feat(api): update OpenAI Codex models for GPT-5.6 (re-land #147)

[BunsDev]

Summary

• Re-lands fork PR [codex] update OpenAI Codex models #147 (update OpenAI Codex models for GPT-5.6) from an internal branch, brought up to date with main.

Context

• Related issue/user request: supersedes [codex] update OpenAI Codex models #147 by @romgenie; branch pr-147-openai-models re-lands that work with local fixes.
• Scope: Codex model catalog/defaults (GPT-5.6 family), provider picker and onboarding labels, TUI test stabilization.
• Non-goals: no new providers, no auth changes.

Changes

• feat(api): update OpenAI Codex models — GPT-5.6 model family in the Codex catalog.
• fix(api) / fix(tui) — clippy formatting lint, provider picker credential test stabilization, test-home env race fix.
• docs: update codex default model to gpt-5.6-sol.
• Merged main back in; conflicts with 4751ee8 ("Refresh Codex defaults for GPT-5.6") resolved in favor of main's newer refresh (372K context windows for the 5.6 family per the upstream models.json catalog, simplified efficient-tier selection, updated labels). The branch's unique test fixes are preserved.

Validation

• git diff --check — clean
• cargo fmt --all — no changes
• cargo check --workspace — Finished, no warnings
• cargo clippy --workspace --all-targets -- -D warnings — Finished, no lints
• cargo test --workspace — 1852 passed, 0 failed
• Targeted/manual checks: conflict-resolution hunks reviewed line by line against main's 4751ee8
• Not run: release build (deferred to the v0.6.1 release step)

PR Readiness

• Diff is limited to the intended files.
• Generated files were not edited by hand.
• User-facing docs/help were updated or are not needed.
• No secrets, credentials, local paths, or unrelated logs are included.
• Remaining risks or follow-up work are listed above.

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
docs	Ready	Preview	Jul 11, 2026 3:11pm

[BunsDev]

Reviewed the post-merge delta (4 files, +32/−10) line by line.

Correctness

• model_registry.rs: GPT-5.6 family added at the head of the openai flagship patterns — ordering gives sol > terra > luna > family > 5.5 preference, matching the anthropic list's structure. Test assertions now pin the codex best/small fallbacks to DEFAULT_CODEX_MODEL / gpt-5.6-luna instead of only asserting membership — stronger regression net.
• api/src/lib.rs: two needless_borrow fixes in format! args; behavior-identical.
• tui/src/app.rs: EnvGuard correctly snapshots CLAUDE_BIN_ENV at construction, clears it for the test body, restores on drop — same pattern as the other guarded vars. The rewritten provider-picker test drops the imported-credentials fixture in favor of a local claude binary override, consistent with the no-imported-OAuth-tokens policy (fb989ab). Windows claude.cmd case handled.
• docs/providers.md: description refresh only.

Conventions — commit style, no unwraps on fallible paths, docs updated. Conflict resolution against 4751ee8 kept main's newer sourced context windows (372K for the 5.6 family); nothing from main was regressed.

Verification — local: cargo fmt (no changes), cargo check --workspace, cargo clippy --workspace --all-targets -- -D warnings, cargo test --workspace (1852 passed, 0 failed).

Nitpick, non-blocking: the bare "gpt-5.6" pattern is redundant given the three concrete ids above it, but it is a harmless family fallback.

Ready to merge once CI is green.

[Copilot]

Pull request overview

Re-lands and updates the Codex/OpenAI model catalog and defaults to the GPT-5.6 family, alongside small test-environment and TUI provider-picker test stabilization to keep workspace tests deterministic.

Changes:

• Update model-registry “flagship” heuristics and tests to align Codex defaults with DEFAULT_CODEX_MODEL and the GPT-5.6 family.
• Stabilize TUI provider-picker tests by using deterministic Claude CLI binary discovery via COVEN_CODE_CLAUDE_BIN and ensuring the test env guard restores that env var.
• Minor clippy-driven formatting cleanup for bearer auth header formatting; docs wording tweak for the legacy Codex model.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

File	Description
src-rust/crates/tui/src/app.rs	Test env guard now snapshots/restores COVEN_CODE_CLAUDE_BIN; provider picker test uses a temp “claude” binary path for deterministic local detection.
src-rust/crates/api/src/model_registry.rs	Adds GPT-5.6 family patterns to OpenAI flagship selection; strengthens Codex fallback assertions in tests.
src-rust/crates/api/src/lib.rs	Removes redundant borrows in format! calls for Authorization header (clippy cleanup).
docs/providers.md	Adjusts the gpt-5.2-codex description to “Legacy … compatibility model”.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.