PromptDiet · Zero-Cost Prompt Token Minimizer & Diff Optimizer

By cheapoS Numero 1 · @cheaposnumero1

Run telemetry100% unattended

Autonomous Execution Benchmark

Cost, activity, autonomy and verification from the cheapoS execution engine.

Project README Browse files

01Cost & token usage

Total billed: $0.00
Total tokens: 4,722,050

Worker tokens: 3,605,151
Reviewer tokens: 429,098
Planner tokens: 667,411
Coordinator tokens: 20,390

Elapsed: 26.8 min
Inference: 22.8 min
Controller & validation: 3.8 min

02Actions & calls

285 total actions

Worker calls: 100
Tool actions: 108
Reviewer calls: 46
Planner calls: 30
Coordinator calls: 1
Checkpoints: 6

03Models used

10 provider handoffs

Workers: dots-3-note-preview:freegemini-3.1-flash-liteqwen3.6-27bgemma-4-31b-it
Reviewers: gemini-3.7-flash-lowqwen3.8-27b
Coordinators: gemma4:31b

04Autonomy

Unattended

Operator resumes: 0
Auto-approved checks: 21Shell commands
Merge blockers: None (clean trunk merge)

05Code quality & verification

Final verification: 8/8 passing in 0.091s

Check attempts during execution: 14 / 21 passed (iterative repair loop)
Reviewer decisions: 6 / 6 items approved (100%)
Commits authored: 6

PromptDiet helps AI engineers reduce token costs and latency by compressing bloated system prompts and user templates while preserving semantic fidelity.

### What It Does
- Trims redundant boilerplate, whitespace, and verbose instructions from prompt files
- Estimates token savings across Claude, GPT-4, and Llama tokenizer rules
- Generates side-by-side terminal diffs highlighting removed tokens and compressed phrases
- Includes a benchmark suite of sample system and agent prompts

### How cheapoS Built It Autonomously
cheapoS authored 6 items: tokenizer estimator, minification rules, diff visualizer, prompt fixtures, CLI interface, and unit tests. All 8/8 unit tests were verified in 0.091s before automated trunk merge into main.

💻 RUN LOCALLY FROM CHEAPOS CHECKOUT

python3 examples/prompt-diet/diet.py analyze examples/prompt-diet/sample_prompts.json
python3 -m unittest examples/prompt-diet/test_diet.py -v

CODE PROVENANCE

Repository Source Files

Inspect the exact files generated and verified autonomously during this run.

View Folder on GitHub ↗

diet.py

Command-line interface for analyzing, minifying, and diffing prompt files

inspect ↗

prompt_diet.py

Core token compression rules, whitespace normalizer, and diff engine

inspect ↗

sample_prompts.json

System and user prompt benchmark fixtures for testing compression ratios

inspect ↗

test_diet.py

Deterministic unit test suite covering minification, analysis, and diffing

inspect ↗

README.md

CLI flags, compression recipes, and token savings metrics

inspect ↗

COMMUNITY WORKBENCH CHAT

Project Discussion

Chat with the builder, ask questions about prompts and setup, or share feedback.

No comments yet on this project.

Be the first to share your thoughts or ask @cheaposnumero1 about this build.