AI-Services

AI Services, part of the IBM Open-Source AI Foundation for Power, deliver pre-built AI capabilities and integration with inferencing solutions like Red Hat AI Inference Server. Optimized for IBM Spyre™ on Power, they enable fast deployment and support models such as LLMs, embeddings, and re-rankers—helping enterprises scale AI efficiently.

📺 Demo

RAG-AIServices-PowerSpyre.mp4

Quick Start

Installation

For detailed platform-specific installation instructions, see Installation Guide.

Quick install for your platform:

# Set version
VERSION="v0.2.0"

# macOS (Intel)
curl -LO "https://github.com/IBM/project-ai-services/releases/download/${VERSION}/ai-services-darwin-amd64"
chmod +x ai-services-darwin-amd64
sudo mv ai-services-darwin-amd64 /usr/local/bin/ai-services

# macOS (Apple Silicon)
curl -LO "https://github.com/IBM/project-ai-services/releases/download/${VERSION}/ai-services-darwin-arm64"
chmod +x ai-services-darwin-arm64
sudo mv ai-services-darwin-arm64 /usr/local/bin/ai-services

# Linux (x86_64)
curl -LO "https://github.com/IBM/project-ai-services/releases/download/${VERSION}/ai-services-linux-amd64"
chmod +x ai-services-linux-amd64
sudo mv ai-services-linux-amd64 /usr/local/bin/ai-services

# Linux (ppc64le/Power)
curl -LO "https://github.com/IBM/project-ai-services/releases/download/${VERSION}/ai-services-linux-ppc64le"
chmod +x ai-services-linux-ppc64le
sudo mv ai-services-linux-ppc64le /usr/local/bin/ai-services

Supported Platforms:

• macOS (Intel x86_64, Apple Silicon ARM64) - Client-only mode
• Linux (x86_64) - Client-only mode
• Linux (ppc64le/Power) - Local + Remote mode

Deployment Modes:

• Client-only mode: CLI connects to remote OpenShift cluster for application deployment
• Local + Remote mode: Supports both local Podman deployments and remote OpenShift cluster connections

For signature verification with Cosign, see the Installation Guide.

Run the binary to get started

% ai-services --help
A CLI tool for managing AI Services infrastructure.

Usage:
  ai-services [command]

Available Commands:
  application   Deploy and monitor the applications
  completion    Generate the autocompletion script for the specified shell
  help          Help about any command
  version       Prints CLI version with more info

Flags:
  -h, --help      help for ai-services
  -v, --version   version for ai-services

Use "ai-services [command] --help" for more information about a command.

---

Repository Structure

project-ai-services/
├── README.md          # Project documentation
├── ai-services/       # CLI tool for project-ai-services
│   ├── assets/        # Application template files
├── images/            # Container base images
│   └── service-base/   # Service base image for services
├── services/          # Microservices (backend)
│   ├── common/        # Shared library layer
│   ├── chatbot/       # RAG chatbot service
│   ├── digitize/      # Document ingestion service
│   ├── summarize/     # Summarization service
│   └── similarity/    # Similarity search service
├── ui/                # Frontend applications
│   ├── chatbot/       # Chatbot UI
│   ├── digitize/      # Digitize UI
│   └── catalog/       # Catalog UI
├── test/              # Test assets
│   └── golden/        # Golden dataset
└── docs/              # Documentation
    └── proposals/     # Architecture proposals

Service Architecture

The repository follows a microservices architecture with:

• Layered Container Images: service-base → services-common → individual services
• Independent Services: Each service has its own Containerfile, Makefile, and versioning
• Shared Common Layer: Common utilities and dependencies in services/common/
• Clean Boundaries: No cross-service dependencies, all shared code in common layer

Service Images:

• chatbot-service - RAG chatbot backend
• digitize-service - Document ingestion and processing
• summarize-service - Text summarization
• similarity-service - Semantic similarity search

UI Images:

• chatbot-ui - Chatbot web interface
• digitize-ui - Document upload interface
• catalog-ui - Service catalog interface