{"slug": "i-got-30-better-ai-coding-results", "title": "I got 30% better AI coding results", "summary": "CodeSlimmer v2.2, a new open-source tool by developer akasula09, converts local codebases or public GitHub repositories into a single LLM-ready text file with token estimation and directory mapping, claiming to improve AI coding results by 30%. The tool supports AI models including Claude 3.5, Gemini 3.1 Pro, GPT-4o, and DeepSeek R1, and is installable via pip. It features remote GitHub URL support, cross-platform clipboard integration, overwrite safety, ASCII directory trees, XML output, token estimation, and smart file size formatting.", "body_md": "Safely convert local codebases or public GitHub repos into clean, LLM-ready context with token estimation and directory mapping.\n\nCodeSlimmer packs any code repository—whether stored on your local machine or hosted on GitHub—into a single text file structured specifically for AI models like **Claude 3.5**, **Gemini 3.1 Pro/3.6 Flash**, **GPT-4o**, and **DeepSeek R1**.\n\n- 🌐\n**Remote GitHub URL Support:** Pass any public GitHub URL (`[https://github.com/user/repo](https://github.com/user/repo)`\n\n) to fetch, extract, and slim repositories directly—zero`git`\n\nrequired. - 📋\n**Cross-Platform Clipboard Integration:** Instantly copies output directly to your system clipboard (Windows, macOS, or Linux). - 🛡️\n**Overwrite Safety:** Inspects signature headers to ensure local files are never overwritten accidentally. - 🌳\n**ASCII Directory Tree:** Builds a visual project structure map at the very top of your context output. - 🏷️\n**LLM-Friendly XML Output:** Encloses code in clear`<file path=\"...\">`\n\ntags for max prompt accuracy. - 📊\n**Token Estimator:** Estimates context size (~4 chars/token) and checks capacity against major models. - 📏\n**Smart File Sizes:** Calculates and formats physical sizes (`B`\n\n,`KB`\n\n,`MB`\n\n,`GB`\n\n).\n\nInstall CodeSlimmer using `pip`\n\n:\n\n```\npip install code-slimmer\n```\n\nClone the repository and run the script directly:\n\n```\ngit clone [https://github.com/akasula09/codeslimmer.git](https://github.com/akasula09/codeslimmer.git)\ncd codeslimmer\n```\n\nIf installed via `pip`\n\n:\n\n```\ncodeslimmer\n```\n\nIf running directly from source:\n\n```\npython codeslimmer.py\n```\n\nCodeSlimmer supports two distinct input modes:\n\nEnter a relative or absolute local folder path. CodeSlimmer creates `ai_context.txt`\n\ninside that folder and offers an optional clipboard copy. An example is below.\n\n```\n--- CODE SLIMMER v2.2 ---\nEnter a local folder path OR a GitHub URL: my_project_folder\n\n📄 Loaded rules from .gitignore\n\n--- SUCCESS! ---\nPacked 12 code files.\nFinal Context Size: 18.45 KB\n💾 Added 'ai_context.txt' directly into your folder!\n\n📊 Estimated Token Count: ~4,612 tokens\n--------------------------------------------------\nWhich AI provider are you targeting? (gemini / claude / chatgpt / deepseek): gemini\n--------------------------------------------------\n🎯 Evaluated against Gemini Models:\n\n  ✅ Gemini 3.1 Pro: Fits! (4,612 / 2,000,000 tokens — 0.2306% capacity)\n  ✅ Gemini 3.6 Flash: Fits! (4,612 / 1,000,000 tokens — 0.4612% capacity)\n  ✅ Gemini 3.5 Flash-Lite: Fits! (4,612 / 1,000,000 tokens — 0.4612% capacity)\n\nDo you want to copy the output text directly to your clipboard as well?\nType 'y' for Yes, or press Enter to skip: y\n📋 Copied full context directly to your clipboard! Ready to Ctrl+V into AI.\n```\n\nPaste any public GitHub URL. CodeSlimmer fetches and slims the repository in a temporary workspace, estimates tokens, and prompts you to copy the result straight to your clipboard without leaving messy files behind.\n\n```\n--- CODE SLIMMER v2.2 ---\nEnter a local folder path OR a GitHub URL: [https://github.com/akasula09/codeslimmer](https://github.com/akasula09/codeslimmer)\n\n🌐 Fetching repository 'akasula09/codeslimmer' from GitHub...\n\n--- SUCCESS! ---\nPacked 5 code files.\nFinal Context Size: 12.30 KB\n\n📊 Estimated Token Count: ~3,075 tokens\n--------------------------------------------------\nWhich AI provider are you targeting? (gemini / claude / chatgpt / deepseek): claude\n--------------------------------------------------\n🎯 Evaluated against Claude Models:\n\n  ✅ Claude 3.5 Sonnet / Haiku: Fits! (3,075 / 200,000 tokens — 1.5375% capacity)\n  ✅ Claude 3 Opus: Fits! (3,075 / 200,000 tokens — 1.5375% capacity)\n\nDo you want to copy the output text directly to your clipboard?\nType 'y' for Yes, or press Enter to skip: y\n📋 Copied full context directly to your clipboard! Ready to Ctrl+V into AI.\n```\n\nℹ️\n\nPrivate Repositories:Unauthenticated requests to private or non-existent repositories will output:\n\n`❌ Repository does not exist or is private.`\n\nCodeSlimmer structures the prompt context using this clean, LLM-optimized XML format:\n\n``` php\n<!-- GENERATED BY CODESLIMMER v2.2 - DO NOT EDIT MANUALLY -->\n\n<project_structure_tree>\nmy_project_folder/\n├── src/\n│   ├── index.js\n│   └── utils.js\n├── .gitignore\n└── package.json\n</project_structure_tree>\n\n<file path=\"src/index.js\">\nconsole.log(\"Hello World\");\n</file>\n\n<file path=\"src/utils.js\">\nexport function add(a, b) { return a + b; }\n</file>\n```\n\nDistributed under the **PolyForm Noncommercial License 1.0.0**. Free for personal, educational, research, and internal business use. See [LICENSE](/akasula09/CodeSlimmer/blob/main/LICENSE) for details.", "url": "https://wpnews.pro/news/i-got-30-better-ai-coding-results", "canonical_source": "https://github.com/akasula09/CodeSlimmer", "published_at": "2026-08-09 22:01:47+00:00", "updated_at": "2026-08-09 22:35:03.727764+00:00", "lang": "en", "topics": ["developer-tools", "ai-tools", "artificial-intelligence"], "entities": ["CodeSlimmer", "akasula09", "Claude 3.5", "Gemini 3.1 Pro", "GPT-4o", "DeepSeek R1", "GitHub"], "alternates": {"html": "https://wpnews.pro/news/i-got-30-better-ai-coding-results", "markdown": "https://wpnews.pro/news/i-got-30-better-ai-coding-results.md", "text": "https://wpnews.pro/news/i-got-30-better-ai-coding-results.txt", "jsonld": "https://wpnews.pro/news/i-got-30-better-ai-coding-results.jsonld"}}