Confirm inbound SMS→AMI working live; rule out outbound over SIP

Inbound: a real text through the full path (Anveo webhook -> relay ->
AMI MessageSend -> Sipnetic) landed with a SIP 200 OK, confirmed live.
Removes the last "UNVERIFIED"/"believed correct" hedges from
sms-inbound.sh now that the AMI permission class and the
Destination-not-To fix are both proven, not just plausible.

Outbound over SIP: tested directly by sending a MESSAGE toward Anveo's
trunk (the mirror image of the inbound webhook). Anveo's SBC responded
501 Not Implemented -- a real, unambiguous rejection of the method
itself. Closes off this avenue for good, symmetric with inbound
SMS-over-SIP already being confirmed unavailable on this DID: neither
direction is offered on this account via SIP. Sending still has no
working path here until Anveo activates HTTP SMS-API access.

Also updates docs/anveo-direct-setup-guide.md's SMS section, which still
described the ntfy-based mechanism this session fully replaced with
AMI/Sipnetic delivery, and corrects its stale "not available on Direct"
sending claim with what's actually been confirmed: SIP MESSAGE outbound
is a dead end (501), the HTTP API is real but pending Anveo activation.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JDyKC6Kdg7tofmYSmRtgww

docs/anveo-direct-setup-guide.md

@@ -199,7 +199,7 @@ In the Security Dashboard's PSTN Trunk tab:
   # container is named easy-asterisk-do instead
   ```
 
-## 8. SMS — receiving verification codes
+## 8. SMS — receiving texts in Sipnetic
 
 Voice and SMS are separate features on an Anveo DID and are configured in
 different places. This section covers **receiving** only; see "What about
@@ -242,41 +242,49 @@ Run the installer and follow what it prints:
 sudo ./setup.sh sms-inbound
 ```
 
-It generates a long random ntfy topic, then offers two ways for Anveo to
+A small systemd service on the droplet receives Anveo's webhook request,
-reach it:
+looks up which extension (or Ring Group) currently owns the destination DID
+— the exact same `pstn-personal-dids.conf`/`pstn-groups.conf` data
+inbound-voice ring logic already reads — and delivers the text into
+Asterisk over its Manager Interface as a SIP `MESSAGE`, landing in Sipnetic
+the same way this project's internal extension-to-extension texting already
+does. Not a push notification: a real message in the softphone's own thread.
 
-- **Relay (recommended)** — a small systemd service on the droplet receives
+("SMS over SIP" — Anveo delivering the text as a MESSAGE/INVITE directly to
-  Anveo's request and republishes to ntfy properly. Two concrete wins: a
+this box's SIP trunk, skipping the HTTP webhook entirely — was tried first
-  message body containing `&` survives intact (Anveo interpolates the text
+and confirmed **not offered** on this DID: the SMS tab only ever showed the
-  into the query string unescaped, so an unencoded `&` otherwise truncates
+plain "Forward to URL" option, no SIP-based alternative.)
-  the message), and your ntfy credentials never get stored in Anveo's portal.
-- **Direct** — Anveo calls ntfy itself; nothing runs on the droplet. Simpler,
-  but the URL you paste into Anveo carries your ntfy token, and the `&` case
-  loses the tail of the message.
 
 Then in the Anveo portal: **Phone Numbers → the DID → SMS tab**. The tab has
 exactly one control — a **Forward to URL** checkbox and a text field. Tick the
 box, paste the string the installer printed into the field, and press **SAVE**
 (not RETURN, which discards). Keep the `$[message]$` placeholder **last** in
-that URL — that's what makes the unescaped-`&` case recoverable.
+that URL — Anveo interpolates the message text into the query string
+unescaped, so a body containing `&` otherwise truncates; with the placeholder
+last, everything after it is recoverable verbatim.
 
-The field has no visible length limit, but the generated URLs are long
+Send a text to the number from another phone; it should land in Sipnetic
-(~90 characters in relay mode, ~150+ in direct mode, since that one carries
+within seconds. `journalctl -u sms-inbound -f` shows sender, recipient,
-the ntfy auth parameter). After saving, reopen the tab and confirm the whole
+message length, and the full AMI exchange — never the message body itself,
-string came back intact rather than truncated — if it didn't, relay mode is
+since the journal has a wider audience than the message thread does.
-the shorter of the two.
-
-Send a text to the number from another phone; the notification should arrive
-within seconds. `journalctl -u sms-inbound -f` shows sender, recipient and
-message length (never the body — these are one-time passcodes and the journal
-has a wider audience than the notification does).
 
