first commit

Author: debnsuma

.gitignore

@@ -1,5 +1,4 @@
 **/node_modules
 **/rs-env
 **/bb-env
-**/
 llm-rag-vectordb-python/image-generation-node-js-app/src/frontend/node_modules/*

building-bonds/.gitignore

@@ -0,0 +1,4 @@
+**/.env
+**/bb-env/
+**/.DS_Store
+**/.DS_Store/

building-bonds/README.md

@@ -0,0 +1,69 @@
+# Building Bonds: The Power of Ice-Breakers
+
+![Application Banner](/llm-rag-vectordb-python/building-bonds/boundbuilding.gif)
+
+Welcome to **Building Bonds**, a Streamlit application that harnesses the strengths of Amazon Bedrock and LangChain. Make your introductions more memorable! Enter a name, and let our application search for their LinkedIn profile, then provide you with a concise summary and ice-breaking facts about that person.
+
+## Features
+
+1. **Instant LinkedIn Search**: Just provide a name, and the application will try to locate their LinkedIn profile from the internet.
+2. **Automated Summary**: With the capabilities of Amazon Bedrock and LangChain, receive a detailed overview of the person's career and accomplishments.
+3. **Ice-Breaker Facts**: Start your conversation with a bang! Learn unique and engaging facts related to the individual.
+
+## How It Works
+
+The magic behind **Building Bonds**:
+
+- **Amazon Bedrock**: Empowers our system to deep dive into data and bring out meaningful insights.
+- **LangChain**: Assists with linguistic processing, allowing the app to draw a clear and engaging summary from LinkedIn details.
+
+## Getting Started
+
+### **1. Pre-requisites** 
+
+- Clone the repository to your local machine.
+- Create a `.env` file in the project directory using `env.example` as a reference. Populate the `.env` file with your Proxycurl and Serpa API Key details:
+
+    ```bash
+    PROXYCURL_API_KEY=<YOUR API KEY>
+    SERPAPI_API_KEY=<YOUR API KEY>
+    ```
+
+### **2. Setting Up a Virtual Environment** 
+
+Use `virtualenv` to create an isolated Python environment:
+
+1. Install `virtualenv`:
+    ```bash
+    pip install virtualenv
+    ```
+
+2. Navigate to the directory where you cloned the repository.
+    
+3. Initialize the virtual environment:
+    ```bash
+    virtualenv bb-env
+    ```
+
+4. Activate the environment:
+    ```bash
+    source bb-env/bin/activate 
+    ```
+
+### **3. Installing Dependencies**
+
+With your virtual environment active, install the necessary packages:
+
+```bash
+pip install -r requirements.txt
+```
+
+This command installs all dependencies from the `requirements.txt` file into your `rs-env` environment.
+
+### **4. Usage**
+
+Launch the application using Streamlit:
+
+   ```bash
+   streamlit run app.py
+   ```

building-bonds/agents/__init__.py

@@ -0,0 +1,46 @@
+from tools.tools import get_profile_url
+from dotenv import load_dotenv
+from langchain import PromptTemplate
+from langchain.llms.bedrock import Bedrock
+from langchain.agents import initialize_agent, Tool, AgentType
+import re 
+
+
+
+
+def get_llm():
+
+    bedrock_llm = Bedrock(model_id="anthropic.claude-v2",
+                           model_kwargs={"temperature": 0.1,"max_tokens_to_sample": 4096})
+
+    return bedrock_llm
+
+def lookup(name: str) -> str:
+    load_dotenv()
+
+    template = """given the full name {name_of_person} I want you to get it me a link to their Linkedin profile page.
+                  Your answer should contain only a URL of the LinkedIN profile"""
+
+    tools_for_agent = [
+        Tool(
+            name="Crawl Google 4 linkedin profile page",
+            func=get_profile_url,
+            description="useful for when you need get the Linkedin Page URL",
+        ),
+    ]
+    llm = get_llm()
+    agent_chain = initialize_agent(tools_for_agent, llm, agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION, verbose=True)
+
+    prompt_template = PromptTemplate(
+        input_variables=["name_of_person"], template=template
+    )
+
+    # Get the LLM's output
+    try:
+        linkedin_username = agent_chain.run(handle_parsing_errors=True, input=prompt_template.format_prompt(name_of_person=name))
+
+    except ValueError as e:
+        print("Error while parsing LLM output:", e)
+        return None
+    
+    return linkedin_username

