Message middleware (inbound pipeline)

FluxyChat runs every inbound chat message through a shared pipeline before it is stored or broadcast (WebSocket type: "message" and POST /messages).

Stages

1. Validate — non-empty string, max 4000 chars (validateMessageContent).
2. Filter — optional block when MESSAGE_MIDDLEWARE_BLOCK_ON_MATCH=true and content matches BUILTIN_MODERATION_BLOCKED_SUBSTRINGS (comma-separated, case-insensitive).
3. Enrich — optional metadata tag when MESSAGE_MIDDLEWARE_ENRICH_TAG is set (exposed on the broadcast payload as middleware).

Rejected messages return:

• WebSocket: \{ type: "error", message: "content_blocked: …" \} or invalid_content: …
• REST: 403 / 400 with \{ error, code \}

Environment

Implementation

• apps/worker/src/lib/message-middleware.js — runInboundMessageMiddleware
• Wired in room-do.js and routes/messages-http.js

Post-persist moderation (auto-flag to moderation_events) remains in post-message-automations.js when block-on-match is off.