Add documentation for model selection and custom providers

[Copilot]

Description

Users needed clarity on whether they can use custom local models (e.g., LMStudio instances) or are restricted to predefined models. This adds comprehensive documentation explaining model provider configuration.

Changes

• New documentation: docs/models.md (384 lines) covering:

  • Built-in providers: OpenAI, LMStudio (port 1234, default model openai/gpt-oss-20b), Ollama (port 11434, default model gpt-oss:20b)
  • Custom provider configuration via ~/.codex/config.toml with all available options (auth, wire protocols, HTTP headers, retry settings)
  • Provider examples: Azure OpenAI, OpenRouter, Together.ai, Groq
  • Configuration profiles for quick provider switching
  • Environment variable overrides (CODEX_OSS_BASE_URL, CODEX_OSS_PORT, OPENAI_BASE_URL)
  • Troubleshooting common issues

• Updated links:

  • README.md: Added model selection documentation to docs section
  • docs/config.md: Added reference to model selection guide

Example Configuration

[model_providers.lmstudio]
name = "LMStudio"
base_url = "http://localhost:1234/v1"
wire_api = "responses"
requires_openai_auth = false

[profiles.local]
model_provider = "lmstudio"
model = "llama-3.3-70b-instruct"

codex --oss  # Use LMStudio
codex --oss -m deepseek/r1-distill-qwen-32b  # Specify model
codex -p local  # Use profile

Configuration applies to both CLI and IDE usage.

Original prompt

In this article, the details are present of how to select specific models when using the extension. I would like to review this to see if I can use my own choice of local models (I have an LMStudio instance with a wide range of models available) or if I am restricted only to the provided ones. https://developers.openai.com/codex/ide/features#switch-between-models

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.