v0.1.0

First Release

Building blocks for AI agent swarms in Go.

LLM Integration

• llm - Multi-provider abstraction (Anthropic, OpenAI, Google, Ollama, Groq, Mistral, xAI)

Communication

• bus - Message bus with pub/sub, queue groups, request/reply (NATS + memory backends)
• transport - JSON-RPC transports (stdio, WebSocket, SSE)

Swarm Coordination

• registry - Agent registration and capability-based discovery
• heartbeat - Liveness detection with death callbacks
• state - Distributed key-value store with locks

Task Management

• tasks - Idempotent task handling with deduplication and retries
• results - Task result publication with subscriptions
• ratelimit - Coordinated rate limiting across swarm

Operations

• shutdown - Graceful shutdown with phases and signal handling
• errors - Structured error taxonomy with retry semantics
• logging - Structured real-time logging
• telemetry - OpenTelemetry tracing

Integrations

• mcp - Model Context Protocol client for external tools
• acp - Agent Client Protocol for editor integration
• memory - Semantic memory with BM25 search (FIL model)

Utilities

• credentials - Secure credential management (TOML-based)
• policy - Security policy enforcement
• security - Taint tracking and content verification
• tools - Tool registry with MCP integration