Discourse Chatbot 🤖

Plugin
943

Thank you very much for your help and update @merefield

1 Like
944

That said there is a bug with this release. I will fix and update here. Do not update for now.

945

that’s fixed: FIX: do not exlude single user prompts in submission by merefield · Pull Request #120 · merefield/discourse-chatbot · GitHub

thanks for your patience.

3 Likes
946

I updated the plugin. There is no problem with the chatbot responding. But now the chatbot is giving an error saying that it cannot see the content of the image. I wanted to give you feedback. My vision setting: directly

1 Like
947

Thanks for the report. I’ll take a look

1 Like
948

That should be fixed now as of:

But I noticed something interesting in my testing.

In Discourse, there’s a slight delay in images being fully reflected in a Post’s data so if the bot replies too quickly it can miss seeing that there’s an image.

The solution is to up this value:

chatbot reply job time delay - I’ve tested it at 0 and 2 - and 0 is often too fast, so slow things down a bit.

If you are always refering to an image in a prior Post you should not have a problem, it’s only if you are Posting an image and then referring to it in the same post, e.g. “describe this image”

1 Like
949

This setting was set to 1, now I made it 2. I will update and try. Thanks for the fix.

1 Like
950

I’ve made the default 2 and the max 5 now, just in case some Discourse forums need more time to process images.

1 Like
951

I updated the plugin and tried 4 images for testing, but it did not give any errors. Thank you for the quick fix.

1 Like
952

To clarify what I mean, here’s an example:

"The image captures a striking blue sports wagon with sleek, aggressive lines, parked in a parking lot and the caption suggests the car is a BMW vehicle." (Captioned by AI)
"The image captures a striking blue sports wagon with sleek, aggressive lines, parked in a parking lot and the caption suggests the car is a BMW vehicle." (Captioned by AI)638×844 144 KB

I don’t want to share the mess when reply delay is 0 :smiley:

2 Likes
953

Hi! Are there any plans to add support for o1? it was released to tier 4 customers today. At least one thing o1 complains about is the:

Chatbot: There was a problem with Chat Completion: status: 400, message: Unsupported value: ‘messages[0].role’ does not support ‘system’ with this model.

1 Like
954

This is model is a pain :sweat_smile:

https://platform.openai.com/docs/guides/reasoning/beta-limitations

are you sure you’d find it useful with no tools, no bespoke character (as no system prompt) and such a high $$$ value?

1 Like
955

o1-preview and o1-mini support added.

You must select basic bot type (as tools are not supported by Open AI yet!).

e.g.

This is a screenshot of a webpage or an application, displaying a chatbot interface with a chat history and text input fields. (Captioned by AI)
This is a screenshot of a webpage or an application, displaying a chatbot interface with a chat history and text input fields. (Captioned by AI)1530×486 60.8 KB

NB at time of writing you must have Tier 4 to access these beta models. Watch your email for when it becomes available to you.

2 Likes
956

thank you very much!

oh, absolutely. a lot of our use cases is for helping with scientific questions & problems and o1 is miles ahead of 4o. In fact, we never switched from 4 to 4o despite the latter being cheaper because it was visibly worse at science.

1 Like
957

@hiddenseal (and anyone who likes science stuff)

Found a workaround for using Maths equations with o1 without needing a system prompt:

  • Use the Maths Plugin
  • Tell Chatbot how to format equations and formulae

e.g.:

if x to the power of 5 minus 1 is zero, what is x?

Use the following formatting for the answer:

$$
equations and formulae
$$

Normally you’d put that formatting stuff in the system prompt of course (if your site was a regular user of bots for maths). But o1 doesn’t yet have that facility (yet?) so you need to be explicit at point of request. Seems to work though :rocket:

The image displays a snippet of a text-based conversation in a messaging app, indicating that it's about art and specifically, a conversation about the equation x^2 + y^2 = 5. (Captioned by AI)
The image displays a snippet of a text-based conversation in a messaging app, indicating that it's about art and specifically, a conversation about the equation x^2 + y^2 = 5. (Captioned by AI)818×1518 138 KB

2 Likes
958

After the last update I am getting the following error. I wanted to inform you.

