This is "from-first-principles" code for performing chat flows with generative AI models from Amazon Bedrock. The code here is intended as an exploration of techniques to interact with Amazon Bedrock models, and code to help test other concepts such as agents, tools and function calling.
The repo includes Python class files that support chat, and some sample client implementations.
Currently this code supports Anthropic Claude, and Mistral AI models via Amazon Bedrock.
The main focus of this repo is the Chat() Python class which handles all(?) the logic required to create a chat based application. Supported: Text prompting, text and image prompting, text generation, and text streaming generation.
# For class files:
boto3==1.28.59
botocore==1.31.85
pydantic==2.6.3
pydantic_core==2.16.3
# For clients:
streamlit==1.32.0
streamlit-chat==0.1.1
- Clone repo
pip install -r requirements.txt
- To run the streamlit sample app:
- Ensure you have AWS credentials accessible in the environment.
streamlit run ./demo_ui_chat_stream_claude.py
or
streamlit run ./demo_ui_chat_stream_mistral.py
etc.
- 0.1
- Initial Release (with support for Claude)
- 0.2
- Added support for Mistral AI models.
This project is licensed under the MIT License - see the LICENSE.md file for details
This is a proof of concept. It is not intended to be used in production.