[Support]: everytime i restart frigate, it try to download huggingface

[Spirch]

Describe the problem you are having

i did connect to the internet to do the initial setup and i did a restart while connected to the internet and wait a few minutes to make sure everything was fine

the moment i disconnect the pc from the internet and do a restart the huggingface try to re-download itself

since the pc doesnt have access to the internet, it fail/retry 5 times

Version

0.16.3

What browser(s) are you using?

No response

Frigate config file

not needed i think?

Relevant Frigate log output

'(MaxRetryError('HTTPSConnectionPool(host=\'huggingface.co\', port=443): Max retries exceeded with url: /jinaai/jina-clip-v1/resolve/main/tokenizer_config.json (Caused by NewConnectionError("HTTPSConnection(host=\'huggingface.co\', port=443): Failed to establish a new connection: [Errno 101] Network is unreachable"))'), '(Request ID: c3e7f1e4-68e9-467a-a3e7-e7ea720406cf)')' thrown while requesting HEAD https://huggingface.co/jinaai/jina-clip-v1/resolve/main/tokenizer_config.json

Relevant go2rtc log output

not needed i think?

FFprobe output from your camera

not needed i think?

Frigate stats

not needed i think?

Install method

Docker Compose

docker-compose file or Docker CLI command

not needed i think?

Object Detector

OpenVino

Network connection

Wireless

Camera make and model

not needed i think?

Screenshots of the Frigate UI's System metrics pages

No response

Any other information that may be helpful

No response

[inkeep-ai-assistant[bot]]

Based on the error you're seeing, Frigate is attempting to download the Jina semantic search model from Hugging Face on every restart when it can't find the cached files locally.

The model files should be cached in the model_cache directory within your config folder(1). Frigate will only attempt to download a model if it doesn't exist in the cache(1).

To resolve this issue:

1. Verify the model cache exists: Check that your /config/model_cache directory contains the downloaded Jina model files from your initial setup when you had internet access.

2. Ensure proper volume mounting: Make sure your Docker Compose configuration properly mounts the config directory so the cache persists across restarts. The cache should be at /config/model_cache inside the container.

3. Disable semantic search if not needed: If you don't need semantic search functionality, you can disable it in your config(2):

semantic_search:
  enabled: False

(2)

4. Backup the model cache: Since you can use Frigate+ models offline once downloaded(1), consider backing up your model_cache directory to prevent this issue in the future.

The system is designed to work completely offline once models are cached(1), so if the downloads keep happening, it indicates the cache isn't being preserved between restarts or the volume mount isn't configured correctly.

📚 Sources:

1. FAQ
2. Full Reference Config

---

Have another question? Just tag @inkeep.

[hawkeye217]

If you've already downloaded the necessary models for Semantic Search and your server is no longer internet connected, there are a few environment variables you can use to stop Frigate from reaching out altogether:

HF_HUB_OFFLINE=1
TRANSFORMERS_OFFLINE=1

Setting these without having the correct files downloaded already will cause a failure, though.

[Spirch]

pasted below a screenshot that show what is there, based, on that do you think it is safe to set these variables?

[hawkeye217]

Looks normal to me.

[Spirch]

this is what i have;