xml output for #say miss xml header

[KrionHY]

I use adhearsion with asterisk/UniMRCP. when I call #say function, the xml content sent to asterisk miss xml header. here is out from log:

[2019-08-06 17:13:13.882] DEBUG Adhearsion::Call: b84e7088-655b-4763-a6e5-fb37c60391bb@: Executing command #<Punchblock::Component::Output target_call_id=nil, target_mixer_name=nil, component_id=nil, source_uri=nil, domain=nil, transport=nil, timestamp=Tue, 06 Aug 2019 17:13:13 +0800, request_id="cc7e2e15-b2d2-4e72-ac82-051393063bd4", headers={}, voice=nil, interrupt_on=nil, start_offset=nil, start_paused=nil, repeat_interval=nil, repeat_times=nil, max_time=nil, renderer="unimrcp", render_documents=[#<Punchblock::Component::Output::Document target_call_id=nil, target_mixer_name=nil, component_id=nil, source_uri=nil, domain=nil, transport=nil, timestamp=Tue, 06 Aug 2019 17:13:13 +0800, url=nil, content_type="application/ssml+xml", value=<speak xmlns="http://www.w3.org/2001/10/synthesis" version="1.0" xml:lang="en-US">&#x4F60;&#x597D;&#xFF0C;&#x5927;&#x5BB6;&#x597D;</speak>>]>

I miss something, or is that a bug?

OS: MAC OS

[benlangfeld]

And then what happens? Does it break? What do you see in Asterisk logs?

[KrionHY]

Asterisk didn't log any errors, and foward the xml content to UniMRCP server, but my TTS provider (https://ai.baidu.com/docs#/BSIC-MRCP/top) get the xml content which miss xml header as plain text, and synthesize whole xml content to speech.

[bklang]

By "XML Header", can you be more specific what you are referring to? This sounds more like a configuration problem between UniMRCP and Baidu, rather than something in Adhearsion. Can you get UniMRCP to work from extensions.conf to verify that Asterisk is configured correctly?

[KrionHY]

I'm sorry about my terrible English. I try to describe this issue more clearly.

Yes, UniMRCP can work with Asterisk from extensions.conf.

the xml content from Adhearsion will be forward to Asterisk, and Asterisk forward to UniMRCP/ Baidu.

Baidu expect the xml content following MRCP protocol like:
08-07 00:58:33:118 [baidu/speech-arch/open-mrcp-tts/tts_channel.cpp:439][tts_channel_t::push_request_msg][logid:6c276be2b86b11e9_0] send TTS request msg: {"text":"<?xml version=\"1.0\" ?>\n<speak>\nHello World\n</speak>\n","per":100,"vol":5,"xml":1,"aue":5,"rate":0,"pit":5,"spd":5,"user_sn":"6c276be2b86b11e9_0","lan":"en_US"}

but the xml content from Adhearsion like:
08-07 01:14:42:349 [baidu/speech-arch/open-mrcp-tts/tts_channel.cpp:439][tts_channel_t::push_request_msg][logid:adf7b962b86d11e9_0] send TTS request msg: {"text":"<speak xmlns=\"http://www.w3.org/2001/10/synthesis\" version=\"1.0\" xml:lang=\"en-US\">Hello World</speak>","per":100,"vol":5,"xml":1,"aue":5,"rate":0,"pit":5,"spd":5,"user_sn":"adf7b962b86d11e9_0","lan":"en_US"}

The text value miss the xml header:
<?xml version="1.0" ?>

Baidu synthesize whole text value to speech, rather than Hello World string .

The text value is from Adhearsion:
[2019-08-06 17:13:13.882] DEBUG Adhearsion::Call: b84e7088-655b-4763-a6e5-fb37c60391bb@: Executing command #<Punchblock::Component::Output target_call_id=nil, target_mixer_name=nil, component_id=nil, source_uri=nil, domain=nil, transport=nil, timestamp=Tue, 06 Aug 2019 17:13:13 +0800, request_id="cc7e2e15-b2d2-4e72-ac82-051393063bd4", headers={}, voice=nil, interrupt_on=nil, start_offset=nil, start_paused=nil, repeat_interval=nil, repeat_times=nil, max_time=nil, renderer="unimrcp", render_documents=[#<Punchblock::Component::Output::Document target_call_id=nil, target_mixer_name=nil, component_id=nil, source_uri=nil, domain=nil, transport=nil, timestamp=Tue, 06 Aug 2019 17:13:13 +0800, url=nil, content_type="application/ssml+xml", value=<speak xmlns="http://www.w3.org/2001/10/synthesis" version="1.0" xml:lang="en-US">Hello World</speak>>]>

So, I think the issue has three causes:
1, Do Adhearsion miss to send xml header?
2, Should Asterisk add the xml header?
3, Baidu handle the text value incorrectly.

which cause I should to follow?

Thanks!