@securecoms/agent-bridge 0.26.0 → 0.26.2

package/connectors/hermes/adapter.py

@@ -80,6 +80,27 @@ _PLUGIN_DIR = os.path.dirname(os.path.abspath(__file__))
 _BINDING_PATH = os.path.join(_PLUGIN_DIR, "securecoms.binding.json")
 _CONNECTIONS_DIR = os.path.expanduser("~/.securecoms/connections")
 
+# Voice calls (owner report 2026-10-02): a turn the voice front desk handed to
+# Hermes carries the daemon's compiled-in spoken-reply guidance. It rides as
+# the event's per-message ephemeral system prompt (MessageEvent.channel_prompt)
+# so the answer is fit to be read aloud. Byte-equal to
+# @securecoms/agent-runtime's VOICE_TURN_PREFIX.
+VOICE_TURN_PREFIX = (
+    "This message was spoken on a live voice call and your reply will be read aloud: "
+    "plain spoken sentences, no markdown, lists, links or code. "
+    "Use your tools as usual when the question needs them. "
+)
+_VOICE_GUIDANCE_MAX_CHARS = 1000
+
+
+def voice_prompt_of(frame: Dict[str, Any]) -> Optional[str]:
+    """The voice system prompt for a `voice: true` frame, else None."""
+    guidance = frame.get("guidance")
+    if frame.get("voice") is not True or not isinstance(guidance, str):
+        return None
+    g = guidance.strip()[:_VOICE_GUIDANCE_MAX_CHARS]
+    return VOICE_TURN_PREFIX + g if g else None
+
 
 def _error_class(error: object) -> str:
     """Return a bounded class name without reading exception prose."""
@@ -417,6 +438,9 @@ class SecureComsAdapter(BasePlatformAdapter):
             message_id=str(frame.get("message_id") or int(time.time() * 1000)),
             timestamp=datetime.now(),
         )
+        voice_prompt = voice_prompt_of(frame)
+        if voice_prompt is not None:
+            event.channel_prompt = voice_prompt
         await self.handle_message(event)
 
     # ── Outbound ──────────────────────────────────────────────────────────

package/connectors/hermes/tests/test_adapter_wiring.py

@@ -54,6 +54,7 @@ def _install_gateway_stub():
         source: Any = None
         message_id: Optional[str] = None
         timestamp: Any = field(default_factory=datetime.now)
+        channel_prompt: Optional[str] = None
 
     @dataclass
     class SendResult:
@@ -324,6 +325,34 @@ class TestEchoGuardWiring:
         assert ws.sent == [{"type": "ack", "message_id": "m2"}]
 
 
+class TestVoiceGuidance:
+    @pytest.mark.asyncio
+    async def test_voice_turn_carries_guidance_as_channel_prompt_text_turn_none(self, monkeypatch):
+        a, _ = make_adapter(monkeypatch)
+        a._bound_identity = "A"
+        handled = []
+
+        async def handler(event):
+            handled.append(event)
+
+        a.set_message_handler(handler)
+        frames = [
+            {"type": "message", "sender_id": "B", "channel_id": "c", "text": "what's the weather?",
+             "message_id": "v1", "voice": True, "guidance": "One short spoken sentence."},
+            {"type": "message", "sender_id": "B", "channel_id": "c", "text": "hi",
+             "message_id": "t1", "guidance": "ignored without voice"},
+        ]
+        ws = FakeWS([_FakeWSMsg(aiohttp.WSMsgType.TEXT, json.dumps(f)) for f in frames]
+                    + [_FakeWSMsg(aiohttp.WSMsgType.CLOSED)])
+        a._ws = ws
+
+        await a._pump()
+
+        assert [e.text for e in handled] == ["what's the weather?", "hi"]
+        assert handled[0].channel_prompt == adapter.VOICE_TURN_PREFIX + "One short spoken sentence."
+        assert handled[1].channel_prompt is None
+
+
 # ---------------------------------------------------------------------------
 # (2) send() success fires the status manager's on_reply_posted
 # ---------------------------------------------------------------------------

package/connectors/openclaw/openclaw-entry.inbound.test.ts

@@ -255,6 +255,22 @@ describe("openclaw-entry — inbound sender, session, authority, dedup (P0 Task
     expect(rt.recorded.map((r) => r.sessionKey)).toEqual([a.SessionKey, b.SessionKey]);
   });
 
+  it("a voice turn carries the daemon's spoken-reply guidance as GroupSystemPrompt; a text turn carries none", async () => {
+    const rt = makeRuntime();
+    const { instances } = await start(rt.runtime);
+    const ws = instances[0];
+
+    deliver(ws, { channel_id: "chan-V", message_id: "m-V", sender_id: OWNER_ID, text: "what's the weather?", voice: true, guidance: "One short spoken sentence.", authorization: auth("owner_dm") });
+    deliver(ws, { channel_id: "chan-T", message_id: "m-T", sender_id: OWNER_ID, text: "hi", guidance: "ignored without voice", authorization: auth("owner_dm") });
+    await waitFor(() => acksOf(ws).length === 2);
+
+    const { VOICE_TURN_PREFIX } = await import("./openclaw-entry.js");
+    const [v, t] = rt.contexts;
+    expect(v.GroupSystemPrompt).toBe(`${VOICE_TURN_PREFIX}One short spoken sentence.`);
+    expect(v.RawBody).toBe("what's the weather?");
+    expect(t.GroupSystemPrompt).toBeUndefined();
+  });
+
   it("two DMs from different humans do not share a session", async () => {
     const rt = makeRuntime();
     const { instances } = await start(rt.runtime);

package/connectors/openclaw/openclaw-entry.js

@@ -85,6 +85,25 @@ const _channels = new Map();
 // ---------------------------------------------------------------------------
 
 export const MAX_INBOUND_ATTEMPTS = 3;
+
+/**
+ * Voice calls (owner report 2026-10-02): prefixes the daemon's compiled-in
+ * spoken-reply guidance on a live voice turn, so a turn the voice front desk
+ * handed to OpenClaw is answered in a form fit to be read aloud. Byte-equal
+ * to @securecoms/agent-runtime's VOICE_TURN_PREFIX.
+ */
+export const VOICE_TURN_PREFIX =
+  "This message was spoken on a live voice call and your reply will be read aloud: plain spoken sentences, no markdown, lists, links or code. Use your tools as usual when the question needs them. ";
+
+/** Cap on the daemon-supplied voice guidance carried into the prompt. */
+const VOICE_GUIDANCE_MAX_CHARS = 1000;
+
+/** The guidance of a `voice: true` frame (trimmed, capped), else undefined. */
+export function voiceGuidanceOf(frame) {
+  if (frame?.voice !== true || typeof frame.guidance !== "string") return undefined;
+  const g = frame.guidance.trim().slice(0, VOICE_GUIDANCE_MAX_CHARS);
+  return g || undefined;
+}
 const SEEN_MAX = 5000;
 
 /** Thrown by the inbound handler when register() has not injected a runtime. */
@@ -519,6 +538,7 @@ export class LoopbackChannel {
         senderName: frame.sender_name,
         authorization: frame.authorization,
         timestamp: frame.ts,
+        voiceGuidance: voiceGuidanceOf(frame),
       });
       this._failures.delete(messageId);
       this._failStreak = 0;
@@ -750,6 +770,9 @@ async function _handleInbound({ text, metadata, channel, account, cfg }) {
     // authorization block (isOwnerAuthorization). MLS membership proves a
     // sender is in the room, not that they own this agent.
     CommandAuthorized: commandAuthorized,
+    // Voice turn: OpenClaw folds GroupSystemPrompt into this turn's extra
+    // system prompt (get-reply's extraSystemPromptParts) for any chat type.
+    ...(metadata.voiceGuidance ? { GroupSystemPrompt: `${VOICE_TURN_PREFIX}${metadata.voiceGuidance}` } : {}),
   });
 
   await core.channel.session.recordInboundSession({