[GH-ISSUE #65] Сonnection testing error with Ollama #35

Closed
opened 2026-06-06 22:08:09 -04:00 by yindo · 6 comments
Owner

Originally created by @kolovr14 on GitHub (Oct 24, 2025).
Original GitHub issue: https://github.com/vxcontrol/pentagi/issues/65

When running the configuration test, an error occurs: Error unmarshal: invalid character '<' looking for beginning of value.

Image
Originally created by @kolovr14 on GitHub (Oct 24, 2025). Original GitHub issue: https://github.com/vxcontrol/pentagi/issues/65 When running the configuration test, an error occurs: Error unmarshal: invalid character '<' looking for beginning of value. <img width="898" height="521" alt="Image" src="https://github.com/user-attachments/assets/7eb3dfd8-7c3c-4a9f-a03a-ffe6d8682908" />
yindo closed this issue 2026-06-06 22:08:10 -04:00
Author
Owner

@asdek commented on GitHub (Oct 24, 2025):

Hello @kolovr14,

This looks like an issue on the LLM side - I faced something similar when the context size was too low. Please check out the guide with supported models.

Btw, the latency is way too high... I think it won't work well :(

<!-- gh-comment-id:3442411710 --> @asdek commented on GitHub (Oct 24, 2025): Hello @kolovr14, This looks like an issue on the LLM side - I faced something similar when the context size was too low. Please check out [the guide with supported models](https://github.com/vxcontrol/pentagi?tab=readme-ov-file#creating-custom-ollama-models-with-extended-context). Btw, the latency is way too high... I think it won't work well :(
Author
Owner

@kolovr14 commented on GitHub (Oct 24, 2025):

I use llama318b

<!-- gh-comment-id:3442556651 --> @kolovr14 commented on GitHub (Oct 24, 2025): I use llama318b
Author
Owner

@asdek commented on GitHub (Oct 24, 2025):

it's just for testing Ollama connectivity, not for production use...

according to my tests, all models under 32B aren't suitable for Automation mode, though gpt-oss-20b can work for Assistant mode

<!-- gh-comment-id:3442599346 --> @asdek commented on GitHub (Oct 24, 2025): it's just for testing Ollama connectivity, not for production use... according to my tests, all models under 32B aren't suitable for Automation mode, though gpt-oss-20b can work for Assistant mode
Author
Owner

@kolovr14 commented on GitHub (Oct 24, 2025):

I got it. Thank you. I'll try it. No specific settings are required for gpt-oss-20b, please let me know

<!-- gh-comment-id:3442752799 --> @kolovr14 commented on GitHub (Oct 24, 2025): I got it. Thank you. I'll try it. No specific settings are required for gpt-oss-20b, please let me know
Author
Owner

@asdek commented on GitHub (Oct 24, 2025):

I used vLLM to run it, probably it should work on Ollama too.
vLLM version was 0.10.1+gptoss, now you can use something newer

<!-- gh-comment-id:3443021433 --> @asdek commented on GitHub (Oct 24, 2025): I used vLLM to run it, probably it should work on Ollama too. vLLM version was 0.10.1+gptoss, now you can use something newer
Author
Owner

@asdek commented on GitHub (Feb 21, 2026):

Closes #102

<!-- gh-comment-id:3939797229 --> @asdek commented on GitHub (Feb 21, 2026): Closes #102
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: vxcontrol/pentagi#35