[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-93258":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":16,"stars7d":17,"stars30d":18,"stars90d":16,"forks30d":16,"starsTrendScore":16,"compositeScore":19,"rankGlobal":10,"rankLanguage":10,"license":20,"archived":21,"fork":21,"defaultBranch":22,"hasWiki":23,"hasPages":21,"topics":24,"createdAt":10,"pushedAt":10,"updatedAt":25,"readmeContent":26,"aiSummary":27,"trendingCount":16,"starSnapshotCount":16,"syncStatus":15,"lastSyncTime":28,"discoverSource":29},93258,"Codex-Orchestration","Cjbuilds\u002FCodex-Orchestration","Cjbuilds","Bring any model to Codex, assign them any role, use them in \u002Fgoal or any workflow.","",null,"Python",279,18,108,2,0,77,171,73.84,"MIT License",false,"main",true,[],"2026-07-22 04:02:08","# Codex Orchestration\n\nBring models like Claude Fable 5 into Codex, give them roles, and make them work together.\n\n## What does it do?\n\nCodex Orchestration turns one Codex task into a multi-model workflow.\n\n- Bring Fable 5 or another compatible model into Codex.\n- Assign roles such as advisor, executor, researcher, writer, designer, or reviewer.\n- Choose the order in which those roles work.\n- Let Codex manage the handoffs and return one verified result.\n\nThe model selected for the task stays in charge. It plans, decides what feedback to use, delegates work, tests the result, and gives you the final answer.\n\n## How it works\n\nHere is one example:\n\n```text\n                 YOUR TASK OR GOAL\n                         |\n                         v\n              SOL — ROOT ORCHESTRATOR\n                    creates the plan\n                         |\n                         v\n              FABLE 5 — PLAN ADVISOR\n                 finds gaps and risks\n                         |\n                         v\n              SOL — ROOT ORCHESTRATOR\n             improves the plan and decides\n                         |\n              +----------+----------+\n              |                     |\n              v                     v\n       LUNA EXECUTOR 1        LUNA EXECUTOR 2\n          builds a part          builds a part\n              |                     |\n              +----------+----------+\n                         |\n                         v\n              SOL — ROOT ORCHESTRATOR\n                 tests and delivers\n```\n\nYou choose the models, roles, and order. Codex follows the workflow while keeping final decisions with the root model.\n\n## Why use it?\n\n- **Better plans:** Fable 5 can challenge the root model before implementation begins.\n- **More perspectives:** use different models for planning, research, design, writing, review, or execution.\n- **Faster implementation:** independent executors can work in parallel—up to 2x faster on suitable tasks.\n- **Less limit pressure:** move repeated implementation work away from the root model and potentially hit premium-model limits about 40% less often.\n\nResults depend on the models, task, context, retries, and how much work can run in parallel. The speed and limit figures are targets, not guarantees.\n\n## Install\n\n```bash\ncodex plugin marketplace add Cjbuilds\u002FCodex-Orchestration\ncodex plugin add codex-orchestration@codex-orchestration\n```\n\nStart a new Codex task after installation so the plugin loads.\n\nSetup requires Python 3.11 or newer.\n\n## Quick start with Fable 5\n\nSelect the model you want to lead the task, then run:\n\n```text\n\u002Fcodex-orchestration setup executor: GPT-5.6 Luna Extra High, advisor: Claude Fable 5 High\n```\n\nThis creates the default workflow:\n\n```text\nSelected Codex model -> Fable 5 review -> selected model decides -> Luna executes -> selected model verifies\n```\n\nFable 5 uses the official Claude Code CLI. You need a compatible first-party Claude login, but you do not need to add an Anthropic API key to Codex.\n\nAfter setup, start another new task and work normally.