Conversation
- New model catalog (sol/astra/luna/terra/codex aliases, gpt-6 and
gpt-5.6 slugs, gpt-5.5); default model is gpt-6-astra.
- --effort {low,medium,high,xhigh,max,ultra} on sc_codex_task.py and
ai_cli run, plus optional reasoning_effort payload/schema field,
validated per model and passed as -c model_reasoning_effort="...".
Omitted when unset so ~/.codex/config.toml applies.
- CLI flags override the payload; effort survives background runs and
is logged as a top-level field.
- ChatGPT-account fallback now retries once on gpt-5.5 with effort
clamped to xhigh.
BREAKING CHANGE: legacy aliases gpt-5.2-codex, codex-max/max,
codex-mini/mini, gpt-5, gpt-5.2 and gtp-5 are removed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Bump version via set-package-version.py, add CHANGELOG entry, and register the sc-codex agent in .claude/agents/registry.yaml. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Apply --model/--effort to the raw payload before validation in sc_codex_task.py and ai_cli run, persisting the resolved model so background payloads and logs reflect what ran. - Add cmd_run tests (effort override, unsupported effort, override of invalid payload values) and schema nullability parity checks. - Allow null for optional fields in task_tool.schema.json. - Docs: effort-vs-model disambiguation, "alongside --json" wording, minimal effort note, --runner codex in ai_cli examples, argv-accurate -c quoting; CHANGELOG note replaces range heading. - Rename fixture test_parallel_minis.yaml -> test_parallel_luna.yaml. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Lets callers pick the Codex model and reasoning effort through sc-codex. For example, "use sc-codex with sol low" runs
codex exec --yolo --model gpt-6-sol -c model_reasoning_effort="low" <prompt>.Package: sc-codex · Type: feat (breaking) · Version: 0.13.0 → 0.14.0
Models
ai_cli/task_tool.py. The Pydantic model, the argparse help and the JSON schema all come from it, and a test checks that they match.sol→gpt-6-sol,astra→gpt-6-astra,luna→gpt-6-luna,terra→gpt-5.6-terra.codexor no model means the default.gpt-6-{sol,astra,luna},gpt-5.6-{sol,terra,luna},gpt-5.5.gpt-6-astra.mini,max,codex-mini,codex-max,gpt-5,gpt-5.2,gpt-5.2-codex,gtp-5). An unknown model now fails with an error that lists the valid names.gpt-5.5, with effort lowered toxhigh, and the retry is logged.Reasoning effort
--effortoption onsc_codex_task.pyandai_cli run, plus areasoning_effortfield in the JSON payload.low|medium|high|xhigh|max|ultra, checked per model before launch. The luna models stop atmaxandgpt-5.5stops atxhigh.minimalis rejected because no current model supports it.-csetting is passed, so~/.codex/config.tomlapplies.--modeland--effortoverride the payload, including a payload with an invalid value. The overrides are saved in the payload, so background runs and logs show what actually ran.Docs / housekeeping
SKILL.md, the package README and the ai_cli README are updated. They include how to turn phrases like "sol low" into flags, and a note thatmaxandultraare efforts, not models.sc_codex_task.py, off forai_cli run.sc-codexagent is added to.claude/agents/registry.yaml.scripts/set-package-version.py, which also updated the registries.minibecomesluna, andtest_parallel_minis.yamlbecomestest_parallel_luna.yaml.Review
gpt-6-astra(high effort): one should-fix, the override order.2534477.Testing
pytest tests/: 1444 passed, 2 failed. Both also fail on a cleanorigin/developin this environment (gonot installed; sc-composerender_templatecan't be imported).tests/test_ai_cli_task_runner.py: 61 tests covering aliases and the default, unknown-model errors,-cpresent or absent, unsupported effort per model, command-line overrides of the payload (including invalid payloads), the background payload carrying effort, the fallback, and schema/code parity.scripts/validate-all.py: 9/9 pass.scripts/audit-versions.py --verbose: 0 failures.--model luna --effort lowreturned output; the log showsgpt-6-luna/low.Out of scope (follow-ups)
{success, data, error}objectshell=Truein hook execution (task_runner.py)scripts/sc_shared.py, which appears unusedcodex-agentskill naming conventioncodex→gpt-5.6-terra)Notes for reviewer
sol/lunanow mean gpt-6. The uncommitted edit in~/.claude/scriptsused gpt-5.6. Usegpt-5.6-sol/gpt-5.6-lunaexplicitly for the older ones.~/.claude/scriptshas an uncommitted 0.12.0 hand-edit. Reinstall sc-codex after merging to replace it.🤖 Generated with Claude Code