Non GPT model from custom provider cannot edit file using `apply_patch`, also cannot use `multi_tool_use.parallel`

What version of Codex CLI is running?

0.118.0

What subscription do you have?

plus

Which model were you using?

Fireworks Kimi K2.5

What platform is your computer?

_No response_

What terminal emulator and version are you using (if applicable)?

_No response_

What issue are you seeing?

I tried using the Fireworks Kimi k2.5 model. Fireworks already supports the Responses API and parallel tool call.

[model_providers.fireworks-ai]
name = "Fireworks"
base_url = "https://api.fireworks.ai/inference/v1"
env_key = "FIREWORKS_API_KEY"
wire_api = "responses"

[profiles.kimi]
model = "accounts/fireworks/routers/kimi-k2p5-turbo"
model_provider = "fireworks-ai"
model_reasoning_effort = "high"
model_context_window = 256000
model_auto_compact_token_limit = 128000

However, when the model attempts to edit a file, it consistently resorts to shell-based methods such as sed or cat > file.name << EOF rather than using the apply tool. Moreover, when I asked the model which tools it had available and whether it has a tool capable of applying patches, it responded that it did not have such a tool.

There is also a second issue: the model lacks access to the parallel tool-calling mechanism, despite Fireworks' Responses API supporting this capability.

What steps can reproduce the bug?

1. codex --profile kimi

2.

> List tools you are available to call, do you have tool that can apply patch?

> • Here are the tools available to me:

  1. exec_command – Run shell commands in a PTY (pipelines, redirections, etc.)
  2. write_stdin – Write input to an ongoing exec_command session
  3. update_plan – Update progress plans (steps & status)
  4. request_user_input – (Plan mode only) Ask the user up to 3 questions
  5. view_image – Display a local image file
  6. spawn_agent – Create sub-agents for parallel tasks
  7. send_input – Send messages to running agents
  8. resume_agent – Re‑open a closed agent
  9. wait_agent – Block until specific agents finish
  10. close_agent – Shut down an agent

  No native apply_patch tool – I don’t have a built‑in editing primitive.

As shown, the model cannot access functions.apply_patch to edit files, nor does it have the multi_tool_use.parallel tool for parallel tool invocation.

What is the expected behavior?

Non OpenAI model should be able to use the apply patch tool to edit file.

Additional information

_No response_