame
6.48 kB
from typing import TypedDict, Optional, List
from langchain_core.messages import AnyMessage, ToolMessage, HumanMessage, AIMessage
from langgraph.graph.message import add_messages
from typing import Sequence, Annotated
from langchain_core.messages import RemoveMessage
from langchain_core.documents import Document
from src.config.llm import get_llm
from src.utils.logger import logger
from src.utils.helper import extract_transcript, extract_comment
from .prompt import *
import operator
class State(TypedDict):
video_link: str
messages: Annotated[Sequence[AnyMessage], add_messages]
transcript: str
comment: str
script_structure_analyzer_response: str
comment_insight_extractor_response: str
research_insight_response: str
script_re_outline_response: str
script_writer_response: List[str]
target_word_count: int
script_count: int
current_script_index: int
def trim_history(state: State):
history = state.get("messages", [])
if len(history) > 20:
num_to_remove = len(history) - 20
remove_messages = [
RemoveMessage(id=history[i].id) for i in range(num_to_remove)
]
return {
"messages": remove_messages,
"selected_ids": [],
"selected_documents": [],
}
return {}
def extract_transcript_and_comment(state: State):
transcript = extract_transcript(state["video_link"])
comment = extract_comment(state["video_link"])
# Calculate script count based on target word count
# Assume each script is around 200-300 words
avg_words_per_script = 1000
script_count = max(1, state.get("target_word_count", 8000) // avg_words_per_script)
return {
"transcript": transcript,
"comment": comment,
"script_count": script_count,
"messages": HumanMessage(
content=f"Will generate {script_count} scripts for {state.get('target_word_count', 8000)} words target"
),
}
def script_structure_analyzer(state: State):
transcript = state["transcript"]
response = chain_script_structure_analyzer.invoke({"script": transcript})
return {
"script_structure_analyzer_response": response.content,
"messages": HumanMessage(
content="Script Structure Analyzer Response: " + response.content
),
}
def comment_insight_extractor(state: State):
response = chain_comment_insight_extractor.invoke(
{
"comment": state["comment"],
"script_structure_analyzer_response": state[
"script_structure_analyzer_response"
],
}
)
return {
"comment_insight_extractor_response": response.content,
"messages": HumanMessage(
content="Comment Insight Extractor Response: " + response.content
),
}
def scientific_fact_finder(state: State):
input_message = {}
input_message["messages"] = [
{
"role": "user",
"content": f"""Hãy tìm 3-5 nghiên cứu khoa học thực tế (PubMed, JAMA, Circulation, Nutrients…),
Tóm tắt số liệu, trích nguồn, gợi ý số liệu phù hợp cho từng đoạn trong script mới. Dựa trên các thông tin sau:
Script Structure Analyzer Response: {state["script_structure_analyzer_response"]}
Comment Insight Extractor Response: {state["comment_insight_extractor_response"]}
""",
}
]
response = scientific_fact_agent.invoke(input_message)
research_insight = response["messages"][-1].content
return {
"research_insight_response": research_insight,
"messages": HumanMessage(
content="Scientific Fact Finder Response: " + research_insight
),
}
def script_re_outline(state: State):
response = chain_script_re_outline.invoke({"messages": state["messages"]})
return {
"script_re_outline_response": response.content,
"messages": HumanMessage(
content="Script Re-Outline Response: " + response.content
),
}
def script_writer_init(state: State):
"""Initialize script writing process"""
return {
"script_writer_response": [],
"current_script_index": 0,
"messages": HumanMessage(content="Starting script generation process..."),
}
def script_writer_single(state: State):
"""Generate a single script"""
current_index = state.get("current_script_index", 0)
script_count = state.get("script_count", 10)
target_word_count = state.get("target_word_count", 8000)
words_per_script = target_word_count // script_count if script_count > 0 else 1000
# Get existing scripts
script_out = list(state.get("script_writer_response", []))
current_messages = list(state["messages"])
# Add word count guidance to the prompt
if current_index == 0:
word_prompt = f"Hãy viết script đầu tiên với khoảng {words_per_script} từ."
else:
word_prompt = f"ok, viết cho tôi phần tiếp theo, bám sát cấu trúc, khoảng {words_per_script} từ cho script này, các công thức tạo cảm xúc và đừng quên đối tượng khán giả là người Mỹ,giới tính nữ, trên 20 tuổi, bắt đầu, trình bày thành dạng câu văn liền mạch, dùng để làm văn nói cho video YouTube, không dùng icon"
current_messages.append(HumanMessage(content=word_prompt))
# Generate script
response = chain_script_writer.invoke({"messages": current_messages})
script_out.append(response.content)
# Add response to message history
current_messages.append(AIMessage(content=response.content))
return {
"script_writer_response": script_out,
"current_script_index": current_index + 1,
"messages": current_messages
+ [
HumanMessage(content=f"Script {current_index + 1}/{script_count} completed")
],
}
def should_continue_writing(state: State):
"""Check if we should continue writing more scripts"""
current_index = state.get("current_script_index", 0)
script_count = state.get("script_count", 10)
return (
"script_writer_single" if current_index < script_count else "script_writer_end"
)
def script_writer_end(state: State):
"""Finalize script writing"""
script_count = len(state.get("script_writer_response", []))
return {"messages": HumanMessage(content=f"All {script_count} scripts completed!")}