Chatbot: There was a problem with Chat Completion: status: 400, message: Unrecognized request argument supplied: max_completion_tokens

959

Again, you must keep up with the latest API version

https://platform.openai.com/docs/api-reference/chat/create#chat-create-max_tokens

max_tokens is deprecated and it’s being replaced it with max_completion_tokens

NB o1-mini won’t work with max_tokens at all.

Make sure you are on the latest Azure API.

If you are not able to do that, create a branch on your fork that corresponds to the commit prior to the change and git clone that fork and branch in app.yml.

2 Likes
960

I just tested Chatbot with the default brand new 3B llama3.2 locally and I must say it’s INCREDIBLE.

Tools appear to work and it’s very fast.

The image shows a screenshot of a chatbot conversation, where a user has asked about the root of the number 65. (Captioned by AI)
The image shows a screenshot of a chatbot conversation, where a user has asked about the root of the number 65. (Captioned by AI)409×637 16.6 KB

Great work Meta!

This is a fantastic option for development or experimentation at low cost or where you have limited internet.

Settings:

This image shows a screenshot of a chatbot interface with three different chatbot types and their respective chatbot open ai model high trust trust model high trust model custom model name advanced users only high trust users, custom model name advanced users only high trust users, name of model - (Captioned by AI)
This image shows a screenshot of a chatbot interface with three different chatbot types and their respective chatbot open ai model high trust trust model high trust model custom model name advanced users only high trust users, custom model name advanced users only high trust users, name of model - (Captioned by AI)724×528 17.9 KB

(this model dropdown is ignored here as you’ve specified the option of a custom model)

You must have ollama installed and running

4 Likes
961

@merefield

Thanks for this great plugin, I am working on a self-hosting version of our site so I can play around with it, I am aiming to replicate what you did here, Building a technical support chatbot

I am following the instructions, added

- git clone https://github.com/merefield/discourse-chatbot.git

to my app.yml.

However, when I ./launcher rebuild app I get an error

Gem::LoadError: can’t activate multipart-post-2.2.3, already activated multipart-post-2.4.0 (Gem::LoadError)

Full Error Messages

I, [2024-10-04T14:03:14.844712 #1] INFO – : > exec chpst -u redis -U redis /usr/bin/redis-server /etc/redis/redis.conf

I, [2024-10-04T14:03:14.846521 #1] INFO – : > sleep 10

2259:C 04 Oct 2024 14:03:14.854 # oO0OoO0OoO0Oo Redis is starting oO0OoO0OoO0Oo

2259:C 04 Oct 2024 14:03:14.854 # Redis version=7.0.7, bits=64, commit=00000000, modified=0, pid=2259, just started

2259:C 04 Oct 2024 14:03:14.854 # Configuration loaded

2259:M 04 Oct 2024 14:03:14.855 * monotonic clock: POSIX clock_gettime

2259:M 04 Oct 2024 14:03:14.855 # Warning: Could not create server TCP listening socket *:6379: bind: Address already in use

2259:M 04 Oct 2024 14:03:14.855 # Failed listening on port 6379 (TCP), aborting.

I, [2024-10-04T14:03:24.849199 #1] INFO – :

I, [2024-10-04T14:03:24.849755 #1] INFO – : > cd /var/www/discourse && su discourse -c ‘bundle exec rake db:migrate’

rake aborted!

Gem::LoadError: can’t activate multipart-post-2.2.3, already activated multipart-post-2.4.0 (Gem::LoadError)

/var/www/discourse/lib/plugin_gem.rb:25:in `load’

/var/www/discourse/lib/plugin/instance.rb:824:in `gem’

/var/www/discourse/plugins/discourse-zendesk-plugin/plugin.rb:13:in `activate!’

