deploy at 2024-08-20 21:53:26.699758

Dockerfile

@@ -0,0 +1,14 @@
+FROM python:3.10
+RUN useradd -m -u 1000 user 
+USER user 
+ENV HOME=/home/user 	PATH=/home/user/.local/bin:$PATH
+WORKDIR /code
+COPY --link --chown=1000 . .
+ENV HOME=/home/user
+RUN mkdir -p /tmp/cache/
+RUN chmod a+rwx -R /tmp/cache/
+ENV HF_HUB_CACHE=HF_HOME
+RUN pip install --no-cache-dir -r requirements.txt
+
+ENV PYTHONUNBUFFERED=1 PORT=7860
+CMD ["python", "main.py"]

main.py

@@ -0,0 +1,284 @@
+from fasthtml_hf import setup_hf_backup
+from timelinestyle import TimelineStyle
+import os
+import json
+import pandas as pd
+import traceback
+from datetime import datetime
+from typing import Literal
+from pydantic_core import from_json
+from PyPDF2 import PdfReader
+from langchain_core.prompts import PromptTemplate
+from langchain.chains import LLMChain
+from langchain.output_parsers import PydanticOutputParser
+from langchain.chains.summarize import load_summarize_chain
+from langchain_openai import ChatOpenAI
+from langchain_anthropic import ChatAnthropic
+from pydantic import BaseModel, Field, ValidationError
+from langchain_openai import ChatOpenAI
+from fasthtml.common import * 
+from fasthtml.components import Svg
+from langchain_community.utilities.wikipedia import WikipediaAPIWrapper
+from langchain_community.tools.wikipedia.tool import WikipediaQueryRun
+
+# Set up the app, including daisyui and tailwind for the chat component
+tlink = Script(src="https://cdn.tailwindcss.com"),
+dlink = Link(rel="stylesheet", href="https://cdn.jsdelivr.net/npm/[email protected]/dist/full.min.css")
+
+assets_dir = "/Users/manaranjanp/Documents/Work/MyLearnings/fastHTML/llmtimeline/assets"
+
+app = FastHTML(hdrs=(tlink, dlink, picolink))
+
+svg = Svg(
+    Path(fill_rule='evenodd', d='M10 18a8 8 0 100-16 8 8 0 000 16zm3.857-9.809a.75.75 0 00-1.214-.882l-3.483 4.79-1.88-1.88a.75.75 0 10-1.06 1.061l2.5 2.5a.75.75 0 001.137-.089l4-5.5z', clip_rule='evenodd'),
+    xmlns='http://www.w3.org/2000/svg',
+    viewbox='0 0 20 20',
+    fill='currentColor',
+    cls='h-5 w-5'
+)
+
+print(type(svg))
+
+# Pydantic models
+class Event(BaseModel):
+    time: datetime = Field(description="When the event occurred")
+    description: str = Field(description="A summary of what happened. Not more than 20 words.")
+    sentiment: Literal["Positive", "Negative"] = Field(..., description="Categorization of the event sentiment")
+
+class EventResponse(BaseModel):
+    events: List[Event] = Field(max_length=20, description="List of events extracted from the context")
+
+# Set up the Pydantic output parser
+parser = PydanticOutputParser(pydantic_object=EventResponse)
+
+# LangChain prompt template with format instructions
+event_extraction_template = """
+Extract the time based informations or events from the context and return a list of events with time,  event description and event sentiment type whether it was positive or negative event.
+The context may contain information about people, organization or any other entity. Try to get detailed and unique list of events as possible.
+
+<context>
+{context}
+</context>
+
+The response must follow the following schema strictly. There will be penalty for not following the schema.
+
+<schema>
+{format_instructions}
+</schema>
+
+Output:
+"""
+
+event_prompt = PromptTemplate(
+    input_variables=["topic", "context"],
+    partial_variables={"format_instructions": parser.get_format_instructions()},
+    template=event_extraction_template
+)    
+
+# Function to get the appropriate language model based on user selection
+def getModel(model, key):
+    if(model == 'OpenAI'):
+        os.environ['OPENAI_API_KEY'] = key
+        return ChatOpenAI(temperature=0,  # Set to 0 for deterministic output
+                    model="gpt-4o-2024-08-06",  # Using the GPT-4 Turbo model
+                    max_tokens=8000)  # Limit the response length
+    else:
+        os.environ['ANTHROPIC_API_KEY'] = key
+        return ChatAnthropic(model='claude-3-5-sonnet-20240620')  # Limit the response length
