Modules

Tramai is modular by design. Pick what you need.

Core

tramai-core

Core annotations, data models, provider SPI, structured-output contracts, and exceptions for the Tramai AI library.

tramai-engine

Runtime engine that turns annotated service interfaces into AI-backed proxies with retry, circuit-breaking, caching, and token budgeting.

Providers

tramai-anthropic

Provider for Anthropic's Messages API (Claude models).

tramai-azure-openai

Provider for Azure OpenAI — deployment-based endpoints with dual API key / Entra ID authentication.

tramai-bedrock

Provider for Amazon Bedrock using the InvokeModel API.

tramai-deepseek

Provider for DeepSeek's OpenAI-compatible chat completion API.

tramai-gemini

Provider for Google Gemini API — translates TramAI's unified message model to Gemini's endpoints.

tramai-ollama

Provider for locally-hosted Ollama models via the Ollama chat API.

tramai-openai

OpenAI-compatible model provider with support for OpenAI API, Together AI, Groq, vLLM, and any /chat/completions endpoint.

Infrastructure

tramai-bom

Bill of materials — a single-import Maven BOM that aligns versions of all Tramai publishable modules.

tramai-dashboard

Vue 3 admin UI for managing TramAI workflows, schedules, and workers.

tramai-mcp

Exposes registered Tramai workflows as MCP tools over stdio or SSE transport.

tramai-observability

Optional, opt-in OpenTelemetry integration — traces, metrics, and events for operations and workflows.

tramai-orchestration

Multi-step workflow engine with checkpoint/resume, distributed worker support, and a declarative DSL.

tramai-platform

Multi-tenant operational layer above the workflow server with API key authentication, rate limiting, audit logging, and a runtime plugin system.

tramai-scheduler

Time-based workflow triggers with cron expressions, delay wakeups, and a durable tick-claiming scheduler loop.

tramai-server

HTTP REST API surface for Tramai workflows — start, inspect, resume, cancel, and stream runs.

tramai-spring

Spring Boot auto-configuration that wires Tramai's runtime, scans @AiService interfaces, discovers @AiTool beans, and binds tramai.* application properties.

tramai-standalone

Minimal framework-free entry point that wires core, engine, and structured output into a single Tramai.create<T>() call.

tramai-structured

Generates JSON schemas from Kotlin types, extracts and validates structured objects from LLM responses.

tramai-testing

Mock providers, recording observers, and fluent assertions for deterministic Tramai integration tests.