Version 1.1.16
- Update dependencies
Thanks for the app — works great with a local Ollama box.
One thing that would help a lot: recent models (Gemma 4, Qwen3-VL, Qwen3.5) have reasoning enabled by default. For automation use cases like “describe who’s at my front door” the reasoning tokens are pure overhead — the flow just needs one short sentence back.
Quick benchmark on my machine, same model, same prompt:
| eval count | total | |
|---|---|---|
| default | 85 tokens | 5.6s |
think: false |
10 tokens | 0.86s |
Same answer, 6x faster. With an image attached the difference is even bigger.
Would it be possible to add "think": false as a toggle on the flow card?
I think I’ll disable it on the Flowcards by default, since that would be unnecessary, and make it a toggle on the chat widget. I’ll publish the update soon!
Hi @silsil ,
Can you please try the test version?
Works like a charm! My flow now runs in 3s instead of 7s
One more thing: my Ollama instance sits behind a reverse proxy, so it’s reachable over HTTPS instead of plain HTTP on the internet. Would it be possible to add a “Use HTTPS” (or “Use TLS”) checkbox in the settings, next to the IP-adres field?
Happy to test a beta again if that helps.