No tooling is called when using local providers like Ollama

What version of Codex is running?

codex-cli 0.5.0

Which model were you using?

qwen2

What platform is your computer?

Linux 5.15.0-141-generic x86_64 x86_64

What steps can reproduce the bug?

The filtering stop models that support tool calling if they don't have codex in the name.

For example, running a session (which is annoying because of all of the other ENVs that must be set to by-pass other defects for local models):

OPENAI_API_KEY=sk-none OPENAI_API_BASE=http://localhost:11434/v1 OLLAMA_BASE_URL=http://localhost:11434/v1 codex --profile qwen2 --full-auto

Then request to make changes to a repo. The expected result is that changes are made, but instead a patch is created by not applied:

update all the words of "chunk" in the readme.md to "chonk", apply the patch automatically

<img width="942" height="708" alt="Image" src="https://github.com/user-attachments/assets/1d049222-7c61-473a-bea1-d9496c7644c8" />

An the result:

<img width="654" height="470" alt="Image" src="https://github.com/user-attachments/assets/00e1a8ee-a576-4ee5-a769-cf1b3d223872" />

What is the expected behavior?

The expected actions would be that there is consistent behavior regardless of the provider.

What do you see instead?

No modifications to the local files or execution of targeted executables.

Additional information

# config.toml
# actual configured execution
model_provider = "ollama"
model = "qwen2"
# approval_policy = "untrusted"
wire_api = "chat"

[model_providers.ollama]
name = "ollama"
base_url = "http://localhost:11434/v1"

[profiles.qwen2]
model_provider = "ollama"
model = "qwen2"

[profiles.codellama]
model_provider = "ollama"
model = "codellama"