replaced telegram with an async version

[tmbo]

Proposed changes:

• use async communication when talking to telegram to avoid blocking sanic workers

[ancalita]

⚡ Nice one!

[github-actions]

🚀 A preview of the docs have been deployed at the following URL: https://11552--rasahq-docs-rasa-v2.netlify.app/docs/rasa

[tmbo]

this looks great, thanks a lot 👍 (I can't approve it since I created the PR but please go ahead and merge it)

[benos]

Hi,

This is failing for me. I get the following error when rasa starts:

And the bot is unresponsive when I type something in Telegram. Downgrading to 3.2.9 fixes the problem.

My credentials look like this:

Thanks!

Ben

[ancalita]

@benos thanks for flagging! I'll open a fix PR asap for this on 3.3.x and hopefully that should fix this issue.

[benos]

Hi @ancalita

That sounds great, but do you have any idea when this fix might be made live? I don't see any change on 3.3.1

Thanks!

Ben

[ancalita]

@benos Apologies for the delay, other priorities took hold 🙈
I've opened this PR, once it gets reviewed and merged, I'll release a new patch 🤞🏻

[benos]

@ancalita Awesome! Many thanks

[benos]

Hi @ancalita,

Sorry to reopen this, but this change has created an additional bug. When triggering an intent like so:

http://localhost:5005/conversations/[redacted]/trigger_intent?output_channel=telegram

This fails with the following error:

{"version":"3.4.1","status":"failure","message":"An unexpected error occurred. Error: asyncio.run() cannot be called from a running event loop","reason":"ConversationError","details":{},"help":null,"code":500}

I assume get_output_channel is re-run by the triggering code.

Best,

Ben

[benos]

@ancalita While I'm at it, might this line cause a leak if called repeatedly?

channel = TelegramOutput(self.access_token)

I've actually noticed a steadily rising CPU consumption which seems linked to calling trigger_intent, but could never really pinpoint the problem.

[ancalita]

@benos I'm so sorry to hear about these new difficulties 🙈 I didn't envisage the potential issues of this change when using the HTTP API. Could you please create a ticket in this Jira board, the new home for OSS issues, with steps for reproducing the error.

I've actually noticed a steadily rising CPU consumption which seems linked to calling trigger_intent

A separate ticket for this would be welcome too 🙏🏻 with the reference to the particular line in the codebase if possible. Then you can link these two tickets as related, so they can inform each other.

[benos]

Hi @ancalita ,

I created two tickets:

https://rasa-open-source.atlassian.net/jira/software/c/projects/OSS/boards/1?modal=detail&selectedIssue=OSS-677&quickFilter=5

https://rasa-open-source.atlassian.net/jira/software/c/projects/OSS/boards/1?modal=detail&selectedIssue=OSS-676&quickFilter=5

The silver lining could be that this might have revealed the origin of this CPU link that has been plaguing us since we launched.

Best,

Ben