This server does not support reranking

[lijianxin520]

LocalAI version:
localai/localai:master-aio-cpu (Docker image, pulled on: 2026-02-26)

Environment, CPU architecture, OS, and Version:
CPU info:
model name : AMD Ryzen 5 7530U with Radeon Graphics
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good nopl tsc_reliable nonstop_tsc cpuid extd_apicid tsc_known_freq pni pclmulqdq ssse3 fma cx16 sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm cmp_legacy svm cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw topoext perfctr_core ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 erms invpcid rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 xsaves clzero xsaveerptr arat npt nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold v_vmsave_vmload umip vaes vpclmulqdq rdpid fsrm
CPU: AVX found OK
CPU: AVX2 found OK
CPU: no AVX512 found

• Hardware: x86_64 CPU (8 cores / 16GB RAM), no GPU
• Operating System: Windows 11
• Docker version 29.2.1, build a5c7197

Describe the bug
When attempting to enable the reranking feature, calling the /v1/rerank API returns an RPC error: rpc error: code = Unimplemented desc = This server does not support reranking. Start it with --reranking and without --embedding.
However:

1. The latest localai/localai:latest image is used, which has removed the --reranking startup flag — adding this flag will result in an unknown flag --reranking error;
2. The reranking model has been configured with known_usecases: [rerank] as per the official documentation, and the model file & configuration file have correct paths/naming;
3. Automatic model download was disabled at startup, but the error still occurs, and the reranking API remains unavailable.

To Reproduce
Steps to reproduce the behavior:

1. Prepare the reranking model file: Place bge-reranker-v2-m3-Q4_K_M.gguf in the ./models directory;
2. Create the model configuration file ./models/bge-reranker-v2-m3-Q4_K_M.gguf.yaml with the following content:

backend: llama-cpp
description: Imported from file:///models/bge-reranker-v2-m3-Q4_K_M.gguf
function:
  grammar:
    disable: true
known_usecases:
  - rerank
name: bge-reranker-v2-m3-Q4_K_M.gguf
options:
  - use_jinja:true
parameters:
  model: bge-reranker-v2-m3-Q4_K_M.gguf
template:
  use_tokenizer_template: true
embeddings: false

[lijianxin520]