[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-2635":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":9,"language":10,"languages":9,"totalLinesOfCode":9,"stars":11,"forks":12,"watchers":13,"openIssues":14,"contributorsCount":14,"subscribersCount":14,"size":14,"stars1d":15,"stars7d":16,"stars30d":17,"stars90d":14,"forks30d":14,"starsTrendScore":18,"compositeScore":19,"rankGlobal":9,"rankLanguage":9,"license":20,"archived":21,"fork":21,"defaultBranch":22,"hasWiki":23,"hasPages":21,"topics":24,"createdAt":9,"pushedAt":9,"updatedAt":25,"readmeContent":26,"aiSummary":27,"trendingCount":14,"starSnapshotCount":14,"syncStatus":28,"lastSyncTime":29,"discoverSource":30},2635,"claude-coworker-model","imkunal007219\u002Fclaude-coworker-model","imkunal007219","Give Claude Code a cheap coworker. CLI tools that delegate bulk I\u002FO to cheap LLMs (Kimi, DeepSeek, Ollama). Save 60-70% of your token budget.",null,"Python",150,38,3,0,7,19,45,21,4.77,"MIT License",false,"master",true,[],"2026-06-12 02:00:42","# Claude Coworker Model\n\n[![License: MIT](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-MIT-yellow.svg)](https:\u002F\u002Fopensource.org\u002Flicenses\u002FMIT)\n\nOffload bulk I\u002FO from Claude Code to cheap LLMs. Save thousands of tokens on file reading, boilerplate generation, and doc updates. Worker calls cost ~$0.02; primary model focuses on architecture.\n\n## Quick Start\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fimkunal007219\u002Fclaude-coworker-model.git\ncd claude-coworker-model\n.\u002Fsetup.sh\n\nexport WORKER_API_KEY=\"your-key\"\nexport WORKER_BASE_URL=\"https:\u002F\u002Fapi.moonshot.ai\u002Fv1\"\nexport WORKER_MODEL=\"kimi-k2.5\"\n\nask-kimi --paths src\u002F*.py --question \"Find all SQL injection risks\"\n```\n\n## How It Works\n\nThe expensive model (Claude) handles reasoning and architecture. The cheap worker model handles token-heavy I\u002FO:\n\n1. **Read**: Worker ingests large codebases, returns structured summaries with file paths and line numbers\n2. **Generate**: Worker produces boilerplate using existing files as style references  \n3. **Extract**: Worker parses session transcripts for documentation\n\nPattern: Claude decides *what* to do; the worker does the *reading\u002Fwriting*.\n\n## Configuration\n\nThree environment variables configure any OpenAI-compatible provider:\n\n| Variable | Purpose | Example |\n|----------|---------|---------|\n| `WORKER_API_KEY` | API authentication | `sk-abc123` |\n| `WORKER_BASE_URL` | Provider endpoint | `https:\u002F\u002Fapi.moonshot.ai\u002Fv1` |\n| `WORKER_MODEL` | Model identifier | `kimi-k2.5` |\n\n## Provider Examples\n\n**Kimi (Moonshot AI)**\n```bash\nexport WORKER_API_KEY=\"$MOONSHOT_API_KEY\"\nexport WORKER_BASE_URL=\"https:\u002F\u002Fapi.moonshot.ai\u002Fv1\"\nexport WORKER_MODEL=\"kimi-k2.5\"\n```\n\n**DeepSeek**\n```bash\nexport WORKER_API_KEY=\"$DEEPSEEK_API_KEY\"\nexport WORKER_BASE_URL=\"https:\u002F\u002Fapi.deepseek.com\u002Fv1\"\nexport WORKER_MODEL=\"deepseek-chat\"\n```\n\n**Ollama (local)**\n```bash\nexport WORKER_API_KEY=\"ollama\"\nexport WORKER_BASE_URL=\"http:\u002F\u002Flocalhost:11434\u002Fv1\"\nexport WORKER_MODEL=\"qwen2.5-coder:14b\"\n```\n\n## Tools\n\n### ask-kimi\nDelegate bulk reading to the worker model. Returns structured bullets, not prose.\n\n```bash\n# Analyze multiple files for security issues\nask-kimi \\\n  --paths auth.py database.py utils.py \\\n  --question \"Identify all unvalidated inputs\" \\\n  --max-tokens 8192\n\n# Generate API documentation from source\nask-kimi \\\n  --paths src\u002F**\u002F*.ts \\\n  --question \"List all exported functions with their arguments\"\n```\n\nFlags:\n- `--paths`: Files to ingest (supports globs)\n- `--question`: Specific extraction query\n- `--max-tokens`: Total budget including reasoning tokens\n- `--model`: Override `WORKER_MODEL`\n\n### kimi-write\nGenerate code or documentation using an existing file as a style reference.