Lota SDKAI Runtime Framework

Build production-grade agent-powered applications with persistent workstreams, long-term memory, multi-agent orchestration, and structured execution plans.

Get Started

API Reference

GitHub

🧵 Workstreams

Persistent conversation threads with full message history, cursor-paginated retrieval, real-time streaming, and structured state tracking. Supports direct 1:1, group, and core workstream modes.

🧠 Memory System

Durable fact extraction pipeline with HNSW vector search, BM25 full-text search, and LLM-based reranking. Memories are typed, scoped, and automatically consolidated over time.

🤖 Agent Orchestration

Define agent rosters with custom tools, skills, and behavioral rules. Fan out tasks to specialist teams in parallel with real-time streaming progress.

📋 Execution Plans

Structured task management with ordered steps, status transitions, event auditing, and plan replacement with carried context from previous work.

🔌 Plugin Architecture

Extend the runtime with typed plugins that contribute services, tools, database schemas, and environment variables. First-class support for integration plugins.

🌉 Bifrost Gateway

Route all LLM requests through a unified AI gateway with model-level configuration, provider routing, and centralized key management.