refactor: unify inline directives and media fetch

src/auto-reply/reply/agent-runner.ts

@@ -33,9 +33,8 @@ import {
 import { stripHeartbeatToken } from "../heartbeat.js";
 import type { OriginatingChannelType, TemplateContext } from "../templating.js";
 import { normalizeVerboseLevel, type VerboseLevel } from "../thinking.js";
-import { isSilentReplyText, SILENT_REPLY_TOKEN } from "../tokens.js";
+import { SILENT_REPLY_TOKEN } from "../tokens.js";
 import type { GetReplyOptions, ReplyPayload } from "../types.js";
-import { parseAudioTag } from "./audio-tags.js";
 import {
   createAudioAsVoiceBuffer,
   createBlockReplyPipeline,
@@ -48,6 +47,7 @@ import {
   type QueueSettings,
   scheduleFollowupDrain,
 } from "./queue.js";
+import { parseReplyDirectives } from "./reply-directives.js";
 import {
   applyReplyTagsToPayload,
   applyReplyThreading,
@@ -550,28 +550,39 @@ export async function runReplyAgent(params: {
                       !payload.audioAsVoice
                     )
                       return;
-                    const audioTagResult = parseAudioTag(taggedPayload.text);
-                    const cleaned = audioTagResult.text || undefined;
+                    const parsed = parseReplyDirectives(
+                      taggedPayload.text ?? "",
+                      {
+                        currentMessageId: sessionCtx.MessageSid,
+                        silentToken: SILENT_REPLY_TOKEN,
+                      },
+                    );
+                    const cleaned = parsed.text || undefined;
                     const hasMedia =
                       Boolean(taggedPayload.mediaUrl) ||
                       (taggedPayload.mediaUrls?.length ?? 0) > 0;
                     // Skip empty payloads unless they have audioAsVoice flag (need to track it)
-                    if (!cleaned && !hasMedia && !payload.audioAsVoice) return;
                     if (
-                      isSilentReplyText(cleaned, SILENT_REPLY_TOKEN) &&
-                      !hasMedia
+                      !cleaned &&
+                      !hasMedia &&
+                      !payload.audioAsVoice &&
+                      !parsed.audioAsVoice
                     )
                       return;
+                    if (parsed.isSilent && !hasMedia) return;
 
                     const blockPayload: ReplyPayload = applyReplyToMode({
                       ...taggedPayload,
                       text: cleaned,
-                      audioAsVoice:
-                        audioTagResult.audioAsVoice || payload.audioAsVoice,
+                      audioAsVoice: parsed.audioAsVoice || payload.audioAsVoice,
+                      replyToId: taggedPayload.replyToId ?? parsed.replyToId,
+                      replyToTag: taggedPayload.replyToTag || parsed.replyToTag,
+                      replyToCurrent:
+                        taggedPayload.replyToCurrent || parsed.replyToCurrent,
                     });
 
                     void typingSignals
-                      .signalTextDelta(taggedPayload.text)
+                      .signalTextDelta(cleaned ?? taggedPayload.text)
                       .catch((err) => {
                         logVerbose(
                           `block reply typing signal failed: ${String(err)}`,
@@ -735,11 +746,21 @@ export async function runReplyAgent(params: {
       currentMessageId: sessionCtx.MessageSid,
     })
       .map((payload) => {
-        const audioTagResult = parseAudioTag(payload.text);
+        const parsed = parseReplyDirectives(payload.text ?? "", {
+          currentMessageId: sessionCtx.MessageSid,
+          silentToken: SILENT_REPLY_TOKEN,
+        });
+        const mediaUrls = payload.mediaUrls ?? parsed.mediaUrls;
+        const mediaUrl = payload.mediaUrl ?? parsed.mediaUrl ?? mediaUrls?.[0];
         return {
           ...payload,
-          text: audioTagResult.text ? audioTagResult.text : undefined,
-          audioAsVoice: audioTagResult.audioAsVoice,
+          text: parsed.text ? parsed.text : undefined,
+          mediaUrls,
+          mediaUrl,
+          replyToId: payload.replyToId ?? parsed.replyToId,
+          replyToTag: payload.replyToTag || parsed.replyToTag,
+          replyToCurrent: payload.replyToCurrent || parsed.replyToCurrent,
+          audioAsVoice: payload.audioAsVoice || parsed.audioAsVoice,
         };
       })
       .filter(isRenderablePayload);
@@ -775,8 +796,7 @@ export async function runReplyAgent(params: {
 
     const shouldSignalTyping = replyPayloads.some((payload) => {
       const trimmed = payload.text?.trim();
-      if (trimmed && !isSilentReplyText(trimmed, SILENT_REPLY_TOKEN))
-        return true;
+      if (trimmed) return true;
       if (payload.mediaUrl) return true;
       if (payload.mediaUrls && payload.mediaUrls.length > 0) return true;
       return false;