+    
+
+# Function to generate an HTML table from the summary object
+#def generate_timeline_html(timeline):
+#    rows = []
+#    for idx, tline in timeline.iterrows(): 
+#        if(tline['Sentiment'] == "Positive"):           
+#            rows.append(Div(Div( H2(tline['Time']), P(tline['Event']), cls = 'content'), cls = "container left"))                
+#        else:
+#            rows.append(Div(Div( H2(tline['Time']), P(tline['Event']), cls = 'content'), cls = "container right"))                
+#
+#    return  Div(*rows, cls="timeline")
+
+
+# Function to generate an HTML table from the summary object
+def generate_timeline_html(timeline):
+    rows = []
+    
+    for idx, tline in timeline.iterrows(): 
+        if idx % 2 == 0:
+            rows.append(Li(Div(File("./assets/circle.svg"), cls = "timeline-middle"), 
+                        Div(Time(tline['TimeStr'], cls = "font-mono italic"), Div(tline['Event'], cls = 'text-lg font-black'), cls = "timeline-start mb-10 md:text-end"),                            
+                        Hr()))        
+        else:
+            rows.append(Li(Div(File("./assets/circle.svg"), cls = "timeline-middle"), 
+                        Div(Time(tline['TimeStr'], cls = "font-mono italic"), Div(tline['Event'], cls = 'text-lg font-black'), cls = "timeline-end mb-10"),                            
+                        Hr()))                                
+
+    # for idx, tline in timeline.iterrows(): 
+    #     if idx % 2 == 0:
+    #         rows.append(Li(Div(svg, cls = "timeline-middle"), 
+    #                     Div(Time(tline['TimeStr'], cls = "font-mono italic"), Div(tline['Event'], cls = 'text-lg font-black'), cls = "timeline-start mb-10 md:text-end"),                            
+    #                     Hr()))        
+    #     else:
+    #         rows.append(Li(Div(svg, cls = "timeline-middle"), 
+    #                     Div(Time(tline['TimeStr'], cls = "font-mono italic"), Div(tline['Event'], cls = 'text-lg font-black'), cls = "timeline-end mb-10"),                            
+    #                     Hr()))                                
+
+
+
+
+
+#    for idx, tline in timeline.iterrows(): 
+#        if idx % 2 == 0:
+#            rows.append(Li(#Div(Img(src="/assets/icons/circle.svg", cls="w-5 h-5"), cls = "timeline-middle"), 
+#                        Div(Time(tline['TimeStr'], cls = "font-mono italic"), Div(tline['Event'], cls = 'text-lg font-black'), cls = "timeline-start timeline-box"),                            
+#                        Hr()))        
+#        else:
+#            rows.append(Li(#Div(Img(src="/assets/icons/circle.svg", cls="w-5 h-5"), cls = "timeline-middle"), 
+#                        Div(Time(tline['TimeStr'], cls = "font-mono italic"), Div(tline['Event'], cls = 'text-lg font-black'), cls = "timeline-end timeline-box"),                            
+#                        Hr()))                                
+
+               
+    return  Ul(*rows, cls="timeline timeline-vertical")
+
+
+def get_timeline_df(result):
+    
+    results_data = []
+    # Parse the final result into GradedQAPair objects
+    try:        
+        if not isinstance(result, EventResponse):
+            raise ValueError(f"Expected a list, but got {type(result)}")
+        
+    except Exception as e:
+        print(f"An error occurred during analysis: {str(e)}")
+        raise
+    
+    except Exception as e:
+        print(f"An error occurred during analysis: {str(e)}")
+        raise     
+            
+    if isinstance(result, EventResponse):
+        # Create a list to hold the data for the DataFrame
+        
+        for event in result.events:
+            results_data.append({
+                'Time': event.time,
+                'Event': event.description,
+                'Sentiment': event.sentiment
+            })
+
+    df = pd.DataFrame(results_data)      
+    df = df.sort_values("Time", ascending = True).reset_index()
