wacrm
Get started

Reference

Changelog

Notable user-facing changes to the wacrm template, newest first. For exact release tags and the full commit history, see the GitHub releases.

July 2026

AI knowledge base

The AI Assistant can now answer from your own content. Add FAQs, policies, or product details under Settings → AI Assistant → Knowledge base, and the relevant excerpts are retrieved into every draft and auto-reply — so the assistant answers instead of guessing or handing off. Hybrid retrieval: keyword full-text search works for every account with no extra key; add an embeddings key and semantic search turns on (matches by meaning, not just wording). Anthropic-only accounts keep the keyword path with no extra setup. Requires migration 030_ai_knowledge.sql (enables pgvector). See AI Assistant → Knowledge base.

AI Assistant (bring-your-own-key)

Bring your own OpenAI or Anthropic key — set it under Settings → AI Assistant — and wacrm calls the provider directly. No per-seat AI fee, no wacrm-run service in the middle; your key is stored encrypted and never shown again after saving. Two capabilities:

  • Drafted replies — a ✨ button in the inbox composer reads the recent conversation and writes a suggested reply for the agent to edit and send. Read-only; nothing sends on its own.
  • Auto-reply bot — optionally answers inbound messages that no Flow or Automation already handles and no agent has taken, with a per-conversation cap and a clean handoff to a human when it can't confidently help.

Deterministic responders win: Flows, then Automations, then the AI bot as the fallback. Requires migration 029_ai_reply.sql. See AI Assistant.

Public REST API (/api/v1)

Drive wacrm from your own scripts and automations — no dashboard required. Create a scoped, revocable API key under Settings → API keys, then call the REST endpoints:

  • Messages — POST /api/v1/messages sends text, template, or media to an E.164 number, finding-or-creating the contact + conversation.
  • Contacts — list (with search + tag filters), create (find-or-create by phone), read, and update, including tags.
  • Conversations & messages — browse conversations and their message history with delivery status.
  • Broadcasts — POST /api/v1/broadcasts launches a template campaign to a recipient list; poll for progress.

Keys carry only the scopes you grant, are account-scoped, and are shown exactly once (stored hashed). Every list endpoint shares one cursor-pagination format. Requires migration 026_api_keys.sql. See Public API.

Outbound event webhooks

Register an HTTPS endpoint and wacrm POSTs to it when things happen — message.received, message.status_updated, conversation.created — so your automations can react to inbound activity instead of polling.

  • Manage endpoints with POST /api/v1/webhooks (scope webhooks:manage); the signing secret is returned once and stored encrypted.
  • Each delivery is signed (X-Wacrm-Signature, HMAC-SHA256) so you can verify authenticity and reject replays, and carries a unique id to dedupe on.
  • Delivery is best-effort with auto-disable of persistently-failing endpoints; targets must be public https:// URLs.
  • Requires migration 028_webhook_endpoints.sql.

See Public API → Webhooks.

June 2026

Upload an image for template headers

Creating a template with an image header now works — you can upload a JPEG/PNG (≤ 5 MB) right in the builder, or paste a public URL.

  • Fixes the long-standing bug where an image-header template failed at submission: Meta only accepts a Resumable-Upload media handle (not a plain URL) as the review sample, so the app now performs that upload for you automatically — for both uploaded files and pasted URLs.
  • Approved image templates also send correctly (the public image URL is attached on every send).
  • Requires META_APP_ID (see Environment variables); text/body-only templates are unaffected.

See Templates → Header.

Send media in chat

Agents can now send photos, videos, documents, and voice notes directly from the inbox composer — previously media could only be received from customers or sent via a Flow.

  • Click the paperclip to attach a photo, video, or document, or to record a voice note. Optional caption on everything except voice.
  • Voice notes are encoded in the browser (Ogg/Opus), so there's no server-side audio tooling to install — works on any deployment, including Hostinger shared hosting. Recording auto-stops at 5 minutes.
  • Per-type size limits mirror WhatsApp's: photos 5 MB; video, documents, and voice 16 MB.
  • Uploads land in a dedicated, account-scoped chat-media Storage bucket (migration 023_chat_media.sql).
  • Rolled out behind a per-account beta flag first, now available to every account — no opt-in required.

See Inbox → Sending media for the full walkthrough.