Files
Read, write, edit, delete, and list files directly in your working folder.
Product
machine learning pipeline for your own projects
01
Model Ledger helps you build the scripts of a machine learning pipeline by reading, editing files and running your commands in your own folder and shows the real output.
Shell
Run terminal commands and see the real output and exit code.
Approval
Pause before changes, plan first, or run actions automatically.
Features
- Filesread_file, write_file, edit_file, delete_file, make_dir, move_file, copy_file, list_dir, grep, glob work in the chat folder.
- Shellruns terminal commands like builds and tests, judging success by the real exit code.
- Webweb_search finds information and http_get reads official documentation pages.
- Screenshots and imagesgen_image saves real image files and ocr reads text from screenshots.
- Orchestrationtask splits work into parallel sub-agents and run_on_model sends hard steps to bigger models.
- Approval modesAuto-run acts at once, Ask first pauses before changes, Plan only writes a plan.
- Cronruns prompts on a schedule.
- Modelsuse a cloud provider, a local model, or the built-in model with a Test button.
How it works
The agent inspects your project, runs commands, edits scripts, and shows every change as a diff; in Ask first mode the diff waits for your OK.
Create a baseline script that reads a CSV and prints the most common label.
- list_dir
.1 match - write_file
baseline.pysaved - edit_file
baseline.py3 changes - Shell
python baseline.pyexit 0 - read_file
baseline.py8 lines
waiting for your OK before the next stepRejectAccept
Walkthroughs
Beginner steps
Start your machine learning pipeline by picking a small idea like a spam filter, and the agent uses list_dir to see your folder structure. It is real. The agent uses glob to find relevant scripts in your project folder. It then uses read_file to inspect your data loading code carefully. Next, it uses write_file to create a simple baseline script for you. People who search for 'practical resume worthy ml projects' often begin with this kind of clear structure. It uses edit_file to change the script if needed, showing a diff in Ask first mode. Finally, it runs your training command in the Shell tool for real output.
Advanced review
The agent uses grep to find where a setting is used in your codebase and web_search to find official documentation for your library. Simple. It uses task to split a review into parts for parallel processing. http_get reads specific pages while gen_image saves a plot. ocr reads text from a screenshot and run_on_model handles hard steps. The version-control skill manages branches and commits for your machine learning projects for beginners. Cron schedules tasks so you review the plan in Plan only mode. Real world ml project ideas emerge when you switch to Auto-run to let it finish.
What Model Ledger does not do
Use it only on your own files, data and systems or ones you have written permission for, and you are responsible for that. The app does not guarantee that code or analysis is correct, so review the diff and run the tests. It is useful for finding good topics for ml projects on resume.
