sshh12 / llm_convo

Use ChatGPT over Twilio to create an AI phone agent (works for incoming or outgoing calls).
MIT License
98 stars 19 forks source link

Setup? #3

Open FreshImmuc opened 1 year ago

FreshImmuc commented 1 year ago

Hey, how exactly do i set this up. Looks really cool! What services do i need to buy in twilio? Is there any instructions?

sshh12 commented 1 year ago

Hey @FreshImmuc, high-level setup is:

  1. Create a Twilio account
  2. Buy a phone number
  3. Go to Phone Numbers > Manage > Active Numbers > Configure and set the "a call comes in" webhook to https://<your app id>.ngrok.app/incoming-voice chrome_VZSfJHN6FV
sshh12 commented 1 year ago

For getting the ngrok server, you can just do set it up here: https://ngrok.com/ with ngrok http <port>

Abe-Telo commented 1 year ago

I was programming my own Twilio with chatGPT system, with appointments settings email etc when i found your code, I decided to give it a shot, However, i see that nothing happens when calls come in, the server see the calls, however it just holds the calls. I see the log as a 405. "GET /incoming-voice HTTP/1.1" 405 308 0.001295 Whats the meaning of this? i assume i did something wrong at setup. i used your sample script, is there anything else i need to do besides the keys for openai and twillio? i also noticed the naming of the folder lim_convo. In reality its convo in your scripts. Probably a good idea to update that. The requirements where off as well. i had to do alot on setup. currently i have the path to tws = TwilioServer(remote_host="0.0.0.0", port=5000, static_dir=r"/root/TTS/llm_convo-main/llm_convo/") at first when i did not know what the issue is, i renamed the files in agent and conversation and anywhere else i sew convo to lim_convo.

sshh12 commented 1 year ago

Hopefully, most of these should be fixed. @Abe-Telo you still need help setting this up?

trekrollercoaster commented 1 year ago

I was programming my own Twilio with chatGPT system, with appointments settings email etc when i found your code, I decided to give it a shot, However, i see that nothing happens when calls come in, the server see the calls, however it just holds the calls. I see the log as a 405. "GET /incoming-voice HTTP/1.1" 405 308 0.001295 Whats the meaning of this? i assume i did something wrong at setup. i used your sample script, is there anything else i need to do besides the keys for openai and twillio? i also noticed the naming of the folder lim_convo. In reality its convo in your scripts. Probably a good idea to update that. The requirements where off as well. i had to do alot on setup. currently i have the path to tws = TwilioServer(remote_host="0.0.0.0", port=5000, static_dir=r"/root/TTS/llm_convo-main/llm_convo/") at first when i did not know what the issue is, i renamed the files in agent and conversation and anywhere else i sew convo to lim_convo.

Hi, I think the issue might be with the remote_host. Are you using ngrok? If you are using ngrok, it should resolve this problem. The remote_host here (without http or https) will be assembled as wss://{remote_host}/audiostream and submitted to Twilio, so the remote_host needs to be a publicly accessible address. I hope this helps you.

Abe-Telo commented 6 months ago

Hopefully, most of these should be fixed. @Abe-Telo you still need help setting this up?

I am reattempting your installation and get back to you; this time, my environment is pycharm on Windows.

I am personally new to Pycharm, so no luck yet with the isolating environment. In linux i did source /directory here. I will make some research and continue tonight.

Abe-Telo commented 6 months ago

On WSL (windows), I had some bumps, but in Pycharm and in WLS, I finally got it working, Hopefully this helps someone.

Here is my bumps. image

To resolve using WSL ( A linux terminal for windows ) you can download here PyCharm will not let you install evrything.

cd into
cd /mnt/c/Users/user/PycharmProjects/TTS/llm2/

source .venv/bin/activate

I had a issue with not having weel so pip install wheel

for environment, as I understand.

pip install git+https://github.com/sshh12/llm_convo.git

to get the files in directory.

git clone https://github.com/sshh12/llm_convo.git

pip install git+https://github.com/openai/whisper.git

sudo apt update sudo apt install ffmpeg

Then i had to remove the environment, and redo the processes for some reason ( chat gpt told me to do so.)

Finlay got something working with no errors. My next step is figuring out how to use the environment in pycharm or setup Ngrok on my WLS

Update: I got nigrok working but no idea why there is no audio working,