Hi,
I have localhost up and running and Ollama selected, but the dropdown list doesnt show any models. I have qwen2.5-coder:32b-base-q2_K installed.
It seems correctly installed, since in CMD when I type ollama run qwen2.5-coder:32b-base-q2_K I get the message prompt and can run command queries.
When I run ollama ps in cmd within my bolt.new-any-llm directory, there’s nothing listed. Does Quen2.5 need to be saved in a specific directory?
Thanks.
Hey @itqandas, there is a small bug we are working through right now where the Ollama list doesn’t load because of stale state. Could you try selecting another provider and then going back to Ollama? When I’ve had this happen that fixes it for me.
I dont have any other providers installed. Can I override this by adding a fake entry, and in which file would I do this ?
If I understand correctly, ollama ps isn’t showing you any models running after trying to use ollama run with the 32B model. The 32B model is huge. Are you sure that your hardware is capable of running the model? When the ollama run command completes you should be able to chat with it. How’s that going?
I’m not sure what you’re talking about with a “fake entry”, but that’s not really a good approach to these things regardless. Try to understand your context (whether the model is running), explore it (chat with the LLM), and if all of that is going well then it’s time to determine if there’s an issue at the oTToDev level. Hope this helps.
Yes, as Cole mentioned there’s a bug in Ottodev. Thats why I asked for the work around to be able to select then go back again (refesh).
I dont know why ollama ps isnt showing the model, but the run command is working.
Ps would work after the run command, which will kick off the model just FYI. Taking your issue into account for ollama support!
I had a similar issue like this today and it was somewhat on me, as I am new to Linux. I fixed it by doing a fresh install of Ubuntu since I did not know how to clean up my path variable, I had multiple user shell (I don’t know what this means) in Ollama.service.
After reinstalling Ollama I was able to run ollama command from anywhere on my system.
As for the Models not showing up, I added my computer ip to “bolt.new-any-llm/app/utils/cat constants.ts”, but be warned it is not an approved answer.
const getOllamaBaseUrl = () => {
const defaultBaseUrl = import.meta.env.OLLAMA_API_BASE_URL || ‘http://192.168.1.49:11434’;
// Check if we’re in the browser
if (typeof window !== ‘undefined’) {
// Frontend always uses localhost
return defaultBaseUrl;
Can you try with localhost or 127.0.0.1 instead of the 192.168.. address? It’s possible that your ollama instance is only running locally, which won’t expose the server on that 192 address.
Tips
pnpm run dev, and ollama is running on the same machine, use 127.0.0.1.Hope this helps
Also, you will want to set these base urls within an .env.local file (duplicate .env.example as a start if need be), versus modifying the get methods in constants.ts. That will give you problems down the line for sure as the project evolves, and the .env files will be the “root” of where you want to set values based on your specific environment.
I run Ollama manually using this command:
OLLAMA_ORIGINS=“*” OLLAMA_TMPDIR=/home/user/tmp OLLAMA_HOST=0.0.0.0 ollama serve
This enables me to run Ollama on another host and access it without any issues.
You can see the options using
% ollama help serve
Start ollama
Usage:
ollama serve [flags]
Aliases:
serve, start
Flags:
-h, --help help for serve
Environment Variables:
OLLAMA_DEBUG Show additional debug information (e.g. OLLAMA_DEBUG=1)
OLLAMA_HOST IP Address for the ollama server (default 127.0.0.1:11434)
OLLAMA_KEEP_ALIVE The duration that models stay loaded in memory (default “5m”)
OLLAMA_MAX_LOADED_MODELS Maximum number of loaded models per GPU
OLLAMA_MAX_QUEUE Maximum number of queued requests
OLLAMA_MODELS The path to the models directory
OLLAMA_NUM_PARALLEL Maximum number of parallel requests
OLLAMA_NOPRUNE Do not prune model blobs on startup
OLLAMA_ORIGINS A comma separated list of allowed origins
OLLAMA_SCHED_SPREAD Always schedule model across all GPUs
OLLAMA_TMPDIR Location for temporary files
OLLAMA_FLASH_ATTENTION Enabled flash attention
OLLAMA_LLM_LIBRARY Set LLM library to bypass autodetection
OLLAMA_GPU_OVERHEAD Reserve a portion of VRAM per GPU (bytes)
OLLAMA_LOAD_TIMEOUT How long to allow model loads to stall before giving up (default “5m”)
will try i jus posted about this error
Thanks for the info, but seems cumbersome having to run on another host. I want to keep it simple and keep within the threshold of initial instructions.
@ColeMedin : were you able to fix the bug and does it require a new reinitialisation?
Yes I believe this has been fixed, just have to pull the latest code and rerun the container! Please let me know if it persists.
I did a ‘git pull’ and re-ran “pnpm run dev”, but unfortunately the Gwen model still isn’t showing up next to Ollama dropdown. I also cleared my browser cache, but no success.
curl http://localhost:11434/api/tags is without error and .env config is set.
I am also running it without Docker.
any solution to the missing model?
I’m running qwen2.5-coder:32b-base-q2_K, but I assume I don’t need to add the model to the utils\constants.ts file as before since it seems to now be retreiving all Ollama models, correct?:
name: 'Ollama',
staticModels: [],
getDynamicModels: getOllamaModels,
getApiKeyLink: 'https://ollama.com/download',
labelForGetApiKey: 'Download Ollama',
icon: 'i-ph:cloud-arrow-down', },
Thanks for sharing this @gngglobetech!
As far as I understand, @gngglobetech solution might have worked for other device connection within same network. Since I don’t have this option, I never tried it.
Despite not being able to spend time on this since about a week, I now got around to doing the latest ‘git pull’ and I now see the model in the Ollama dropdown, so thanks to whoever solved this. 
I wanted to mark the topic as SOLVED, but I couldnt find any way to do this within this ThinkTank Ottomator forum. Could we maybe implement this (new) feature, so that whoever created the topic can mark as solved and there appears a green tick at the top of the thread?
My Ollama model now generates the code, but the artifact doesn’t show the preview yet for me. But that could just be either a hardware issue on my end (due to lack of memory), for others the prompt just needs to be followed up with another prompt to finish the coding.
Maybe you could add [SOLVED] at the start of your title? I think that is a good solution but yeah I can look into if there is a Discourse plugin for this kind of thing!
If you aren’t seeing a preview that generally means the LLM hallucinated either bad code or bad commands. A lot of time you can see an error in the Bolt terminal where it tried to start the site.
