Using remote Llama-cpp, endpoint is "completion" not "completions". The error is a 404. I tested with llama-serve and llama-cpp-python. Both work with http://localhost:23451/v1/completions and CURL.

Error: `HTTP 404 Not Found from http://127.0.0.1:23452/v1/completion {"detail":"Not Found"}`

Environment:
- OS: OS X
- Continue version: 1.0.5 (pre-release)
- IDE version: VSCode 1.98.2
- config:
  
models:
  - name: CodeLlama7bChat
    provider: llama.cpp
    model: codellama-7b-instruct-Q6_K
    apiBase: http://127.0.0.1:23452/v1
    roles:
      - chat
      - edit
      - apply
tabAutoCompleteModel:
    title: DeekSeekCoder
    provider: llama.cpp
    model: deepseek-coder:1.3b-base-q6_K
    apiBase: http://127.0.0.1:23451/v1
context:
  - provider: code
  - provider: docs
  - provider: diff
  - provider: terminal

  
I also tested with:


    env:
      useLegacyCompletionsEndpoint: false

in the config (based on an open PR).