 ### What about sending?
 
-Not covered, on purpose. Outbound SMS isn't available on Anveo Direct — Anveo
+Two things confirmed this way, neither of them a working path yet:
-support directs users to an Anveo **Retail** account for it, which is a
+
-second account to fund and manage. Any of the free texting apps covers
+- **Outbound over SIP** — sending a `MESSAGE` toward Anveo's trunk (the
-sending without involving this box.
+  mirror image of the inbound webhook) was tried directly: Asterisk placed
+  the request fine, but Anveo's SBC came back `501 Not Implemented`. A real,
+  unambiguous rejection of the method itself — not a config or auth problem,
+  and not something any amount of retrying the request format will get past.
+- **Outbound via Anveo's HTTP API** (`api/v1.asp?action=sms`) — real and
+  documented, but gated behind Anveo support activating SMS-API access for
+  this account; still pending as of this writing (limited support tickets on
+  this account, so this has been a slow back-and-forth rather than a quick
+  toggle).
+
+Until that activation lands, there's no outbound path from this box. Any of
+the free texting apps covers sending in the meantime without involving it.
 
 ### MMS and group texts
 

services/sms-inbound.sh

@@ -60,11 +60,8 @@ _sms_detect_container_name() {
 # ── AMI: a scoped manager.conf user, MessageSend only ───────────────────────
 # Localhost-only (bindaddr + permit below) since the relay runs on this same
 # host, not over the network — no firewall port to open for this. "message"
-# is the AMI permission class MessageSend needs; this hasn't been confirmed
+# as the AMI permission class is confirmed live: MessageSend succeeds with
-# against a live MessageSend call yet (see the relay's own comment on
+# just read/write = message, no wider class needed.
-# ami_deliver) — if the very first real delivery attempt gets an AMI
-# permission error in the journal, widen read/write here first before
-# looking anywhere else.
 #
 # Idempotent: creates manager.conf fresh if absent, otherwise ensures
 # enabled=yes and the [smsrelay] section exist without disturbing anything
@@ -333,21 +330,25 @@ def _ami_read_response(sock_file):
 def ami_deliver(from_number, to_exts, body):
     """Logs into AMI once and sends one MessageSend action per recipient.
 
-    "message" as the AMI permission class is confirmed live (login succeeds).
+    Confirmed working end-to-end live (2026-07-27): a real inbound text
-    Destination (not To) is what actually resolves an outgoing message's
+    delivered through this exact path landed in Sipnetic with a SIP 200 OK.
-    endpoint/technology -- confirmed against this box's own
+    Two things worth keeping straight if this ever needs touching again:
-    `manager show command MessageSend`: To alone is documented as a
+
-    backward-compatible fallback for the destination, but live testing
+    - "message" as the AMI permission class is enough on its own (login +
-    (bare "pjsip:212", then "pjsip:212@domain", both for To with no
+      MessageSend both succeed with just read/write = message).
-    Destination) produced zero SIP wire traffic in either case -- `pjsip
+    - Use Destination, not To, to resolve the endpoint. `manager show
-    set logger on` during a real attempt showed Asterisk never even tried
+      command MessageSend` documents To-alone as a backward-compatible
-    reaching the target's registered contact, so that fallback path isn't
+      fallback for the destination when Destination is omitted, but live
-    actually wired up on this Asterisk version regardless of what the docs
+      testing that fallback (bare "pjsip:212", then "pjsip:212@domain",
-    promise. Destination's documented "endpoint" form -- bare "pjsip:<ext>",
+      both as To with no Destination) produced zero SIP wire traffic
-    no domain -- resolves via the endpoint's own default aor/contact, which
+      either way -- Asterisk never even attempted to reach the target's
-    is exactly the live, registered contact `pjsip show contacts` already
+      registered contact. Destination's documented "endpoint" form (bare
-    confirmed exists. Every AMI response is still logged in full so the
+      "pjsip:<ext>", no domain) is what actually resolves via the
-    next attempt is self-diagnosing if this isn't the whole fix either.
+      endpoint's default aor/contact and is what's used below.
+
+    Every AMI response is still logged in full, since a wrong recipient,
+    an unregistered contact, etc. all still need to be diagnosable from
+    the journal alone.
 
     Returns (delivered_count, total_count)."""
     if not AMI_SECRET: