frappe-live-code-extractor

Queries the live Frappe database (read-only) and writes each artifact's code into source files and meta.json under <app>/live/<doctype-slug>/. Makes all DB-resident code visible to grep, LSP, and AI assistants. Safe to run at any time.

All doctypes — built-in Frappe types and custom app types — are treated identically. Each doctype has a config.json at <app>/live/<doctype-slug>/config.json. The script resolves it via a 3-step fallback:

1. App config — <app>/live/<doctype-slug>/config.json if it exists
2. Skill asset — copies the bundled default from assets/ into the app tree, then uses it
3. Agent — interview the user / infer from context to create a config (see REFERENCE.md)

Quick start

--app is required. The agent must determine the app before calling the script.

Agent: inferring the app

If the user has not specified an app, determine it from context:

1. From the workspace folder — use the workspace folder the user is currently in. The app that owns the DocType being extracted is irrelevant — always use the user's active workspace app.
2. From the conversation — if the user has explicitly mentioned an app name, use it.
3. From installed apps — as a last resort, query the DB and pick the first non-framework app:

cd /workspace/development/frappe-bench
./env/bin/python - <<'EOF'
import os; os.environ['FRAPPE_STREAM_LOGGING'] = '1'
import frappe
frappe.init(site='development.localhost', sites_path='sites'); frappe.connect()
FRAMEWORK = {'frappe', 'erpnext', 'payments', 'hrms'}
apps = [a for a in frappe.get_installed_apps() if a not in FRAMEWORK]
print(apps[0] if apps else '')
frappe.destroy()
EOF

Once the app is known, run:

cd /workspace/development/frappe-bench
./env/bin/python skills/.agents/skills/frappe-live-code-extractor/scripts/extract.py \
  --app myapp --site development.localhost

To extract a single DocType (faster, useful after editing one doctype):

./env/bin/python skills/.agents/skills/frappe-live-code-extractor/scripts/extract.py \
  --app myapp --doctype "Report"

--doctype accepts the exact DocType name (case-insensitive). Works for both standard Frappe DocTypes (Server Script, Report, …) and custom ones registered via config.json.

When --doctype is specified, config seeding from assets is limited to that DocType only — no other config.json files are written.

Workflows

Refresh before a refactor

When the user says "refresh the live code snapshot":

1. Run the extraction command above
2. Grep across apps/<app>/<app>/ — the live/ subtree is included automatically
3. Proceed with the refactor

Add a new DocType for extraction

When the user asks to "mirror" or "extract" a DocType that has no config yet:

1. Discover fields — query the schema for Code/Text/HTML fields:

   cd /workspace/development/frappe-bench
   ./env/bin/python -c "
   import os; os.environ['FRAPPE_STREAM_LOGGING'] = '1'
   import frappe
   frappe.init(site='development.localhost', sites_path='sites'); frappe.connect()
   meta = frappe.get_meta('Your DocType Name')
   for f in meta.fields:
       if f.fieldtype in ['Code','Text','Long Text','Small Text','HTML','HTML Editor','Markdown Editor']:
           print(f.fieldname, f.fieldtype, f.label)
   frappe.destroy()
   "
   

2. Create config.json at <app>/live/<doctype-slug>/config.json — see REFERENCE.md for schema and examples.

   ⚠️ Never add configs to skills/.agents/skills/frappe-live-code-extractor/assets/ unless explicitly told to. That directory is reserved for configs bundled with the skill itself. User-defined configs always go into the app's own live/<doctype-slug>/config.json.

3. Run extraction — the file is auto-discovered and processed
4. Commit config.json and extracted files

Advanced features

See REFERENCE.md for:

• Supported artifact types table
• Output directory structure and meta.json format
• config.json schema (all fields, advanced options)
• Config examples (dynamic extensions, mutual exclusion, per-group filtering)
• Integration with git, multi-app extraction, troubleshooting