\n\n```bash\n# Generate tests matching existing style\nkimi-write \\\n  --spec \"Write pytest tests for auth.py covering OAuth2 flow\" \\\n  --context tests\u002Ftest_main.py \\\n  --target tests\u002Ftest_auth.py\n\n# Create API docs matching current format\nkimi-write \\\n  --spec \"Document the new \u002Fv2\u002Fusers endpoint\" \\\n  --context docs\u002Fendpoints.md \\\n  --target docs\u002Fendpoints_v2.md\n```\n\nFlags:\n- `--spec`: What to write (generation instructions)\n- `--context`: Reference file to mimic (style, imports, structure)\n- `--target`: Output file path\n- `--max-tokens`: Token budget for reasoning + output (default 16384)\n\n### extract-chat\nConvert Claude Code JSONL session logs to human-readable text.\n\n```bash\n# Extract last session to stdout\nextract-chat ~\u002F.claude\u002Fprojects\u002Fmy-project\u002Fsession.jsonl\n\n# Write to file\nextract-chat ~\u002F.claude\u002Fprojects\u002Fmy-project\u002Fsession.jsonl -o \u002Ftmp\u002Fchat.txt\n\n# Pipe to ask-kimi for doc updates\nextract-chat session.jsonl -o \u002Ftmp\u002Fchat.txt && \\\n  ask-kimi --paths \u002Ftmp\u002Fchat.txt docs\u002FREADME.md --question \"What doc updates are needed?\"\n```\n\n## CLAUDE.md Setup\n\nCopy `CLAUDE.md.template` to your project root as `CLAUDE.md`. This provides routing rules that tell Claude when to delegate:\n\n```markdown\n## Worker Delegation Rules\n\nWhen asked to analyze, summarize, or search across multiple files:\nDELEGATE to ask-kimi with relevant file paths.\n\nWhen asked to generate boilerplate, tests, or documentation:\nDELEGATE to kimi-write with appropriate reference files.\n\nWhen asked to review session history:\nDELEGATE to extract-chat.\n\nDO NOT delegate:\n- Architecture decisions\n- Debugging complex logic\n- Refactoring plans\n```\n\nAdd `CLAUDE.md` to your repository so Claude Code loads it automatically on startup.\n\n## Results\n\n| Metric | Before | After |\n|--------|--------|-------|\n| Claude Pro weekly limit | Hit by Wednesday | Never hit |\n| Token usage per session | 80%+ on file reading | 20% (summaries only) |\n| 3-week worker API cost | — | $0.38 total |\n| Context window usage | 80% reading files | 20% reading summaries |\n\nBased on the pattern described in [this implementation (medium link)](https:\u002F\u002Fmedium.com\u002F@kunalbhardwaj598\u002Fi-was-burning-through-claude-codes-weekly-limit-in-3-days-here-s-how-i-fixed-it-0344c555abda) [Reddit link](https:\u002F\u002Fwww.reddit.com\u002Fr\u002FClaudeAI\u002Fcomments\u002F1t1o43w\u002Fi_gave_claude_code_a_002call_coworker_and_stopped\u002F?utm_source=share&utm_medium=web3x&utm_name=web3xcss&utm_term=1&utm_content=share_button) (567K views Reddit, 7.2K Medium).\n\n## Author\n\n**Kunal Bhardwaj** — Systems engineer working on autonomous drones and AI-powered developer tools. Building at the intersection of embedded systems and LLM workflows.\n\n- Blog: [medium.com\u002F@kunalbhardwaj](https:\u002F\u002Fmedium.com\u002F@kunalbhardwaj598\u002Fi-was-burning-through-claude-codes-weekly-limit-in-3-days-here-s-how-i-fixed-it-0344c555abda)\n- LinkedIn: [linkedin.com\u002Fin\u002Fkunalbhardwaj](https:\u002F\u002Fwww.linkedin.com\u002Fin\u002Fkunal-bhardwaj-61433818b)\n\n## Contributing\n\nPRs welcome. Focus areas: additional provider templates, token usage optimization, and extracting structured data from more session formats.\n\nMIT License. See [LICENSE](LICENSE).\n","Claude Coworker Model 是一个通过将大量I\u002FO任务委派给成本较低的语言模型（如Kimi、DeepSeek、Ollama），从而节省60-70%令牌预算的CLI工具。该项目的核心功能包括使用廉价模型处理文件读取、代码生成及文档更新等高消耗任务，而主要模型Claude则专注于架构设计与决策逻辑。此项目特别适合需要频繁处理大规模代码库或文档更新但又希望控制成本的开发场景。项目采用Python编写，并且遵循MIT许可证开放源代码。",2,"2026-06-11 02:50:34","CREATED_QUERY"]