Embedding mode dosn't work.

[kelei999999]

LocalAI version:
<LocalAI 4.0.0>

Environment, CPU architecture, OS, and Version:
<Windows Server 2025 WSL Ubuntu 22.04 Nvidia RTX4090>

Describe the bug
<我在使用Qwen3-Embedding-4B时，发现无法返回正确的结果>

To Reproduce

Expected behavior

Logs

Additional context
<参数配置如下：
name: Qwen3-Embedding-4B
backend: llama-cpp
embeddings: true
context_size: 40960
parameters:
model: Qwen3-Embedding-4B/Qwen3-Embedding-4B-q4_k_m.gguf

请问Embedding模型应该怎么配置？

[scetu]

Facing same issue, Qdrant Vector Store retrieval results in corrupted tools call to Qdrant by Embedding model.

Switching to different embeding models did not solved the issue.
Reverting to version v3.12.1 fixed the issue and document retrieval from Qdrant is working again.

Output from embedding model during tools call is

Let me query the knowledge base for information.

<tool_call>
<function=Qdrant_Vector_Store1

resulting in errors

Error: Input data must contain a "input" field with the search query at handleRetrieveAsToolExecuteOperation