AI DIAL Documentation

Project Overview and Contribution Guide

• Contribution Guide

Architecture

• Architecture Overview

Quick Start

• Quick Start Guide

Refer to AI DIAL Chat Repository to learn how to launch AI DIAL Chat with default configurations.

Run AI DIAL Locally

• Launch AI DIAL Chat with an Azure model
• Launch AI DIAL Chat with a self-hosted model
• Launch AI DIAL Chat with a sample application
• Launch AI DIAL Chat with a sample addon

Helm Deployment

• AI DIAL Generic Installation Simple Guide
• GCP Installation
• Azure Installation
• AWS Installation

Configuration

• Refer to Configuration to see configuration guidelines for AI DIAL components.

AI DIAL Chat Application User Manual

• AI DIAL Chat User Manual

Other AI DIAL Project Open Source Repositories

Here is the current list of repositories where you can find more details.

You can also refer to the open source repository map on the AI DIAL website.

• DIAL Helm - helm chart, find stable assemblies here.
• DIAL Core - the main component that exposes API
• DIAL SDK - development kit for applications and model adapters
• DIAL Interceptors Python SDK - framework for creating DIAL Interceptors in Python for chat completion and embedding models.
• DIAL Chat - default UI
• DIAL Overlay - a library for using AI DIAL Chat in an overlay format
• DIAL Chat Themes - static content and UI customizations for default UI
• Visualizer Connector - a library for connecting custom visualizers
• DIAL CI - github CI commons
• DIAL Assistant - model agnostic assistant/addon implementation for DIAL. It allows to use self-hosted OpenAI plugins as DIAL addons.
• DIAL Analytics Realtime - simple real-time usage analytics. That transforms logs into InfluxDB metrics
• DIAL Auth Helper - AuthProxy is a proxy service that implements OpenID-compatible Web API endpoints to avoid direct interaction with the AuthProviders' APIs, such as the KeyCloak API. _ App Controller - a Java-based web service application that orchestrates the building and deployment of Python applications in Kubernetes.
• App Builder - a Python-based application designed to download source code from AI DIAL file storage and prepare files to build a container image.
• Python Code Interpreter - uses Jupiter Kernel to execute arbitrary python code.
• DIAL-to-DIAL Adapter - adapter for a local development against a remote AI DIAL Core.
• Model adapters:
  • DIAL Azure OpenAI Adapter - pluggable Azure ChatGPT adapter
  • DIAL GCP VertexAI Adapter - pluggable Google LLMs adapter
  • DIAL AWS Bedrock Adapter - pluggable Amazon LLMs adapter (Anthropic Claude 1/2 is included)
  • More LLM adapters will be released (you may contribute)