Skip to content

techdivision/opencode-plugin-config

BranchesTags

Repository files navigation

@techdivision/opencode-plugin-config

Automatic project configuration sync for the OpenCode AI Plugin System.

Overview

This plugin eliminates manual configuration for OpenCode plugins. Instead of hand-maintaining JIRA workflow statuses, Tempo accounts, agent defaults, and pricing data in every project, the config plugin makes a single webhook call at startup, receives all remote defaults, deep-merges them with the local configuration (local always wins), and exposes the result as process.env.OPENCODE_PROJECT_CONFIG for all downstream plugins.

No files are written -- the merged config lives entirely in memory, consistent with how shell-env handles environment variables. For architectural details see the arc42 documentation.

How It Works

shell-env (init)
  --> process.env populated from .env files

config (init)
  1. Read local config cascade (Global + Project opencode-project.json)
  2. Plugin discovery via discoverPlugins()
  3. POST webhook call to n8n
  4. Schema validation per section (Ajv + plugin-owned JSON Schemas)
  5. Deep-merge: remote as base, local wins
  6. --> process.env.OPENCODE_PROJECT_CONFIG = JSON string

consumer plugins (init)
  --> JSON.parse(process.env.OPENCODE_PROJECT_CONFIG)
  --> Fallback: read local opencode-project.json directly

Installation

# Global installation (recommended)
cd ~/.config/opencode
npm install @techdivision/opencode-plugin-config

Then link all plugins:

opencode-link all

Load Order

The plugin must be listed after shell-env and before all consumer plugins in .opencode/package.json:

{
  "dependencies": {
    "@techdivision/opencode-plugin-shell-env": "^1.2.0",
    "@techdivision/opencode-plugin-config": "^0.1.0",
    "@techdivision/opencode-plugin-time-tracking": "^1.4.0",
    "@techdivision/opencode-plugins": "github:techdivision/opencode-plugins"
  }
}

Order: shell-env (1st) --> config (2nd) --> all other plugins (3rd+)

Configuration

Plugin Config (opencode-project.json)

The plugin reads its own settings from the config section:

{
  "config": {
    "sync_url": "https://n8n.example.com/webhook/oc-config-sync",
    "sync_token": "optional-bearer-token"
  }
}

Or via environment variables (loaded by shell-env from .env):

OC_CONFIG_SYNC_URL=https://n8n.example.com/webhook/oc-config-sync
OC_CONFIG_SYNC_TOKEN=optional-bearer-token
OPENCODE_USER_EMAIL=user@example.com

Resolution order: config.sync_url in opencode-project.json --> {env:VAR} placeholder --> process.env.OC_CONFIG_SYNC_URL fallback --> skip webhook.

Config Cascade (3 Layers)

Layer 1 (Base):     ~/.config/opencode/opencode-project.json    Global defaults
Layer 2 (Override): <project>/.opencode/opencode-project.json   Project overrides
                    ──────────────────────────────────────────
                    = "Local Config" (deep-merge, Layer 2 wins)

Layer 3 (Remote):   Webhook response                            Remote defaults
                    ──────────────────────────────────────────
                    = "Final Config" (deep-merge: remote as base, local wins)

Merge rule: deepmerge(remote, local) -- local always wins.

  • Both values are objects --> merge recursively
  • Local has a value --> local wins (regardless of type)
  • Only remote has a value --> remote value is adopted (new defaults)
  • Arrays --> local array replaces remote array completely

Protected top-level fields ($schema, version) are never overwritten by the webhook.

Minimal Project Config (Seed)

The minimum a project needs in <project>/.opencode/opencode-project.json:

{
  "jira": {
    "project": "MYPROJ",
    "base_url": "https://your-instance.atlassian.net"
  },
  "time_tracking": {
    "valid_projects": ["MYPROJ"]
  }
}

The webhook uses these seed values (e.g. jira.project) to look up all remaining configuration from JIRA API and Google Sheets automatically.

