Commit 3ebc5da

Francisco

committed

fix: refactor multimodal handling in vLLMRawStream, improve content merging logic, and enhance conversation truncation

1 parent 2728e30 commit 3ebc5daCopy full SHA for 3ebc5da

6 files changed

scripts
- hugging_face.py
src/api/entities_api
- clients
  - vllm_raw_stream.py
- orchestration/mixins
  - service_registry_mixin.py
- services
  - file_service.py
- utils
  - conversation_truncator.py
tests/integration
- vllm_orc_vision_sdk_managed_turns.py

`‎scripts/hugging_face.py‎`

-Original file line number
+Diff line change
@@ @@ -0,0 +1,4 @@ @@
 +# From inside WSL or your container
 +from huggingface_hub import snapshot_download
++
 +snapshot_download(repo_id='Qwen/Qwen2.5-VL-3B-Instruct')

Comments

(0)