<![CDATA[Discourse AI (translation)]]>Hello.

Some time ago, I came across a blog post from Discourse where they present a translation feature using the AI plugin. It basically enables that each user can converse seamlessly in a chosen language: https://blog.discourse.org/2026/03/building-for-every-language/

The Discourse AI plugin is now bundled and can be enabled manually for available features.

As I take care of a multilingual forum, I was wondering if it is doable to have the AI translation run with a Cloudron Ollama instance:

  • Did someone manage connect a local Ollama instance to the Discourse AI Plugin ?
  • What would be a recommended requirement in RAM and model (SotA ?) to make this work gracefully ?
  • Otherwise, which third-party provider would you recommend using for a low traffic Discourse site ?

Thanks.

]]>https://forum.cloudron.io/topic/15646/discourse-ai-translationRSS for NodeSat, 08 Aug 2026 07:36:32 GMTThu, 25 Jun 2026 19:54:03 GMT60<![CDATA[Reply to Discourse AI (translation) on Wed, 08 Jul 2026 11:51:04 GMT]]>Hello @mononym
I did not yet look into this myself.

]]>https://forum.cloudron.io/post/126546https://forum.cloudron.io/post/126546Wed, 08 Jul 2026 11:51:04 GMT<![CDATA[Reply to Discourse AI (translation) on Wed, 08 Jul 2026 10:33:02 GMT]]>@james I would suggest adding a section in the Ollama or Discourse documentation explaining how to reach the Ollama service, internally and externally. For now, I only test whatever comes to my mind but unfortunately, I can't figure it out. Feels like magic 🙂

]]>https://forum.cloudron.io/post/126543https://forum.cloudron.io/post/126543Wed, 08 Jul 2026 10:33:02 GMT<![CDATA[Reply to Discourse AI (translation) on Mon, 06 Jul 2026 18:37:41 GMT]]>First trial and error.

Added a new LLM entry in /admin/plugins/discourse-ai/ai-llms/new?llmTemplate=none.

  • Named it "Ollama" and selected OpenAI as provider and qwen3.5:2b (pulled) as model.
  • Set the URL to the Ollama instance and put the API key from it's cat /app/data/.api_key.
  • Chose QwenTokenizer and 262000 context.
  • Saved settings with the Submit button.

The "Run test" fails with

Non-streaming request failed: <html> <head><title>405 Not Allowed</title></head> <body> <center><h1>405 Not Allowed</h1></center> <hr><center>nginx/1.24.0 (Ubuntu)</center> </body> </html>
  • Changed URL to Ollama's ifconfig | grep "inet " | grep -Fv 127.0.0.1 | awk '{print $2}', like http://172.18.16.111:11434

The "Run test" fails with "Internal Server Error"

Jul 06 20:35:43 172.18.0.1 - - [06/Jul/2026:18:35:43 +0000] "POST /admin/plugins/discourse-ai/ai-llms/test.json HTTP/1.1" 500 46 "https://eupalinos.eu/admin/plugins/discourse-ai/ai-llms/1/edit" "Mozilla/5.0 (X11; Linux x86_64; rv:152.0) Gecko/20100101 Firefox/152.0"
Jul 06 20:35:43 Started POST "/admin/plugins/discourse-ai/ai-llms/test.json" for 185.244.213.58 at 2026-07-06 18:35:43 +0000
Jul 06 20:35:43 Processing by DiscourseAi::Admin::AiLlmsController#test as JSON
Jul 06 20:35:43   Parameters: {"ai_llm" => {"max_prompt_tokens" => "262000", "max_output_tokens" => "", "ai_secret_id" => "2", "tokenizer" => "DiscourseAi::Tokenizer::QwenTokenizer", "url" => "http://172.18.16.111:11434", "display_name" => "Ollama", "name" => "qwen3.5:2b", "provider" => "open_ai", "vision_enabled" => "false", "input_cost" => "", "output_cost" => "", "cached_input_cost" => "", "cache_write_cost" => "0", "provider_params" => {"organization" => "", "disable_native_tools" => "", "reasoning_effort" => "default", "disable_streaming" => "false", "service_tier" => "default"}}}
Jul 06 20:35:43 Completed 500 Internal Server Error in 22ms (ActiveRecord: 0.0ms (0 queries, 0 cached) | GC: 0.0ms)
Jul 06 20:35:44 172.18.0.1 - - [06/Jul/2026:18:35:44 +0000] "POST /message-bus/99cb84f5c8754d6caf75a4140828e566/poll HTTP/1.1" 200 2 "https://eupalinos.eu/admin/plugins/discourse-ai/ai-llms/1/edit" "Mozilla/5.0 (X11; Linux x86_64; rv:152.0) Gecko/20100101 Firefox/152.0"

More reading:
https://forum.cloudron.io/topic/14645/how-to-configure-ollama-in-openwebui
https://forum.cloudron.io/topic/13066/auto-tagging-ollama-setup

]]>https://forum.cloudron.io/post/126483https://forum.cloudron.io/post/126483Mon, 06 Jul 2026 18:37:41 GMT<![CDATA[Reply to Discourse AI (translation) on Mon, 06 Jul 2026 16:36:56 GMT]]>Came across the topic Self-Hosting an OpenSource LLM for DiscourseAI and will maybe give it a try with an Ollama instance if possible.

]]>https://forum.cloudron.io/post/126477https://forum.cloudron.io/post/126477Mon, 06 Jul 2026 16:36:56 GMT