+    df['TimeStr'] = df['Time'].map(lambda x: x.strftime('%d/%m/%Y'))
+
+    return df  
+        
+# Placeholder function for Q&A generation
+def generate_timeline(topic, llm):
+    # This function will be implemented later
+    # For now, return a sample DataFrame
+    
+    wikipedia = WikipediaQueryRun(api_wrapper=WikipediaAPIWrapper())
+    wiki_content = wikipedia.run(topic)
+    
+    chain = event_prompt | llm | parser
+
+    result = chain.invoke({"context" : wiki_content})
+    
+    try:
+        # Parse the output using PydanticOutputParser
+        # response = parser.parse(result)
+        # Create the DataFrame
+
+        print(f"Results: {result}")
+
+ #       timeline = parser.parse(result)
+        
+        df = get_timeline_df(result)
+                
+        # Optionally, save the DataFrame to a CSV file
+        df.to_csv(f"timeline.csv", index=True)
+        print("Results saved to 'results.csv'")    
+    except Exception as e:
+        print(f"Error parsing LLM output: {str(e)}")
+        return None
+
+    return df
+
+# Function to generate the configuration form for the web interface
+def getConfigForm():
+    return Card(Form(hx_post="/submit", hx_target="#result", hx_swap_oob="innerHTML", hx_indicator="#indicator")(
+            Div(
+                Label(Strong("Model and Topic: "), style="color:#3498db; font-size:25px;")
+            ),
+            Div(
+                Span(Strong('Model: '), cls ="badge"),
+                Select(Option("OpenAI"), Option("Anthropic"), id="model", cls = 'select w-full max-w-xs')
+            ),
+            Div(
+                Span(Strong('Topic for timeline (Person/Organization/Event): '), cls ="badge"),
+                Input(id="secret", type="password", placeholder="Key: "),
+            ),
+            Div(
+                Span(Strong('Provide the topic.: '), cls ="badge"),
+                Input(type = 'text', 
+                      id="topic",
+                      cls = "input w-full max-w-xs",
+                      placeholder = "Type here")
+            ),
+            Div(                
+                Button("Generate Timeline", cls = 'btn')
+            ),
+            Div(              
+                Br(),                  
+                A("Developed by Manaranjan Pradhan", href="http://www.manaranjanp.com/", 
+                  target="_blank", 
+                  style = 'color: red; font-size: 16px;')                      
+            )))
+
+# Define the route for the homepage
+@app.get('/')
+def homepage():
+    return Titled('Generate a timeline ', Grid( getConfigForm(),
+        Div(
+            Div(id="result"),
+            Div(Label(Strong('Generating timeline for the topic.... take a deep breath....')),
+            Progress(), id="indicator", cls="htmx-indicator")
+        )
+        , style="grid-template-columns: 400px 1000px; gap: 50px;"
+    ))
+
+@app.get('/assets/{fname:path}.{ext}')
+async def get(fname: str, ext: str):
+    fpath:str = (assets_dir)+'/'+str(fname)+'.'+str(ext)
+    if os.path.isfile(fpath):
+        response = FileResponse(fpath, media_type="image/svg")
+        print("file sent:"+fpath)
+    else:
+        print("file failed:"+fpath)
+        response = HTTPException(status_code=404, detail="File not found")
+
+# Define the route for form submission
+@app.post('/submit')
+async def post(d:dict):
+    try:            
+        # Get the appropriate language model
+        model = getModel(d['model'], d['secret'])    
+        
+        # Perform one-pass summarization
+        timeline_df = generate_timeline(d['topic'], model)
+        #qas = pd.read_csv("results_tesla.csv")
+
+        timeline_df.head(10)
+        
+        # Generate and return the HTML table with the summaries
+        return generate_timeline_html(timeline_df)
+
+    except BaseException as e:
+        print(traceback.format_exc())
+        return str(e)
+
+setup_hf_backup(app)
+
+# Start the FastAPI server
+serve()

