Home

GolemCore Bot

AI assistant framework with intelligent model routing, multi-LLM support, and autonomous execution capabilities

GolemCore Bot is an extensible AI assistant built with Spring Boot 4.0.2 (Java 17). It processes messages through an ordered pipeline of specialized systems, supports multiple LLM providers, and offers powerful tool integrations including MCP protocol support. Configuration is managed through a built-in web dashboard.

---

Key Features

Category	Highlights
Web Dashboard	Built-in UI for configuration, chat, and monitoring — no env vars needed for most settings
Intelligent Processing	4-tier model selection with per-user overrides, dynamic tier escalation, context overflow protection
Tools	12+ built-in tools (incl. IMAP/SMTP email), MCP protocol support, sandboxed execution, tool confirmation workflow
Advanced Capabilities	Plan mode (review-before-execute), auto mode (autonomous goals), skill pipelines, RAG integration, LLM-powered compaction, modular system prompt
Multi-LLM & Channels	OpenAI, Anthropic, Google, custom endpoints; Telegram channel with voice; webhooks for external integrations
Security	5 security layers, rate limiting, user allowlists, tool confirmation

---

Quick Links

Getting Started

• Quick Start — Get running in 5 minutes
• Dashboard — Web UI for configuration and chat
• Configuration — Runtime config, environment variables, storage layout
• Deployment — Docker, systemd, production setup

Core Concepts

• Architecture — System design, pipeline, patterns
• Model Routing — Tier selection, dynamic upgrades, context overflow
• Skills — Create, configure, and orchestrate skills
• Tools — Built-in tools and workspace sandboxing
• MCP — Model Context Protocol integration

Features

• Plan Mode — Review-before-execute workflow
• Auto Mode — Autonomous goal-driven execution
• Memory System — Sessions, daily notes, long-term memory
• RAG Integration — LightRAG for semantic memory
• Webhooks — Inbound HTTP webhooks for external systems
• Security — Security layers and hardening

Development

• Coding Guide — Code style, conventions, commit messages
• Contributing — Development workflow, PR process
• Commands — Slash commands reference
• FAQ — Common questions and troubleshooting

---

Tech Stack

Component	Technology	Version
Language	Java	17+
Framework	Spring Boot	4.0.2
LLM Integration	LangChain4j	1.0.0-beta1
Dashboard	React + TypeScript + Vite	18 / 5.7 / 6
Messaging	Telegram Bots	8.2.0
Browser	Playwright	1.49.0
HTTP Client	Feign + OkHttp	13.5 + 4.12
Voice	ElevenLabs (STT + TTS)	API v1
Testing	JUnit 5 + Mockito	—
Code Quality	SpotBugs + PMD + JaCoCo	—

---

Getting Started (Quick Reference)

# Build & run
git clone https://github.com/alexk-dev/golemcore-bot.git
cd golemcore-bot
./mvnw compile jib:dockerBuild

docker run -d --name golemcore-bot \
  --shm-size=256m --cap-add=SYS_ADMIN \
  -e STORAGE_PATH=/app/workspace \
  -e TOOLS_WORKSPACE=/app/sandbox \
  -v golemcore-bot-data:/app/workspace \
  -v golemcore-bot-sandbox:/app/sandbox \
  -p 8080:8080 --restart unless-stopped \
  golemcore-bot:latest

# Open dashboard, login with temp password from logs
open http://localhost:8080/dashboard

Most settings (LLM API keys, Telegram token, tool toggles, etc.) are configured in the dashboard and persisted to preferences/runtime-config.json. See Dashboard and Configuration for details.

---

License

Apache License 2.0 — commercial use allowed, patent protection for contributors. See LICENSE.