Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/SamBleed/opencode-obsidian/llms.txt

Use this file to discover all available pages before exploring further.

Bunker OS is a local-first knowledge operating system built on Obsidian and OpenCode. Every source you ingest, every research session you run, and every architectural decision you make gets filed into a structured, cross-referenced wiki that compounds in value over time. Nothing leaves your machine unless you push it.

Quickstart

Clone the vault, open it in Obsidian and OpenCode, and run your first research in under 5 minutes.

Skills Reference

Explore the 13 bundled OpenCode skills — from autonomous research to wiki linting.

n8n Automation

Async pipelines, multi-channel alerts, and the AOC v4 Enterprise workflow.

Architecture

Understand the layers: capture, normalize, classify, evidence, automation, and retrieval.

What Bunker OS Does

Bunker OS is not a traditional app. It is an operations system that wraps Obsidian (your vault) and OpenCode (your AI agent) into a disciplined, automated pipeline:
  • Captures sources into .raw/ and ingests them into a structured wiki
  • Researches topics autonomously across 3 iterative rounds with Exa
  • Retrieves knowledge instantly with BM25 text search — no embeddings, no LLM calls
  • Automates async workflows with n8n: alerts, issue creation, error handling
  • Tests its own integrity with 430 tests across 5 suites on every push
1

Clone the vault

git clone https://github.com/SamBleed/opencode-obsidian.git
cd opencode-obsidian
2

Open in Obsidian

In Obsidian: Manage Vaults → Open folder as vault → select opencode-obsidian/
3

Open OpenCode in the same folder

Launch OpenCode from inside the opencode-obsidian/ directory so it picks up the AGENTS.md, skills, and hooks automatically.
4

Start your first session

autoresearch zero-trust architecture
Bunker OS will decompose the topic, run 3 rounds of web research, and file structured wiki pages — sources, concepts, entities — automatically.

Key Capabilities

Autoresearch

3-round autonomous research with Exa + webfetch. Files sources, concepts, and entities directly to the wiki.

BM25 Retrieval

Pure Python stdlib search over the entire wiki. No embeddings, no API keys, no external services.

/think Framework

10-principle decision framework for architectural choices, audits, and post-mortems.

Wiki Ingest

Drop any file in .raw/ and say ingest [filename]. Entities and concepts extracted automatically.

n8n Pipelines

AOC v4 Enterprise: 37-node pipeline with AI triage, GitHub issue creation, and Slack/Telegram/Discord alerts.

430 Tests + CI

5 test suites: n8n workflow connections, vault integrity, script validation, YAML, and BM25 retrieval.

Built on the LLM Wiki Pattern

Bunker OS is grounded in Andrej Karpathy’s LLM Wiki pattern: the agent writes to a persistent knowledge base rather than answering in chat. Answers cite specific wiki pages, not model training data. Knowledge compounds like interest — each session makes every future session smarter.
Bunker OS complements rather than replaces claude-obsidian. Where claude-obsidian is a researcher and second brain, Bunker OS is an operator. Both can point at the same vault simultaneously.

Build docs developers (and LLMs) love