Module: Tina4::Feedback

Defined in:

lib/tina4/feedback.rb

Overview

── Customer feedback widget — server-side plumbing ─────────────────

Mirrors tina4_python/dev_admin/init.py lines 1436-1645. The widget is for END USERS of a shipped Tina4 app (not developers). Whitelisted users get a floating bubble that proxies one conversational turn at a time to the Rust agent's intake endpoint at /feedback/intake.

Flow:

1. Framework middleware injects <script src="/__feedback/widget.js">
 into HTML responses for whitelisted users.
2. Widget POSTs to /__feedback/api/turn for each conversational turn.
3. The Ruby handler verifies whitelist + rate-limit, stamps the
 user identity server-side (client cannot fake `sender`), then
 forwards to the Rust agent's /feedback/intake.
4. Finalised tickets land in .tina4/chat/threads.json with
 kind:"feedback". Developer sees them in the dev admin sidebar.

Constant Summary

RATE_WINDOW =

1 hour

3600

RATE_MAX =

submissions/turns per user per hour

5

Class Attribute Summary

• .rate_mutex ⇒ Object readonly

  Returns the value of attribute rate_mutex.

Class Method Summary

• .feedback_enabled? ⇒ Boolean

  Hard master switch.

• .feedback_identify_user(request) ⇒ Object

  Best-effort user identity from auth headers.

• .feedback_is_whitelisted(request) ⇒ Object

  Returns [allowed, identity].

• .feedback_rate_limit_ok(user) ⇒ Object

  5 turns/hour per user.

• .feedback_whitelist ⇒ Object

  Comma-separated emails / user IDs in env.

• .handle_request(env) ⇒ Object

  Dispatcher used by RackApp — returns a Rack triple if the path matches a /__feedback route, else nil.

• .handle_turn(env) ⇒ Object

  POST /__feedback/api/turn — whitelist check + rate-limit + stamp sender server-side, forward to Rust agent /feedback/intake.

• .handle_widget_js(_env) ⇒ Object

  GET /__feedback/widget.js — serve the bundle at lib/tina4/public/__feedback/widget.js.

• .inject_feedback_widget(request, html) ⇒ Object

  Insert the widget