Soothe Wiki

Goal-driven orchestration framework for building 24/7 long-running autonomous agents

Welcome to the Soothe documentation. This wiki provides comprehensive guides for users, developers, and operators.

🚀 Quick Start

New to Soothe? Start with Quick Start — install CLI, start the daemon (Docker or local), run your first prompt. Then browse Getting Started for installation details and core concepts.

📖 Documentation Index

🚀 Getting Started

🤖 Core Capabilities

  • Autonomous Mode - Multi-step autonomous task execution
  • Subagents - Specialized subagents (planner, deep_research, academic_research, veritas, etc.)
  • Thread Management - Conversation threads and session resumption

🔧 Configuration & Management

🚀 Deployment & Operations

🏗️ Architecture & Core Modules

🔧 Backend Implementations

🛠️ Troubleshooting

📚 Reference & Community

🏗️ Architecture

Three-Level Execution Model 

┌─────────────────────────────────────────────────────────────┐
│ ContextEngine: Autonomous Goal Management (RFC-624)        │
│ • Manages goal DAGs, delegates single goals to StrangeLoop   │
│ • Loop: Goal/Goals → PLAN → PERFORM → REFLECT → Update     │
└─────────────────────────────────────────────────────────────┘
                          ↓ PERFORM (full delegation)
┌─────────────────────────────────────────────────────────────┐
│ StrangeLoop: Agentic Goal Execution (RFC-201)                 │
│ • Executes single goals through Plan → Execute iterations   │
│ • Loop: Plan → Execute (max ~8 iterations)                 │
└─────────────────────────────────────────────────────────────┘
                          ↓ EXECUTE (step execution)
┌─────────────────────────────────────────────────────────────┐
│ CoreAgent: Runtime (RFC-100)                                │
│ • Model → Tools → Model loop (LangGraph native)            │
│ • Foundation: create_soothe_agent() → CompiledStateGraph    │
└─────────────────────────────────────────────────────────────┘

Key Principles:

  • Protocol-first design - All modules defined as protocols with pluggable implementations
  • Durable by default - Agent state persists and recovers from crashes
  • Plan-driven execution - Complex goals decomposed into plans with steps
  • Least-privilege delegation - Fine-grained permissions for tools and subagents
Learn more: Architecture Overview RFC-000

Framework Stack 

+------------------------------------------------------+
|  Soothe (orchestration framework)                    |
|  - ContextEngine: Autonomous Goal Management         |
|  - StrangeLoop: Agentic Goal Execution                 |
|  - CoreAgent: Runtime                                |
|  - MemoryProtocol, PlannerProtocol,                 |
|    PolicyProtocol, DurabilityProtocol               |
+------------------------------------------------------+
|  deepagents (agent framework)                        |
|  - BackendProtocol, AgentMiddleware,                 |
|    SubAgent/CompiledSubAgent, SummarizationMiddleware|
+------------------------------------------------------+
|  langchain / langgraph (runtime layer)               |
|  - BaseChatModel, BaseTool, StateGraph,              |
|    Checkpointer, BaseStore, RemoteGraph              |
+------------------------------------------------------+

Plan → Execute Loop 

User Query
    ↓
PLAN (LLM plans, assesses progress, decides steps)
    ↓
EXECUTE (execute tools, collect results)
    ↓
Assess Progress → More steps needed? → PLAN
                  ↓
              Complete → Return Result

Benefits:

  • Automatic strategy adjustment
  • Structured tool outputs for reliable evaluation
  • Sub-second responses for simple queries
  • Intelligent iteration for complex tasks

🔌 Plugin System

Extend Soothe with custom tools and subagents:

from soothe_sdk.plugin import plugin, tool

@plugin(name="my-plugin", version="1.0.0")
class MyPlugin:
    @tool(name="my_tool", description="Custom tool")
    def my_tool(self, arg: str) -> str:
        return f"Result: {arg}"

Learn more: RFC-600: Plugin Extension System

📚 Additional Resources

🆘 Getting Help

  • Documentation: You’re here! Browse the guides above.
  • Troubleshooting: See Troubleshooting Guide
  • Issues: Report bugs on GitHub Issues
  • Community: Join discussions on GitHub Discussions

Feature Status

Feature Status Documentation
Intelligent Execution Loop ✅ Production Ready RFC-201
Research Subagent ✅ Production Ready Subagents Guide
Plugin System ✅ Production Ready RFC-600
Multi-Transport Daemon ✅ Production Ready Multi-Transport Setup
Thread Management ✅ Production Ready Thread Management
Security Policies ✅ Production Ready RFC-102
Autonomous Mode 🚧 Experimental Autonomous Mode

Additional Resources

📖 Extended Documentation

🔗 External Links

Getting Help

Common Issues

Community

  • Report issues: GitHub Issues
  • Ask questions: Use GitHub Discussions or check the Troubleshooting guide

Contributing

Interested in contributing to Soothe? See:

Table of contents