Skip to content

OpenCode_Integration.md

Latest commit

 

History

History
43 lines (24 loc) · 3.1 KB

OpenCode_Integration.md

File metadata and controls

43 lines (24 loc) · 3.1 KB
Raw

OpenCode Agent Integration Guide

OpenCode is an open-source AI coding and terminal agent. It supports autonomous development and multi-agent workflows. When you configure NoteBrain as a dedicated OpenCode agent (notebrain-chat), you make your Obsidian vault an interactive semantic knowledge base for your AI pair programming sessions.

Why Use NoteBrain with OpenCode?

Obsidian vaults have thousands of interlinked notes, design documents, code snippets, and meeting records. Traditional AI agents cannot process these vaults correctly because:

  1. Context window flooding: The agents run grep or read full markdown files. This puts thousands of irrelevant lines into the context and increases token consumption.
  2. Missing semantic context: Keyword search (grep) does not find synonyms, conceptual bridges, or structural graph hops (wikilinks).

NoteBrain avoids these problems because it queries a local ChromaDB HNSW vector index. When you use the permission system of OpenCode, you can sandbox the agent. Then the agent uses only high-precision semantic queries (notebrain search, hidden, connections, backlinks).

Agent Configuration (notebrain-chat)

In OpenCode, agent configurations are Markdown files. They are in .opencode/agents/ (for project-specific agents) or ~/.config/opencode/agents/ (for global agents). The example agent file is Notebrain-Chat in this repository. Copy it to .opencode/agents/notebrain-chat.md (project-specific) or ~/.config/opencode/agents/notebrain-chat.md (global). The file name becomes the agent name.

An OpenCode agent file has two parts:

  1. YAML Frontmatter: This part gives agent metadata (name, description, mode, temperature, model) and the strict permission sandbox.
  2. Markdown System Instructions: This part gives the role of the agent, its operating rules, its tiered retrieval strategy, and the required response formatting.

Invoke the Agent in OpenCode

When the configuration is complete, you can interact with your notebrain-chat agent in two ways:

  1. Explicit Mode Selection (primary mode): The example agent file sets mode: all. Because of this, notebrain-chat works as a primary agent and as a subagent. You can select notebrain-chat as your active primary agent in the OpenCode CLI session. Alternatively, you can use the /notebrain-chat command. The agent answers all prompts with data from your Obsidian vault.

  2. Automatic Router Delegation (subagent mode): If you change mode: all to mode: subagent, the routing engine reads the description frontmatter:

    "Use NoteBrain to search, summarize, and explore an Obsidian vault. Invoke this agent whenever the user asks about their notes, knowledge base..."

    When you ask a question like "What architectural decisions did I write down about Redis Streams in my vault?", OpenCode spawns notebrain-chat in the background automatically. The agent executes the semantic query with notebrain search and sends citations back to your coding context.

Example

Notebrain Chat