Spaces:
Sleeping
Sleeping
File size: 4,788 Bytes
946d35b 8039e4b 6c655a3 946d35b 708437f efe9a51 72a7f4f 96b0973 563ce7c 8039e4b 6874dac efe9a51 6874dac 8039e4b 6874dac 8039e4b 6874dac 8039e4b 6874dac 8039e4b 6874dac efe9a51 6c655a3 a0929ab 8039e4b 96b0973 8039e4b 6874dac 8039e4b 6874dac 946d35b 708437f 3c1150c 708437f 3c1150c 8039e4b 3c1150c 8039e4b 708437f 3c1150c 708437f 32131c3 708437f 3c1150c 708437f 563ce7c |
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 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 |
from langchain_core.messages import SystemMessage, ToolMessage, HumanMessage, FunctionMessage
from .tools import retrieve_tool
import base64
from PIL import Image
from io import BytesIO
from fastapi import UploadFile
from huggingface_hub import InferenceClient
from .prompts import story_to_prompt , final_story_prompt
import os
from langgraph.prebuilt import create_react_agent
import pandas as pd
from datasets import load_dataset
from src.genai.utils.models_loader import llm_gpt
def generate_final_story(final_state):
if 'preferred_topics' in final_state:
if len(final_state['preferred_topics'])>0:
template = final_story_prompt(final_state)
messages = [SystemMessage(content=template),
HumanMessage(content=f'''The idea of the video is:\n{final_state['idea']}\n '''),
FunctionMessage(content=f'''The business details is:\n{final_state['business_details']}\nThe data of influencers is:\n{final_state['retrievals'][-1]}''',name='final_story_tool')]
print('The message of final story:',messages)
react_agent=create_react_agent(
model=llm_gpt,
tools=[])
for message_chunk , metadata in react_agent.stream({'messages':messages},stream_mode='messages'):
yield message_chunk.content
else:
for chunk in final_state['stories'][-1]:
yield chunk
else:
template = final_story_prompt(final_state)
influencers_data = retrieve_tool(final_state)
messages = [SystemMessage(content=template),
FunctionMessage(content=f'''The business details is:\n{str(final_state)}\nThe data of influencers is:\n{influencers_data}''',name='final_story_tool')]
react_agent=create_react_agent(
model=llm_gpt,
tools=[])
for message_chunk , metadata in react_agent.stream({'messages':messages},stream_mode='messages'):
yield message_chunk.content
def encode_image_to_base64(uploaded_file: UploadFile) -> str:
return base64.b64encode(uploaded_file.file.read()).decode("utf-8")
# Convert base64 string to PIL image (optional for LangGraph processing)
def process_image(base64_str: str) -> Image.Image:
image_data = base64.b64decode(base64_str)
return Image.open(BytesIO(image_data))
def generate_prompt(final_story,business_details,refined_ideation):
print('************Entering prompt generator****************')
messages = [SystemMessage(content=story_to_prompt()),
HumanMessage(content=f'''The scene-by-scene video story is {final_story}'''),
FunctionMessage(content=f'''The business details is:\n{business_details}\nThe idea is{refined_ideation}''',name='prompt_generation_id')
]
prompt = llm_gpt.invoke(messages)
print('The prompt is:',prompt)
return prompt.content
def generate_image(final_story, business_details, refined_ideation):
prompt = generate_prompt(final_story, business_details, refined_ideation)
print('************Finished prompt generator****************')
client = InferenceClient(
provider="hf-inference",
api_key=os.environ.get('HUGGINGFACEHUB_ACCESS_TOKEN'),
)
print('************Finished calling generator****************')
# output is a PIL.Image object
image = client.text_to_image(
prompt,
model="black-forest-labs/FLUX.1-schnell",
)
print('*****************Image Created*******************')
# Convert image to BytesIO buffer
buffered = BytesIO()
image.save(buffered, format="PNG") # you can also use "JPEG" if preferred
buffered.seek(0)
# Encode to base64 string
img_base64 = base64.b64encode(buffered.read()).decode("utf-8")
print('*****************Image Encoded to Base64*******************')
return img_base64
def save_to_db(business_details):
dataset = load_dataset("subashdvorak/tiktok-agentic-story")['train']
# dataset = load_influencer_data()
df = pd.DataFrame(dataset)
# 2. Flatten all business detail values to a set of lowercase strings
all_values = set()
for v in business_details.values():
if isinstance(v, str):
all_values.add(v.lower())
elif isinstance(v, list):
all_values.update(map(str.lower, map(str, v)))
# 3. Match rows where ANY column contains ANY of the values
def row_matches(row):
return any(
str(cell).lower().find(val) != -1
for cell in row
for val in all_values
)
# 4. Apply row-wise matching
matched_df = df[df.apply(row_matches, axis=1)]
matched_df.to_csv('extracted_data.csv')
|