Improve natural language of prompt (#19515)

* Make sequence details human-readable so they are used in natural language response

* Cleanup

* Improve prompt and image selection

* Adjust

* Adjust sligtly

* Format time

* Adjust frame selection logic

* Debug save response

* Ignore extra fields

* Adjust docs
This commit is contained in:
Nicolas Mowen
2025-08-15 07:25:49 -06:00
committed by Blake Blackshear
parent 6671984e5a
commit ccbaa74a8b
4 changed files with 70 additions and 27 deletions

View File

@@ -29,12 +29,11 @@ You must use a vision capable model with Frigate. Current model variants can be
The following models are recommended:
| Model | Size | Recommended Features |
| ----------------- | ------ | -------------------- |
| `minicpm-v:8b` | 5.5 GB | Review Summary |
| `qwen2.5vl:3b` | 3.2 GB | Review Summary |
| `gemma3:4b` | 3.3 GB | All Features |
| `llava-phi3:3.8b` | 2.9 GB | All Features |
| Model | Size | Notes |
| ----------------- | ------ | ----------------------------------------------------------- |
| `gemma3:4b` | 3.3 GB | Strong frame-to-frame understanding, slower inference times |
| `qwen2.5vl:3b` | 3.2 GB | Fast but capable model with good vision comprehension |
| `llava-phi3:3.8b` | 2.9 GB | Lightweight and fast model with vision comprehension |
:::note