Component Map¶

What it is¶

The Component Map is the architectural blueprint of the repository's technology stack. It categorizes every tool, service, and protocol into a functional lifecycle: Ingest, Store, Understand, Decide, Act, and Sync.

What problem it solves¶

In a rapidly expanding ecosystem of AI agents and self-hosted services, it is easy to lose track of how individual components interact. This map provides a high-level view of the pipeline, helping users and automated agents identify gaps, avoid duplicates, and understand the flow of information from raw data to autonomous action.

Where it fits in the stack¶

It is a Core Architectural Document that serves as the foundation for documentation taxonomy. It guides where new tools should be placed in mkdocs.yml and how they should be linked in the KnowledgeOps graph.

Typical use cases¶

Onboarding: Helping new contributors understand the relationship between different parts of the stack.
Audit Tool: Identifying categories that are over-saturated or under-represented (e.g., "Decide" vs "Act").
Integration Planning: Determining which tools should be connected via n8n or MCP based on their functional roles.

Strengths¶

Functional Clarity: Groups tools by "What they do" rather than "Who made them".
Pipeline-Oriented: Reflects the real-world flow of data in an automated home-office.
Dynamic: Updated regularly to include the latest frontier models and agent frameworks.

Limitations¶

Oversimplification: Some tools (like Nextcloud) span multiple categories and must be placed in a "primary" category.
Maintenance Overhead: Requires manual updates as new canonical pages are added to the repository.

When to use it¶

When planning a new automation workflow and selecting the best tool for each stage (Ingest, Understand, Act).
When auditing the repository to ensure balanced coverage of the AI and automation landscape.

When not to use it¶

For granular technical configuration or installation steps (refer to individual tool pages instead).
As a real-time system monitor (it is a conceptual map, not a status dashboard).

Functional Categories¶

1. Ingest¶

Tools responsible for receiving or capturing raw information. - Email (IMAP): Paperless-ngx, n8n - Scanning: OCRmyPDF, Docling MCP - Manual Input: Obsidian, Logseq - Inventory: Homebox, Grocy - Bookmarks/Tasks: Linkwarden, Habitica - Downloads: qBittorrent, Jackett - Web Crawling: Crawl4AI, Firecrawl

2. Store¶

Tools where information resides in a persistent state. - Document Archive: Paperless-ngx - File Sync/Cloud: Nextcloud, Syncthing - Calendars/Contacts: Radicale, Google Calendar - Media/Projects: Jellyfin, Focalboard - Distributed: Storj Node - Fine-tuning: OpenPipe

3. Understand (Reasoning Engines)¶

The brains of the stack that process and reason over information. - Proprietary APIs: OpenAI, Anthropic, Mistral AI, DeepSeek, Google Gemini - Local Models: Ollama, Local LLMs (llama.cpp), vLLM, TGI, SGLang, ExLlamaV2, Aphrodite Engine, MLX, ansigpt, ZSE - Aggregators: OpenRouter, Perplexity, Valyu - Semantic Search: Paperless-AI, RAGFlow, PageIndex

4. Decide (Orchestrate & Route)¶

Tools that determine which actions to take and how to route requests. - Routing Layers: LiteLLM, OpenRouter, MCP Registry - Workflow Engines: n8n, Home Assistant, Mycelium, Haystack, Semantic Kernel, DSPy - Cloud Connectors: Zapier, Make - Identity: Authentik

5. Act (Agents & Execution)¶

Tools that perform modifications to the environment. - Autonomous Agents: Mistral AI, OpenHands, Droid, TeamOut, OpenSwarm, OpenClaw, CrewAI, AutoGen, Smolagents, LangGraph, Agency Swarm, Composio, Phidata, Bee Agent Framework, Agno - Browser Agents: Browser Use, Skyvern - Coding Assistants: Aider, Cursor, Zed, VS Code, Claude Code - Custom Orchestration: Custom Agents (SSH + LLM Loop) - Execution Plane: SSH Execution Patterns - Home Control: Home Assistant

6. Sync & Infrastructure¶

Tools that ensure consistency and secure connectivity. - Network Access: Tailscale - Protocols: CalDAV - Data Transfer: rclone Automation

7. Benchmark¶

Tools for evaluating model performance and reasoning. - Reasoning: Humanity's Last Exam (HLE), LangSmith, LM Evaluation Harness, DREAM Benchmark - Agentic: Terminal-Bench, SWE-bench, LongCLI-Bench, PA-bench - Local Performance: Ollama Benchmark CLI, LLMPerf

Sources / References¶

Contribution Metadata¶

Last reviewed: 2026-07-20
Confidence: high