LLM OpenAI API Playground

Contents:

• Streamlit OpenAI Python App
• ChatGPT with PDF Data (Langchain)
• How to Build LLM Apps with Langchain

---

Streamlit OpenAI Python App

Exploring OpenAI via Python Packages (Streamlit, Langchain, OpenAI) and API Endpoint

• See:
  • Streamlit OpenAI Python App
  • OpenAI App Jupyter Notebook

Goals:

• ✅ create appropriate environments for Jupyter Notebook and app(s)
• ✅ create OpenAI interactive bot script using Jupyter Notebooks
• ✅ create simple Streamlit / Langchain / OpenAI App taking text and OpenAI App Key as input
• ✅ embed .env variable for openai_api_key and incorporate check for env versus user input in app
• add ability to upload document & ask questions on data
• build on app to incorporate career history:
  • pre-defined personal context: my resume, linkedin and github readme
  • set prompt for users to ask questions regarding my personal data

Resources:

• How to use ChatGPT API Python for Beginners - Full ChatBOT Tutorial
• Build an LLM Powered App with Langchain, Streamlit & OpenAI
• OpenAI Documentation - API Reference

Create a Python environment

Using Anaconda (for Jupyter Notebook):

1. Open the Anaconda PowerShell prompt.
2. Create a new environment with Python 3.9: conda create --name envpy39 python=3.9
3. Activate the environment: conda activate envpy39
4. Install required packages: pip install -q openai python-dotenv notebook

Using venv (for .py files):

1. Open a terminal or command prompt.
2. Navigate to your project directory: cd <path>
3. Create a new virtual environment: python -m venv venv
4. Activate the environment:
   • On Windows (PowerShell): .\venv\Scripts\Activate
   • On macOS/Linux (bash): source venv/bin/activate
   5. Install required packages: pip install streamlit openai langchain langchain[llms] or pip install -r requirements.txt or python -m pip install langchain streamlit openai
      • langchain llm troubleshooting • also this

Check versions of packages installed in environment:

• pip list --user (specific packages I installed)
• pip list (all packages installed)

Run the project

Jupyter Notebook:

1. Open the Anaconda PowerShell prompt.
2. Navigate to your project directory: cd <path>
3. Activate the environment: conda activate envpy39
4. Start Jupyter Notebook: jupyter notebook

.py files:

1. Open a terminal or command prompt.
2. Navigate to your project directory: cd <path>
3. Activate the environment:
   • On Windows (PowerShell): .\venv\Scripts\Activate
   • On macOS/Linux (bash): source venv/bin/activate
4. Run the server using Framework:
   • DRF: python manage.py runserver
   • Streamlit: streamlit run <projectname.py> view: http://localhost:8501/

Set your OpenAI API Key:

Don't have an OpenAI API Key? Get one here.

Save to your environment:

1. Create a file called: .env
2. Enter the following: OPENAI_API_KEY = <yourkeyhere>

Enter in the App:

1. Run the app
2. Enter the OpenAI API Key in the window to the left

---

ChatGPT with PDF Data (Langchain)

See:

• Jupyter Notebook: ChatGPT with PDF Data

Resources:

• Youtube

How it works:

• take pdf as input
• break into smaller pieces (chunks <=512 tokens)
  • allows receiving smaller responses based on queries
• take chunks and embed one by one (1536 dimension vectors ADA-002)
• put each chunk/embedding into vector db (FAISS) -> ready for recall when user queries
• allow user to query database
  • take in user query
  • put through embedding model
  • get back number of matched docs/embeddings
  • parse through LLM (GPT3.5)
  • send answer back to user

Note:

• need to upload documents to reports folder

---

How to Build LLM Apps with LangChain

See:

• LLM App with Langchain: Jupyter Notebook

Resources:

• How to Build LLM Applications with LangChain (Datacamp)