ChittyBot

Simple context aware chat bot that can carry a conversation with a human. Built with LangChain, Streamlit and AWS Bedrock.

Setup

Dev Environment

1. Setup Python 3.11 virtual environment. I used Anaconda to create mine. Install required dependences afterwards.

   conda create -n py3.11 python=3.11
   
   pip install streamlit
   pip install langchain
   pip install python-dotenv
   pip install boto3>=1.33.13

2. Create .env file and add AWS environment variables required by boto3

   AWS_DEFAULT_PROFILE=<your-aws-profile>
   AWS_DEFAULT_REGION=<your-aws-region>

3. Setup VS Code Debugging

   {
     "version": "0.2.0",
     "configurations": [
       {
         "name": "Debug Streamlit App",
         "type": "python",
         "request": "launch",
         "program": "/path/to/streamlit/executable/py3.11/bin/streamlit",
         "args": [
           "run",
           "<my_app_code>.py"
         ],
         "console": "integratedTerminal",
         "justMyCode": true
       }
     ]
   }

AWS Credentials

The following are required before we can successfully invoke a Bedrock model:

• Ensure you have requested access to the models
• IAM role granted with permissions to invoke specific Bedrock models
• Add trust policy to allow our IAM user to assume the IAM role

Steps

1. If you do not have an IAM role yet, create one and add the following policies. We will assume this role with our AWS identity to invoke Bedrock models. Model IDs can be found in the AWS Docs.

   {
       "Version": "2012-10-17",
       "Statement": [
   				{
               "Sid": "BedrockInvokeModel",
               "Effect": "Allow",
               "Action": [
                   "bedrock:InvokeModel",
                   "bedrock:InvokeModelWithResponseStream"
               ],
               "Resource": [
                   "arn:aws:bedrock:<aws-region>::foundation-model/anthropic.claude-v2",
                   "arn:aws:bedrock:<aws-region>::foundation-model/amazon.titan-embed-text-v1"
               ]
           },
           {
               "Sid": "ReadPermsAndXRay",
               "Effect": "Allow",
               "Action": [
                   "bedrock:ListFoundationModels",
                   "xray:PutTelemetryRecords",
                   "xray:PutTraceSegments"
               ],
               "Resource": "*"
           }
       ]
   }

2. Add a trust policy to the role you just created to allow your user to assume that role

   {
       "Version": "2012-10-17",
       "Statement": [
           {
               "Effect": "Allow",
               "Principal": {
                   "Service": "lambda.amazonaws.com",
                   "AWS": "arn:aws:iam::<aws-acct-no>:user/<aws-user-for-your-profile>"
               },
               "Action": "sts:AssumeRole"
           }
       ]
   }

3. Configure role we will assume locally. Add the following to your ~/.aws/config file. boto3 which LangChain uses to connect to AWS, will pick up this configuration and assume the correct identity.

   [profile <whatever-profile-name-you-like]
   role_arn=arn:aws:iam::<aws-account-number>:role/<iam-role-to-assume>
   source_profile=<your-aws-profile> # must be a valid aws profile from ~/.aws/credentials

Run Streamlit Locally

To run the web application, execute: streamlit run langchain_streamlit_awsbr.py. This will automatically open a page in your browser and direct you to http://localhost:8502.

To kill/stop the app, simply type CTRL + C in the terminal.