Temporal Documentation Helper

NOTE: This branch contains instructions for setting up the non-"Temporalized" version of the temporal-documentation-helper, switch to the temporal branch for the original version.

This Slack bot evaluates questions and provides responses if they can be found in the Temporal Technologies documentation.

Branches

• main: directly queries Ollama to classify questions and generate responses.
• temporal: relies on Temporal workflows to handle LLM processing.

Prerequisites

• Configure a Slack application in your workspace. Include the following permissions and event subscriptions:
  • Actions: chat:write
  • Information: channels:history, reactions:read
  • Events: message.channels, reaction_added
• Setup a free Ngrok account to expose the bot to the world.
• Download and install Ollama

Setup

• Clone repo
• Create a new virtualenv python -m ~/path/to/temporalio
• Activate virtualenv source bin ~/path/to/temporalio/bin/activate
• Install pre-requisites pip install -r requirements.txt
• Create a .env file (touch .env) in the root directory and populate the following fields:
  • SLACK_BOT_TOKEN
  • SLACK_SIGNING_SECRET
  • SLACK_CLIENT_ID
  • SLACK_CLIENT_SECRET
  • SLACK_REDIRECT_URI
  • SLACK_BOT_USER_ID
• Generate the Chroma vectorstore python scripts/build_temporal_index.py
• Install the Temporal CLI brew install temporal

Running

• Start the bot uvicorn main:app --reload --host 0.0.0.0 --port 8000

• Expose it to the world ngrok http --url=${NGROK_URL}$ 8000

• Start Ollama ollama run llama3

Usage

Navigate to the primary channel the bot is configured to listen on and submit a random question, then a question about the Temporal platform.

• Unrelated questions should be ignored
• Questions about Temporal should receive a reply with a response referencing the documentation.