\n\n## Tell Codex your workflow\n\nPaste a workflow at the start of a task:\n\n```text\nUse this workflow:\n\n1. The selected model is the root orchestrator and creates the plan.\n2. Claude Fable 5 reviews the plan as advisor.\n3. The root accepts only useful feedback and improves the plan.\n4. Luna executors build independent parts in parallel.\n5. The root integrates, tests, and verifies the final result.\n```\n\nYou can change the sequence for any task:\n\n```text\nResearcher -> root synthesis -> designer -> writer -> reviewer -> root verification\n```\n\n## Bring another model into Codex\n\nAsk Codex Orchestration to create a role:\n\n```text\n\u002Fcodex-orchestration create these project roles:\n\n- researcher\n  model: \u003Cmodel-id>\n  provider: \u003Cconfigured-provider-id>\n  effort: high\n  job: gather evidence and cite sources\n\n- writer\n  model: \u003Cmodel-id>\n  effort: medium\n  job: turn approved research into a clear draft\n\n- designer\n  model: \u003Cmodel-id>\n  effort: high\n  job: propose and review the user experience\n```\n\nCodex previews the role files before creating them. Project roles live in `.codex\u002Fagents\u002F`. Personal roles live in `~\u002F.codex\u002Fagents\u002F` and can be used across projects.\n\nFable 5 is currently bundled as a plan advisor. Other models must already be available through Codex or an authenticated, compatible provider.\n\n## Use it with Codex Goals\n\nSet a Goal normally, then add your workflow:\n\n```text\n\u002Fgoal Ship the authentication redesign with tests and migration notes.\n\nUse my Fable advisor and Luna executor workflow until the Goal is complete.\n```\n\nCodex still owns the Goal. The plugin controls the model workflow inside it.\n\n## Useful commands\n\n```text\n\u002Fcodex-orchestration status\n\u002Fcodex-orchestration status --require-effective\n\u002Fcodex-orchestration setup executor: GPT-5.6 Luna Extra High, advisor: Claude Fable 5 High\n\u002Fcodex-orchestration disable\n```\n\n`disable` restores the Codex routing values that existed before setup. It does not delete user-owned custom roles.\n\n## Important limits\n\n- Codex remains the root orchestrator.\n- Fable 5 is a root-facing plan advisor, not a second orchestrator.\n- Other providers must already be configured and authenticated.\n- The plugin never creates credentials or bypasses permissions and approvals.\n- Codex decides when delegation or parallel work is useful.\n- If you say `no subagents`, Codex must not delegate.\n\nTechnical details are in [providers and models](plugins\u002Fcodex-orchestration\u002Fskills\u002Fcodex-orchestration\u002Freferences\u002Fproviders-and-models.md).\n\n## Update\n\n```bash\ncodex plugin marketplace upgrade codex-orchestration\ncodex plugin add codex-orchestration@codex-orchestration\n```\n\nStart a new task after updating.\n\n## Uninstall\n\nFirst run:\n\n```text\n\u002Fcodex-orchestration disable\n```\n\nThen remove the plugin:\n\n```bash\ncodex plugin remove codex-orchestration@codex-orchestration\ncodex plugin marketplace remove codex-orchestration\n```\n\nReview and remove any user-owned custom roles separately.\n\n## Development\n\n```bash\npython3 -m pip install -r requirements-dev.txt\npython3 -m compileall -q plugins tests scripts\npython3 -m ruff check plugins tests scripts\npython3 -m unittest discover -s tests -v\npython3 tests\u002Fplugin_lifecycle_smoke.py\npython3 scripts\u002Frelease_check.py\n```\n\nSee the [production-readiness audit](docs\u002Fproduction-readiness-audit.md), [security policy](SECURITY.md), and [release process](RELEASE.md).\n\n## License\n\nMIT\n","Codex Orchestration 是一个用于在 Codex 平台中编排多模型协同工作的插件。它支持将不同大语言模型（如 Claude Fable 5、GPT-5.6 Luna 等）按角色（如顾问、执行者、评审员）接入同一任务流程，由根模型统一规划、分派、验证，实现串行或并行的多阶段推理协作。技术上基于 Python 实现，依赖 Codex 插件机制，无需额外 API 密钥（部分模型复用官方 CLI 认证）。适用于需多视角校验、复杂任务分解、降低主模型调用压力的场景，例如技术方案设计、内容协同生成与自动化报告编制。","2026-07-14 02:30:07","CREATED_QUERY"]