The potential of AI to create truly intelligent chatbots has always fascinated me. Groq's specialized AI processors offered a unique opportunity to push these boundaries. This blog chronicles my journey building a chatbot using Groq's API, crafting a user-friendly interface with Streamlit, and deploying it using Ngrok and Google Colab, making it accessible to everyone.
- Utlize a wide range of LLMs, including Llama-3, Mixtral, etc.
- In-memory converstaion with the friendly chatbot that is profiled by a persona promptining technique.
- Utilize the powerful machine that Colab provides.
First and foremost, we obtain API keys from Groq and Ngrok:
- First, you clone the project to your local machine and install required libraries:
git clone https://github.com/What-s-behind/GroqChatbot-on-Colab.git
cd /GroqChatbot-on-Colab
pip install -r requirements.txt
- Setup
.envfile. To be honest, you have 2 choices: you set theGROQ_API_KEYso that your users do not need to, or you let them type their own api key. If you prefre the first choice, create a.envfile and put the key following this format.
# .env file
GROQ_API_KEY="GROQ_API_KEY"I leave an illustration in .env.example
- Run the app
streamlit run app.py
For those who want to use the app on Colab. You go to colab and upload the Jupyter notebook that you can see from the directory. Remember to initialize all necessary API keys from Colab's Secret zone.
Note: I've also repaired a simple HTML/CSS/Javascript demo at the /app. You can setup the API_KEY in the script.js file and run by starting the Live server in VSCode.
When you success establishing all steps. Here is the result
- Gmail: [email protected]
- LinkedIn: https://www.linkedin.com/in/minh-le-duc-a62863172/