building-bonds/agents/__pycache__/__init__.cpython-310.pyc

@@ -0,0 +1,76 @@
+import streamlit as st
+
+from langchain.prompts import PromptTemplate
+from langchain.chains import LLMChain
+from third_parties.linkedin import scrape_linkedin_profile
+from agents.linkedin_lookup_agent import lookup as linedin_lookup_agent
+from langchain.llms.bedrock import Bedrock
+
+
+def get_llm():
+    bedrock_llm = Bedrock(model_id="anthropic.claude-v2",
+                          model_kwargs={"temperature": 0.1, "max_tokens_to_sample": 4096})
+    return bedrock_llm
+
+
+def ice_break_with(name: str):
+    linkedin_profile_url = linedin_lookup_agent(name=name)
+    linkedin_data = scrape_linkedin_profile(linkedin_profile_url=linkedin_profile_url)
+
+    summary_template = """
+        Given the LinkedIN information {information} about a person from, I wish to create the following:
+            1. A short Summary
+            2. Two interesting facts about them
+    """
+
+    summary_prompt_template = PromptTemplate(
+        input_variables=["information"],
+        template=summary_template,
+    )
+
+    llm = get_llm()
+    chain = LLMChain(llm=llm, prompt=summary_prompt_template)
+
+    result = chain.run(information=linkedin_data)
+    return result
+
+
+def main():
+    
+    st.title('Building Bonds: The Power of Ice-Breakers 💼✨')
+    st.write('An app that uses Amazon Bedrock and LangChain to create summaries based on their social media profile. 🚀')
+
+    st.sidebar.header("🔎 Enter the person's details")
+    name = st.sidebar.text_input("Name (e.g., 'Andy Jassy Amazon'):")
+
+    if st.sidebar.button('Get Summary'):
+        with st.spinner('Fetching LinkedIn data and creating summary... 🔄'):
+            result = ice_break_with(name)
+        st.subheader(f'Summary and couple of interesting facts 📝')
+        st.write(result)
+        st.success('Summary generated successfully! 👍')
+
+        st.markdown(
+            "<h3 style='text-align: center; font-size: 20px;'> To know more about Amazon Bedrock, visit <a href='https://aws.amazon.com/bedrock/' target='_blank'>here</a> </h3>",
+            unsafe_allow_html=True
+        )
+    # Styling the Streamlit page
+    st.markdown("""
+        <style>
+            body {
+                color: #4f4f4f;
+                background-color: #F5F5F5;
+            }
+            .stButton>button {
+                color: #4f4f4f;
+                background-color: #FFD700;
+                border-radius: 30px;
+                padding: 10px 20px;
+                font-size: 1.2em;
+            }
+        </style>
+    """, unsafe_allow_html=True)
+
+
+if __name__ == "__main__":
+    main()

building-bonds/agents/__pycache__/linkedin_lookup_agent.cpython-310.pyc

@@ -0,0 +1,11 @@
+streamlit==1.24.0
+tiktoken==0.5.1
+unstructured==0.10.16
+langchain==0.0.309
+PyPDF2==3.0.1
+python-dotenv==1.0.0
+psycopg2-binary==2.9.6
+altair==4.0.0
+urllib3==1.26.6
+boto3==1.28.61
+google-search-results==2.4.2

building-bonds/agents/linkedin_lookup_agent.py

@@ -0,0 +1,39 @@
+import os
+import requests
+
+
+def scrape_linkedin_profile(linkedin_profile_url):
+    """scrape information from LinkedIn profiles,
+    Manually scrape the information from the LinkedIn profile"""
+    api_endpoint = "https://nubela.co/proxycurl/api/v2/linkedin"
+    header_dic = {"Authorization": f'Bearer {os.environ.get("PROXYCURL_API_KEY")}'}
+
+    # For production
+    response = requests.get(
+        api_endpoint, params={"url": linkedin_profile_url}, headers=header_dic
+    )
+
+    # # For test and development
+    # response = requests.get(
+    #     "https://gist.githubusercontent.com/debnsuma/07afaf3939dcc2b5cc404de58016fdd2/raw/9e4b8f942364f6ee9759d1cdb1b6d7b8078ceb4e/suman.json"
+    # )
+
+    data = response.json()
+    data = {
+        k: v
+        for k, v in data.items()
+        if v not in ([], "", "", None)
+        and k
+        not in [
+            "people_also_viewed",
+            "certifications",
+            "accomplishment_publications",
+            "accomplishment_honors_awards",
+            "accomplishment_projects",
+        ]
+    }
+    if data.get("groups"):
+        for group_dict in data.get("groups"):
+            group_dict.pop("profile_pic_url")
+
+    return data