fix: support AI commands in media captions

app/agentic.py

@@ -437,8 +437,14 @@ async def run_agent_turn(
         context["base_dir"] = config.base_dir
         existing.session.context.clear()
         existing.session.context.update(context)
+        media = await _download_message_media(
+            client,
+            command_message,
+            Path(existing.workspace.name),
+            config.max_media_bytes,
+        )
         return AgentTurnResult(
-            answer=await _run_session(existing.session, invocation.prompt),
+            answer=await _run_session(existing.session, invocation.prompt, media=media),
             handle=existing,
         )
 
@@ -463,6 +469,12 @@ async def run_agent_turn(
             user_template=templates.user,
             template_context=context,
         )
+        media = await _download_message_media(
+            client,
+            command_message,
+            temp_path,
+            config.max_media_bytes,
+        )
         tools = [
             *WebPageTools(
                 timeout_seconds=config.web_page_timeout_seconds,
@@ -501,7 +513,7 @@ async def run_agent_turn(
             session_name=f"telegram:{chat_id}",
             messages=messages,
         )
-        answer = await _run_session(session, invocation.prompt)
+        answer = await _run_session(session, invocation.prompt, media=media)
         return AgentTurnResult(
             answer=answer,
             handle=AgentSessionHandle(session, workspace, set()),
@@ -511,8 +523,11 @@ async def run_agent_turn(
         raise
 
 
-async def _run_session(session: Any, prompt: str) -> str:
-    result = await session.run(prompt)
+async def _run_session(session: Any, prompt: str, *, media: list[Media] | None = None) -> str:
+    if media:
+        result = await session.run(prompt, media=media)
+    else:
+        result = await session.run(prompt)
     if not result.succeeded or result.final_message is None:
         raise AgentError("Agent loop failed") from result.error
     answer = result.final_message.message().strip()

app/commands/dispatcher.py

@@ -59,10 +59,12 @@ class CommandDispatcher:
         account: "AccountState",
     ) -> bool:
         text = getattr(message, "text", None)
-        if not isinstance(text, str):
+        caption = getattr(message, "caption", None)
+        command_text = text if isinstance(text, str) else caption
+        if not isinstance(command_text, str):
             return False
 
-        parsed = parse_command(text)
+        parsed = parse_command(command_text)
         if parsed is None:
             return False
 
@@ -85,7 +87,7 @@ class CommandDispatcher:
             message=message,
             command=definition,
             prefix=parsed.prefix,
-            raw_text=text,
+            raw_text=command_text,
             args=parsed.args,
             alias=parsed.alias,
         )

app/handlers.py

@@ -18,4 +18,4 @@ def register_common_handlers(client: "Client", manager: "AccountManager", accoun
     async def command_handler(client_: "Client", message) -> None:
         await dispatcher.dispatch(client_, message, manager, account)
 
-    client.add_handler(MessageHandler(command_handler, filters.text))
+    client.add_handler(MessageHandler(command_handler, filters.text | filters.caption))