src/auto-reply/reply/reply-directives.ts

@@ -0,0 +1,49 @@
+import { splitMediaFromOutput } from "../../media/parse.js";
+import { parseInlineDirectives } from "../../utils/directive-tags.js";
+import { isSilentReplyText, SILENT_REPLY_TOKEN } from "../tokens.js";
+
+export type ReplyDirectiveParseResult = {
+  text: string;
+  mediaUrls?: string[];
+  mediaUrl?: string;
+  replyToId?: string;
+  replyToCurrent: boolean;
+  replyToTag: boolean;
+  audioAsVoice?: boolean;
+  isSilent: boolean;
+};
+
+export function parseReplyDirectives(
+  raw: string,
+  options: { currentMessageId?: string; silentToken?: string } = {},
+): ReplyDirectiveParseResult {
+  const split = splitMediaFromOutput(raw);
+  let text = split.text ?? "";
+
+  const replyParsed = parseInlineDirectives(text, {
+    currentMessageId: options.currentMessageId,
+    stripAudioTag: false,
+    stripReplyTags: true,
+  });
+
+  if (replyParsed.hasReplyTag) {
+    text = replyParsed.text;
+  }
+
+  const silentToken = options.silentToken ?? SILENT_REPLY_TOKEN;
+  const isSilent = isSilentReplyText(text, silentToken);
+  if (isSilent) {
+    text = "";
+  }
+
+  return {
+    text,
+    mediaUrls: split.mediaUrls,
+    mediaUrl: split.mediaUrl,
+    replyToId: replyParsed.replyToId,
+    replyToCurrent: replyParsed.replyToCurrent,
+    replyToTag: replyParsed.hasReplyTag,
+    audioAsVoice: split.audioAsVoice,
+    isSilent,
+  };
+}

src/auto-reply/reply/reply-payloads.ts

@@ -10,8 +10,23 @@ export function applyReplyTagsToPayload(
   payload: ReplyPayload,
   currentMessageId?: string,
 ): ReplyPayload {
-  if (typeof payload.text !== "string") return payload;
-  const { cleaned, replyToId, hasTag } = extractReplyToTag(
+  if (typeof payload.text !== "string") {
+    if (!payload.replyToCurrent || payload.replyToId) return payload;
+    return {
+      ...payload,
+      replyToId: currentMessageId?.trim() || undefined,
+    };
+  }
+  const shouldParseTags = payload.text.includes("[[");
+  if (!shouldParseTags) {
+    if (!payload.replyToCurrent || payload.replyToId) return payload;
+    return {
+      ...payload,
+      replyToId: currentMessageId?.trim() || undefined,
+      replyToTag: payload.replyToTag ?? true,
+    };
+  }
+  const { cleaned, replyToId, replyToCurrent, hasTag } = extractReplyToTag(
     payload.text,
     currentMessageId,
   );
@@ -20,6 +35,7 @@ export function applyReplyTagsToPayload(
     text: cleaned ? cleaned : undefined,
     replyToId: replyToId ?? payload.replyToId,
     replyToTag: hasTag || payload.replyToTag,
+    replyToCurrent: replyToCurrent || payload.replyToCurrent,
   };
 }
 

src/auto-reply/reply/reply-tags.ts

@@ -1,12 +1,4 @@
-const REPLY_TAG_RE =
-  /\[\[\s*(?:reply_to_current|reply_to\s*:\s*([^\]\n]+))\s*\]\]/gi;
-
-function normalizeReplyText(text: string) {
-  return text
-    .replace(/[ \t]+/g, " ")
-    .replace(/[ \t]*\n[ \t]*/g, "\n")
-    .trim();
-}
+import { parseInlineDirectives } from "../../utils/directive-tags.js";
 
 export function extractReplyToTag(
   text?: string,
@@ -14,31 +6,17 @@ export function extractReplyToTag(
 ): {
   cleaned: string;
   replyToId?: string;
+  replyToCurrent: boolean;
   hasTag: boolean;
 } {
-  if (!text) return { cleaned: "", hasTag: false };
-
-  let sawCurrent = false;
-  let lastExplicitId: string | undefined;
-  let hasTag = false;
-
-  const cleaned = normalizeReplyText(
-    text.replace(REPLY_TAG_RE, (_full, idRaw: string | undefined) => {
-      hasTag = true;
-      if (idRaw === undefined) {
-        sawCurrent = true;
-        return " ";
-      }
-
-      const id = idRaw.trim();
-      if (id) lastExplicitId = id;
-      return " ";
-    }),
-  );
-
-  const replyToId =
-    lastExplicitId ??
-    (sawCurrent ? currentMessageId?.trim() || undefined : undefined);
-
-  return { cleaned, replyToId, hasTag };
+  const result = parseInlineDirectives(text, {
+    currentMessageId,
+    stripAudioTag: false,
+  });
+  return {
+    cleaned: result.text,
+    replyToId: result.replyToId,
+    replyToCurrent: result.replyToCurrent,
+    hasTag: result.hasReplyTag,
+  };
 }

src/auto-reply/types.ts

@@ -31,6 +31,8 @@ export type ReplyPayload = {
   mediaUrls?: string[];
   replyToId?: string;
   replyToTag?: boolean;
+  /** True when [[reply_to_current]] was present but not yet mapped to a message id. */
+  replyToCurrent?: boolean;
   /** Send audio as voice message (bubble) instead of audio file. Defaults to false. */
   audioAsVoice?: boolean;
   isError?: boolean;