Getting CAAL to ask a question with a manual trigger in n8n

[nerd8east-a11y]

Is it possible to get CAAL to ask a question and wait for an answer using a manual trigger in n8n? For example, lets say I ask CAAL for the weather, but do not provide a location. Would it be possible to get CAAL to ask for the missing data point, before moving on with the rest of the workflow?

[cmac86]

I think this is a matter of prompting and having required parameters identified in the tool description (webhook notes). So if the weather_get_forecast() tool requires a city as a parameter and you don't provide one, CAAL should ask you what city in order to run the tool.

If asked to run a tool and all the required information is not provided 
in order to satisfy the required parameters of the tool, first request 
clarification from the user and then run the tool.

Something like that.

I think the tools I've been playing with all have a default if no parameter is provided. I'll have to test this out.

[nerd8east-a11y]

ok, I played around with the add task workflow, and CAAL does ask for additional information when something is missing.
I tried to add another task after CAAL asked for more info, and it seemed to work (CAAL kept a list of two additional tasks I asked to be added), but at the end CAAL only added the first task.

[cmac86]

I thought about your question again. If you mean that in the middle of a workflow, get additional information before proceeding, I don't have an answer to that yet. I have thought about this as well but I haven't spent the time on a solution yet.

[nerd8east-a11y]

i did all you suggested and I cannot get the browser to load the page. I set the .env file like this:

Your machine's IP address

Find it: ip addr show | grep "inet " | grep -v 127.0.0.1

CAAL_HOST_IP=10.0.0.110

HTTPS domain (set to same as CAAL_HOST_IP for LAN, or Tailscale domain)

HTTPS_DOMAIN=10.0.0.110

when I run docker compose --profile https up -d I see this
WARN[0000] No services to build
[+] up 6/6
✔ Network caal-network Created 0.0s
✔ Container caal-kokoro Healthy 22.6s
✔ Container caal-livekit Healthy 12.1s
✔ Container caal-speaches Healthy 12.6s
✔ Container caal-frontend Created 0.1s
✔ Container caal-agent Created 0.1s
Error response from daemon: failed to set up container networking: network 758f5d18ce71bb6f35e26f98d7fcace72051b015ae35a0089a2b7a4912d9c34f not found

but the page https://10.0.0.110:3443 does not load

I tried different combinations for the values of the HTTPS_DOMAIN like
HTTPS_DOMAIN=https://10.0.0.110
HTTPS_DOMAIN=https://10.0.0.110:3443

but still got nothing. Sometimes the livekit server would not start.

Sorry for the back and forth, I am just not sure what else to try.

[cmac86]

The error you're seeing is a Docker networking issue, not a config issue:

Error response from daemon: failed to set up container networking: network ... not found

This happens when containers try to join a network that got partially cleaned up. Let me help you fix it:

1. Full cleanup

docker compose --profile https down
docker network prune -f

2. Your .env should be:

CAAL_HOST_IP=10.0.0.110
HTTPS_DOMAIN=10.0.0.110 # Just the IP, no https:// or port

3. Start fresh

docker compose --profile https up -d --build

4. Check all containers started

docker compose --profile https ps

[nerd8east-a11y]

Good morning. thanks for helping me get the issue sorted out. CAAL is running with the https profile and the new code now. I started testing with the email tool. Here are a few things I have seen:

• The 5 dots on the frontend no longer move side to side when CAAL is thinking. A minor thing, but it made it easier to know that CAAL was processing the request.
• There is a significant increase in the lag time between when CAAL hears a command and when he returns with an answer. I tried just asking the time and it takes him a good 15-20 seconds to report the time. I tried typing in the same "what time is it" question, and the lag seemed shorter but it is still there.
• I tested running the email tool, but it failed at accesing the emails. n8n executed the workflow correctly and the workflow's error handling logic ran properly, so CAAL has access to the tool. I noticed that the credentials are not set, even after providing it during the tool installation. I tried setting them manually, but the workflow seems to not save them for some reason. I tried unpublishing it, setting credentials, and then publishing again, but it did not work. I have a simple test workflow that does the same thing (no webhoook, just a manual trigger) and it is able to reach my email account.

I will try some of the other tools one by one next.

[cmac86]

Hey @nerd8east-a11y,

• I will look into the dots moving side to side. I think it just flashes faster in the middle now. Not sure what happened there.
• The increase in lag time doesn't make sense. What model are you running? Is it slow on just the first request, then faster immediately after? I'm wondering if its just model load time. Then typically Ollama offloads the model after a time period of no use. I've set mine to never offload the model.
• Did you set up the Gmail credential within n8n->Credentials? You need to set up OAuth to Gmail. Its a pain the butt but there is docs in n8n on how to do this.

[nerd8east-a11y]

ok, I will check ollama and see if I need to set the model to not offload.
I checked my gmail credentials and the setup looks like yours. I setup oath2 and all of that on the console api page.

[nerd8east-a11y]

I reinstalled Ollama and set it to never unload the model.

I ran this command to pull the model you trained:

sudo docker exec -it ollama ollama pull hf.co/CoreWorxLab/caal-ministral

when I set the model in the front end CAAL gives me this message:

I encountered an error: hf.co/CoreWorxLab/caal-ministral:latest does not support tools (status code: 400)

I set my model back to ministral-3:8b for now

[nerd8east-a11y]

this is what I see when I open the Get Unread Emails node:
it says "Found credential with no ID. (item 0)" in the error cell

[nerd8east-a11y]

Well, I just checked both gmail nodes and manually set my credentials, published again and it is working now. At this point, you probably think I am crazy. Sorry for the run around, I kept getting the error above all day yesterday and today. I even unpublished, set my credentials and published again yesterday and nothing worked. Not sure what fixed it this time.

[nerd8east-a11y]

One thing I did was to download the workflow data as a JSON file (i did not import it back thought, just exported it). I was getting some weird errors related to some Id and wanted to see which Id n8n was talking about. I could not find anything that made sense. Not saying this was related, but just wanted to mention it just incase. Maybe n8n had an issue importing the code the first time? who knows.

[cmac86]

I don't think you're crazy. This is great info. if you're having issues, others surely will too. I'll watch out for this 'Found credential with no ID' error that you saw in the Gmail node. Maybe something happening with the install. Glad you got it going! Let me know if you have issues with other tools.

[cmac86]

Ok - I found the caal-ministral issue. It should work now that I pushed it to Ollama, not Hugging Face.

Try this:

ollama rm hf.co/CoreWorxLab/caal-ministral:latest
ollama pull coreworxlab/caal-ministral