wvlet/scala-ai: Scala 3 for AI-Driven Coding & Agentic Systems

wvlet/scala-ai empowers Scala 3 for modern AI coding and building AI agent applications. It provides robust core utilities and tools for seamless LLM integration and agent orchestration.

This is achieved through two primary modules:

• ai-core: Foundational utilities adapted from wvlet/airframe for Scala 3, providing:

  • Logging: Performant, structured logging.
  • Object Design: Well-managed object lifecycle and configuration (using type introspection and DI concepts from airframe).
  • SerDe: Efficient serialization/deserialization of Scala case classes between common data exchange formats, JSON, MessagePack, etc.
  • RPC/HTTP Framework: For quickly building RPC/HTTP server and clients without boilerplate.
  • Testing Framework: A lightweight testing framework for Scala 3, inspired by airspec.
  • We also have a collection of prompts so that LLM can produce quality code.

• ai-agent: Tools for building AI agents, offering:

  • Unified LLM APIs: A consistent LLMAgent interface for AWS Bedrock, Google Vertex AI, and future LLM services.
  • Agent Orchestration: Components for advanced agent control flows.
  • Dynamic Tool Usage: Enabling agents to use external tools, including MCP (Model Context Protcotol) or other RCP servers built with ai-core.
  • Common Agent Patterns: Support for RAG (Retrieval Augmented Generation), vector search, and other common agent patterns.

Leveraging Scala 3's powerful type system and airframe's battle-tested components, scala-ai enables the development of efficient, type-safe, and reliable applications.

Note: This library is under active development. APIs may change.

Getting Started

Add the dependencies to your build.sbt:

// build.sbt
val scalaAIVersion = "LATEST_VERSION" // TODO: Replace with the actual latest version

libraryDependencies ++= Seq(
  // Core utilities for Scala 3 development
  "org.wvlet.ai" %% "ai-core" % scalaAIVersion,
  // High-level AI-agent interfaces (includes ai-core)
  "org.wvlet.ai" %% "ai-agent" % scalaAIVersion,
  // Optional: Choose AI service integrations
  "org.wvlet.ai" %% "ai-agent-bedrock" % scalaAIVersion,
  // "org.wvlet.ai" %% "ai-agent-vertexai" % scalaAIVersion,
)