Webhook API

Request

POST {sync_url}
Content-Type: application/json
Authorization: Bearer {sync_token}   (optional)

Payload

{
  "plugin_version": "0.1.0",
  "email": "user@example.com",
  "plugins": ["config", "time-tracking", "jira", "shell-env"],
  "config": {
    "jira": {
      "project": "MYPROJ",
      "base_url": "https://your-instance.atlassian.net"
    },
    "time_tracking": {
      "valid_projects": ["MYPROJ"]
    }
  }
}
Field Source Description
plugin_version plugin.json version Config plugin version for compatibility check
email process.env.OPENCODE_USER_EMAIL Identifies the user (Google Sheet, JIRA lookups)
plugins discoverPlugins() List of all discovered plugin names
config Merged local config (Global + Project) Entire local config as seed for the webhook

Response

{
  "version": "0.1.0",
  "config": {
    "jira": {
      "project": "MYPROJ",
      "workflow": { "status": { "open": { "name": "Open", "id": "1" } } },
      "tempo": { "account_field_id": "customfield_10039", "accounts": {} }
    },
    "time_tracking": {
      "agent_defaults": { "@implementation": { "issue_key": "MYPROJ-5" } },
      "pricing": { "periods": [] }
    }
  }
}
Field Description
version Minimum plugin version this config was generated for
config Remote config sections, keys match plugin names (_ instead of -)

Version compatibility: If response.version > plugin_version, the response is discarded with a warning ("config was generated for a newer plugin version, please update").

Error Behavior

Situation Behavior
sync_url not configured Webhook skipped, local config applies
OPENCODE_USER_EMAIL not set Webhook skipped, local config applies
Webhook unreachable (timeout 5s) Local config applies, warning logged
Webhook returns HTTP 4xx/5xx Local config applies, warning logged
Response validation fails Local config applies, warning logged
Version incompatibility Local config applies, warning logged
No local config present Empty object {}, only remote if available

In all error cases the plugin continues without blocking. The local config is always written to process.env.OPENCODE_PROJECT_CONFIG -- even without a webhook response.

Integration in Existing Plugins

Consumer plugins should read from process.env first, with a file fallback:

// Config from process.env (set by config plugin)
// Fallback to local file if config plugin is not installed
function loadConfig(sectionKey: string): Record<string, unknown> {
  const envConfig = process.env.OPENCODE_PROJECT_CONFIG
  if (envConfig) {
    try {
      const parsed = JSON.parse(envConfig)
      if (parsed[sectionKey]) return parsed[sectionKey]
    } catch { /* fallback */ }
  }
  try {
    const raw = fs.readFileSync('.opencode/opencode-project.json', 'utf-8')
    return JSON.parse(raw)[sectionKey] ?? {}
  } catch { return {} }
}

// Usage
const myConfig = loadConfig('time_tracking')

This change is backwards-compatible: if the config plugin is not installed, the fallback reads the local file directly.

Schema Validation

Each plugin can declare its own JSON Schema for its config section via configSchema in plugin.json. The config plugin validates each webhook response section against the corresponding plugin's schema using Ajv.

Validation happens per section, not all-or-nothing. If the jira section fails validation, time_tracking data can still be used.

Example plugin.json with schema declaration:

{
  "name": "time-tracking",
  "description": "Automatic time tracking plugin for OpenCode.",
  "category": "standard",
  "version": "1.4.0",
  "configSchema": "schemas/config.schema.json"
}
Situation Behavior
Section has schema + validates Section included in merge
Section has schema + fails validation Section skipped, warning logged
Section has no schema Included without validation

The configSchema field is optional and backwards-compatible. Plugins without it continue to work -- their sections are merged without schema validation.

For details on the validation architecture, see arc42 Konzepte (8.10).

Development

git clone https://github.com/techdivision/opencode-plugin-config.git
cd opencode-plugin-config
npm install
npm link  # For local development

Project Structure

opencode-plugin-config/
  package.json
  plugin.json
  tsconfig.json
  schemas/
    config.schema.json         Own config schema (for "config" section)
  src/
    config.ts                  Entry point (plugin function)
    services/
      ConfigLoader.ts          Reads + merges local config cascade
      ConfigSyncer.ts          Webhook call
      ConfigMerger.ts          Deep-merge with local precedence
      SchemaValidator.ts       Ajv-based schema validation per section
    types/
      PluginConfig.ts          { sync_url, sync_token }
      SyncPayload.ts           Webhook request type
      SyncResponse.ts          Webhook response type
  n8n/
    workflow.json              Exported n8n workflow
    README.md                  n8n setup instructions

Architecture Documentation

Detailed architecture docs live alongside the source:

Document Content
Feature: feat-cfg Feature description and scope
Bausteinsicht (5.5) Building block view of the config plugin
Laufzeitsicht (6.6) Runtime scenario: init sequence
Konzepte (8.8, 8.9, 8.10) Config cascade, webhook sync, schema validation
ADR-009 to ADR-012 Architecture Decision Records

License

MIT

About

OpenCode Plugin - Automatic project configuration sync via webhook with deep-merge, schema validation and in-memory delivery

Resources

Readme

License

MIT license
Activity
Custom properties

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages