princeomar9009/sciagent
🔬 Enhance your experiments with SciAgent, an AI-powered tool for automatic tracking, intelligent analysis, and local storage—all without altering your code.
What's novel
🔬 Enhance your experiments with SciAgent, an AI-powered tool for automatic tracking, intelligent analysis, and local storage—all without altering your code.
Code Analysis
10 files read · 3 roundsA CLI tool that wraps ML training commands, auto-extracts metrics from stdout via regex, tracks config/code changes, diffs against historical runs, and uses LLM APIs to generate analysis reports and weekly summaries for researchers.
Strengths
Clean modular architecture with clear separation of concerns (guardian, analyzer, log_parser, diff_engine, code_tracker). The log parser supporting multiple framework output formats including Chinese, and the multi-provider LLM abstraction with auto-detection are genuinely useful for the target audience of Chinese ML researchers.
Weaknesses
Zero test coverage, the README is completely misleading (describes a GUI app with PDF export when it's a CLI tool), the LLM invoke() method is unimplemented (just 'pass'), and the 'curves.png' output is a 1x1 pixel placeholder rather than real visualization.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
14
Contributors
2
Files
31
Active weeks
3
Repository
Language
Python
Stars
1
Forks
0
License
MIT