requirements.txt

@@ -0,0 +1,13 @@
+wikipedia
+pydantic==2.8.2
+python-fasthtml==0.4.4
+fasthtml-hf==0.1.4
+pandas
+pypdf
+PyPDF2
+langchain
+langchain-community
+langchain-core
+langchain-openai
+openai
+langchain_anthropic

timeline.csv

@@ -0,0 +1,21 @@
+,index,Time,Event,Sentiment,TimeStr
+0,6,1999-01-01 00:00:00+00:00,Dhoni made his first-class debut for Bihar.,Positive,01/01/1999
+1,7,2004-12-23 00:00:00+00:00,Dhoni debuted for the Indian cricket team in an ODI.,Positive,23/12/2004
+2,0,2007-09-24 00:00:00+00:00,India won the 2007 ICC World Twenty20 under Dhoni's captaincy.,Positive,24/09/2007
+3,18,2008-01-01 00:00:00+00:00,Dhoni received the Major Dhyan Chand Khel Ratna Award.,Positive,01/01/2008
+4,8,2008-01-01 00:00:00+00:00,Dhoni became captain of the Indian cricket team in all formats.,Positive,01/01/2008
+5,11,2010-04-25 00:00:00+00:00,CSK won the IPL under Dhoni's captaincy.,Positive,25/04/2010
+6,3,2010-06-24 00:00:00+00:00,India won the Asia Cup in 2010 under Dhoni's captaincy.,Positive,24/06/2010
+7,16,2010-09-26 00:00:00+00:00,CSK won the Champions League T20 under Dhoni.,Positive,26/09/2010
+8,19,2011-01-01 00:00:00+00:00,Dhoni received honorary rank of Lieutenant Colonel.,Positive,01/01/2011
+9,1,2011-04-02 00:00:00+00:00,India won the 2011 Cricket World Cup under Dhoni's leadership.,Positive,02/04/2011
+10,12,2011-05-28 00:00:00+00:00,CSK won the IPL again with Dhoni as captain.,Positive,28/05/2011
+11,2,2013-06-23 00:00:00+00:00,India won the 2013 ICC Champions Trophy with Dhoni as captain.,Positive,23/06/2013
+12,17,2014-10-04 00:00:00+00:00,CSK won the Champions League T20 again with Dhoni.,Positive,04/10/2014
+13,9,2014-12-30 00:00:00+00:00,Dhoni retired from Test cricket.,Negative,30/12/2014
+14,4,2016-03-06 00:00:00+00:00,India won the Asia Cup in 2016 with Dhoni as captain.,Positive,06/03/2016
+15,13,2018-05-27 00:00:00+00:00,CSK won the IPL for the third time under Dhoni.,Positive,27/05/2018
+16,5,2018-09-28 00:00:00+00:00,Dhoni was part of the 2018 Asia Cup winning squad.,Positive,28/09/2018
+17,10,2019-07-10 00:00:00+00:00,Dhoni retired from international limited-overs cricket.,Negative,10/07/2019
+18,14,2021-10-15 00:00:00+00:00,CSK won the IPL for the fourth time with Dhoni.,Positive,15/10/2021
+19,15,2023-05-29 00:00:00+00:00,CSK won the IPL for the fifth time under Dhoni.,Positive,29/05/2023

timelinestyle.py

@@ -0,0 +1,124 @@
+class TimelineStyle:
+    def __init__(self):
+        self.style = """
+        * {
+        box-sizing: border-box;
+        }
+
+        body {
+        background-color: #474e5d;
+        font-family: Helvetica, sans-serif;
+        }
+
+        .timeline {
+        position: relative;
+        max-width: 1200px;
+        margin: 0 auto;
+        }
+
+        .timeline::after {
+        content: '';
+        position: absolute;
+        width: 6px;
+        background-color: white;
+        top: 0;
+        bottom: 0;
+        left: 50%;
+        margin-left: -3px;
+        }
+
+        .container {
+        padding: 10px 40px;
+        position: relative;
+        background-color: inherit;
+        width: 50%;
+        }
+
+        .container::after {
+        content: '';
+        position: absolute;
+        width: 25px;
+        height: 25px;
+        right: -17px;
+        background-color: white;
+        border: 4px solid #FF9F55;
+        top: 15px;
+        border-radius: 50%;
+        z-index: 1;
+        }
+
+        .left {
+        left: 0;
+        }
+
+        .right {
+        left: 50%;
+        }
+
+        .left::before {
+        content: " ";
+        height: 0;
+        position: absolute;
+        top: 22px;
+        width: 0;
+        z-index: 1;
+        right: 30px;
+        border: medium solid white;
+        border-width: 10px 0 10px 10px;
+        border-color: transparent transparent transparent white;
+        }
+
+        .right::before {
+        content: " ";
+        height: 0;
+        position: absolute;
+        top: 22px;
+        width: 0;
+        z-index: 1;
+        left: 30px;
+        border: medium solid white;
+        border-width: 10px 10px 10px 0;
+        border-color: transparent white transparent transparent;
+        }
+
+        .right::after {
+        left: -16px;
+        }
+
+        .content {
+        padding: 20px 30px;
+        background-color: white;
+        position: relative;
+        border-radius: 6px;
+        }
+
+        @media screen and (max-width: 600px) {
+        /* Place the timelime to the left */
+        .timeline::after {
+        left: 31px;
+        }
+        
+        .container {
+        width: 100%;
+        padding-left: 70px;
+        padding-right: 25px;
+        }
+        
+        .container::before {
+        left: 60px;
+        border: medium solid white;
+        border-width: 10px 10px 10px 0;
+        border-color: transparent white transparent transparent;
+        }
+
+        .left::after, .right::after {
+        left: 15px;
+        }
+        
+        .right {
+        left: 0%;
+        }
+    }"""
+    
+    def get_style(self):        
+        return self.style