Arti

A serverless AI Slack bot service using Embedchain deployed to AWS Lambda using Pulumi.

Load and interrogate your data using an Artificial Intelligence RAG microservice built on Embedchain, providing CLI, REST API and Slack interfaces, with an option to deploy to AWS Lambda using Pulumi.

Optional configuration for various data sources, LLMs, vector databases, embedding models, and evaluation.

Usage

Requirements

• Docker
• Python 3

Prerequisites

cp -R /path/to/my/assets ./assets
cp sample.env .env

Quickstart

Ensure you have required dependencies installed and have followed the prerequisite steps.

Populate the .env file with, at minimum, the OPENAI_API_KEY is required for arti to answer questions.

docker compose run --rm arti ask "What do these files contain?"

Manual installation

python3 -m venv venv
. venv/bin/activate
make install
arti ask "What do these files contain?"

Configuration

Configuration is found in environment variables, documented in the table below.

Local development

Requirements

• Docker
• Python 3

Running the application locally

1. Copy sample.env to .env, then edit .env to replace the sample with your desired settings:

   cp sample.env .env

   Variable	Description
   AWS_REGION	AWS region
   OPENAI_API_KEY	OpenAI Key (takes precedence over secret)
   OPENAI_API_KEY_SECRET_NAME	OpenAI Key secret name
   PINECONE_API_KEY	Pinecone Key secret (takes precedence over secret)
   PINECONE_API_KEY_SECRET_NAME	Pinecone Key secret name
   SLACK_BOT_TOKEN	Slack bot token (takes precedence over secret)
   SLACK_BOT_TOKEN_SECRET_NAME	Slack bot token secret name
   SLACK_BOT_SIGNING_SECRET	Slack bot signing secret
   LOG_LEVEL	Log level

2. Create a Python virtual environment and activate it (first run only)

   make venv
   . venv/bin/activate

3. Configure the project for development and install dependencies

   make develop

4. Populate the dataset in ./assets with files such as PDFs, Docx, CSV, HTML, Text and more

5. Run the application

   arti arti ask "what can you tell me about fruits and vegetables?"
   # alternatively, run everything using docker compose
   docker compose run --rm arti ask "what can you tell me about fruits and vegetables?"

6. To stop, [CTRL]-C the application

Pre-commit

This project implements pre-commit to manage git hooks. This isn't required, but it will help you catch any issues before you push your commits.

Install pre-commit on MacOS using Homebrew:

brew install pre-commit

Once you have pre-commit installed, install the git hook scripts:

pre-commit install

Slack integration

The slack integration uses Bolt.

Follow their instructions to create a new Slack app.

For local usage, populate your .env file with the appropriate Slack tokens.

Minimum scopes

For reference, these are approximately the expected scopes, depending on your use of the bot.

OAuth & permissions

• chat:write
• channels:read
• commands
• im:read
• im:write
• users:read
• users:write

Event subscriptions

• message.channels
• message.im

Features

Real-time data-loading into the vector database via an S3 file upload bridge

Model configuration

Citations

CLI

A cli entrypoint is added as an example. By default, the assets directory will be loaded into the vector database for search.

Say for example, I had a document containing information about fruits and vegetables:

arti ask "what can you tell me about fruits and vegetables?" 
arti -h

Makefile

A Makefile is provided to ease some common tasks, such as linting and deploying.

To see usage instructions:

make help

Deployment

Deployment happens through the Makefile for convenience. The stack configuration can be found in deploy/pulumi.

Prerequisites

Create the following keys in Secrets Manager. These names are configurable in deploy/pulumi/Pulumi.<stack>.yaml. Slack and Pinecone are only necessary if they are configured for use.

Secret Name	Schema
/catmeme/cloud-platform/sandbox/arti/access-token/openai	{ "apiKey": "" }
/catmeme/cloud-platform/sandbox/arti/access-token/pinecone	{ "apiKey": "" }
/catmeme/cloud-platform/sandbox/arti/access-token/slack	{ "apiKey": "", "signingSecret": "" }

Ensure you have a matching AWS profile name to the one in the stack. The Makefile assumes <environment>-deployment

make deploy DEPLOY_ENVIRONMENT=dev

The above example would expect a dev-deployment AWS profile configured.