/var/www/discourse/lib/plugin/instance.rb:767:in `instance_eval’

/var/www/discourse/lib/plugin/instance.rb:767:in `activate!’

/var/www/discourse/lib/discourse.rb:348:in `block in activate_plugins!’

/var/www/discourse/lib/discourse.rb:345:in `each’

/var/www/discourse/lib/discourse.rb:345:in `activate_plugins!’

/var/www/discourse/config/application.rb:232:in `block in class:Application’

/var/www/discourse/lib/plugin.rb:6:in `initialization_guard’

/var/www/discourse/config/application.rb:232:in `class:Application’

/var/www/discourse/config/application.rb:75:in `module:Discourse’

/var/www/discourse/config/application.rb:74:in `<top (required)>’

/var/www/discourse/Rakefile:7:in `<top (required)>’

/var/www/discourse/vendor/bundle/ruby/3.3.0/gems/rake-13.2.1/exe/rake:27:in `<top (required)>’

/usr/local/bin/bundle:25:in `load’

/usr/local/bin/bundle:25:in `’

(See full trace by running task with --trace)

I, [2024-10-04T14:03:33.989382 #1] INFO – :
…

I, [2024-10-04T14:03:33.989950 #1] INFO – : Terminating async processes

I, [2024-10-04T14:03:33.989982 #1] INFO – : Sending INT to HOME=/var/lib/postgresql USER=postgres exec chpst -u postgres:postgres:ssl-cert -U postgres:postgres:ssl-cert /usr/lib/postgresql/13/bin/postmaster -D /etc/postgresql/13/main pid: 36

I, [2024-10-04T14:03:33.990027 #1] INFO – : Sending TERM to exec chpst -u redis -U redis /usr/bin/redis-server /etc/redis/redis.conf pid: 104

104:signal-handler (1728050613) Received SIGTERM scheduling shutdown…

2024-10-04 14:03:33.990 UTC [36] LOG: received fast shutdown request

2024-10-04 14:03:33.991 UTC [36] LOG: aborting any active transactions

2024-10-04 14:03:33.996 UTC [36] LOG: background worker “logical replication launcher” (PID 51) exited with exit code 1

2024-10-04 14:03:33.997 UTC [46] LOG: shutting down

104:M 04 Oct 2024 14:03:34.007 # User requested shutdown…

104:M 04 Oct 2024 14:03:34.007 * Saving the final RDB snapshot before exiting.

2024-10-04 14:03:34.023 UTC [36] LOG: database system is shut down

104:M 04 Oct 2024 14:03:34.025 * DB saved on disk

104:M 04 Oct 2024 14:03:34.025 # Redis is now ready to exit, bye bye…

FAILED

Pups::ExecError: cd /var/www/discourse && su discourse -c ‘bundle exec rake db:migrate’ failed with return #<Process::Status: pid 2262 exit 1>

Location of failure: /usr/local/lib/ruby/gems/3.3.0/gems/pups-1.2.1/lib/pups/exec_command.rb:132:in `spawn’

exec failed with the params {“cd”=>“$home”, “tag”=>“migrate”, “hook”=>“db_migrate”, “cmd”=>[“su discourse -c ‘bundle exec rake db:migrate’”]}

bootstrap failed with exit code 1

** FAILED TO BOOTSTRAP ** please scroll up and look for earlier error messages, there may be more than one.

./discourse-doctor may help diagnose the problem.

289408d3c0350809f88a235d9c3ae0b23074f4e1a727748bbc3ab3b205e18115

==================== END REBUILD LOG ====================

Failed to rebuild app.

Any ideas if this is related to the plugin?

1 Like
962

Chatbot depends on gem multipart-post-2.4.0.

Looks like your zendesk plugin is lagging behind?

I would contact the author and ask them to update it or PR to the zendesk plugin :slight_smile:

Alternatively you could fork Chatbot and see if it works with the earlier version.