The latest articles on DEV Community by ZeroRanker (@zeroranker). https://dev.to/zeroranker https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3874514%2F88204aba-2299-4743-88dd-52c8119dbfd2.jpg https://dev.to/zeroranker en ZeroRanker Sun, 12 Apr 2026 07:11:50 +0000 https://dev.to/zeroranker/how-i-saved-150-per-api-call-by-building-a-smart-repo-compressor-for-ai-assistants-55po https://dev.to/zeroranker/how-i-saved-150-per-api-call-by-building-a-smart-repo-compressor-for-ai-assistants-55po <h2> What It Does </h2> <p>ContextForge is a Python CLI tool that solves the "my repo is too big for AI context window" problem.</p> <p>The Python <code>requests</code> library = 154,061 tokens ($2.31 per GPT-4 call). 50% of that is test files, docs, changelogs.</p> <p>ContextForge compresses it to 50,406 tokens ($0.77) by:</p> <ul> <li>Analyzing Git history to find important files</li> <li>Auto-excluding tests/binaries/node_modules</li> <li>Extracting function signatures (not implementations)</li> <li>Exporting optimized packages for GPT-4, Claude XML, Gemini, Aider</li> </ul> <h2> Results </h2> <div class="table-wrapper-paragraph"><table> <thead> <tr> <th>Metric</th> <th>Before</th> <th>After</th> </tr> </thead> <tbody> <tr> <td>Tokens</td> <td>154,061</td> <td>50,406 (67% ↓)</td> </tr> <tr> <td>Cost/Call</td> <td>$2.31</td> <td>$0.77</td> </tr> <tr> <td>Files</td> <td>112</td> <td>54</td> </tr> </tbody> </table></div> <p>One command: <code>contextforge forge ./repo --budget 50000 --model gpt4</code></p> <h2> Tech Stack </h2> <p>Python 3.10+ (tiktoken, gitpython, rich, typer, matplotlib)<br> 48 tests passing | MIT License | Fully open source</p> <h2> Links </h2> <p>GitHub: <a href="https://github.com/zeroranker/contextforge" rel="noopener noreferrer">https://github.com/zeroranker/contextforge</a></p> <p>Built in ~2 hours as a side project. Would love feedback!</p> ai llm productivity python