gmakstutis/cognee
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Update LiteLLMEmbeddingEngine.py (#1205)

Browse source 
This will allow to deal with the issue when the user is using custom
embedding and LLM and passes the hosted_vllm option as part of the
LiteLLM documentation

<!-- .github/pull_request_template.md -->

## Description
<!-- This allows the user to use hosted_vllm with respect to LiteLLM
usage - and only gets applicable for custom embedding models -
specifically Hugging Face models -->

## DCO Affirmation
I affirm that all code in every commit of this pull request conforms to
the terms of the Topoteretes Developer Certificate of Origin.
This commit is contained in:
sssaha1989 2025-08-06 10:53:33 -04:00 • committed by GitHub
parent b54e843951
commit a9e74dac42
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 1 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
cognee/infrastructure/databases/vector/embeddings/LiteLLMEmbeddingEngine.py
View file

@ -186,7 +186,7 @@ class LiteLLMEmbeddingEngine(EmbeddingEngine):
tokenizer = MistralTokenizer(model=model, max_tokens=self.max_tokens)
else:
try:
tokenizer = HuggingFaceTokenizer(model=self.model, max_tokens=self.max_tokens)
tokenizer = HuggingFaceTokenizer(model=self.model.replace('hosted_vllm/',""), max_tokens=self.max_tokens)
except Exception as e:
logger.warning(f"Could not get tokenizer from HuggingFace due to: {e}")
logger.info("Switching to TikToken default tokenizer.")