TrinityAI UltraPromptML compiler - Transform .upml files into hyper-efficient LLM-optimized prompts with 60-65% token reduction
npm install ultraprompt

Transform .upml files into hyper-efficient, LLM-optimized prompts using a unified signals engine and streamlined CLI.
> v1.0.14 Update: Major performance improvements with signal caching (40-60% faster), unified API functions (62% code reduction), enhanced error diagnostics, and configurable recursion limits. Full settings documentation and schema versioning added.
``bash`Install globally with npm
npm install -g ultraprompt
`bash
π Minimal UPML Example
`xml
SYSTEM: [ATTRIBUTE: IDENTITY] {
NAME: "Aurora"
AUTHOR: "TrinityAI Research"
AXIOMATIC: "You are Aurora, a concise and helpful assistant."
TONE: "Professional and friendly"
CUTOFF_DATE: "2025-01-01"
CURRENT_DATE: "${getCurrentTimestamp()}"
}
`Compiles to a compact, signal-boosted form:
`
SYSTEM:{NM:"Aurora",AUTHOR:"TrinityAI Research",AX:"ur Aurora, a concise and helpful assistant.",TONE:"Professional and friendly",RK:"2025-01-01",CD:"${getCurrentTimestamp()}"}
`π― Key Features
- Unified signals: Single mapping source in
src/settings.ultraprompt.json drives all abbreviations.
- Aggressive application: Signals apply case-insensitively across the entire output, including inside quotes.
- Attribute system: [ ... ] prelude hoisted to the first tag as attributes using attribute_signals.
- Punctuation cleanup: Removes spaces around punctuation, and removes hyphens/apostrophes.
- Structural emphasis: Uppercases content inside <> and [] in the compiled output.
- Contractions/web-speak: Built-in mappings like you are -> ur, cannot -> cant, etc.π Token Savings Example
| Format | Tokens | Reduction |
|--------|--------|-----------|
| Original UPML | 245 tokens | - |
| Compiled Output | 89 tokens | 63.7% |
| Manual Optimization | 156 tokens | 36.3% |
ποΈ Configuration
-
src/settings.ultraprompt.json is the single source of truth.
- Sections:
- quickstart_template: a string containing a starter UPML template wrapped in { ... }.
- signals: unified mapping for tags, keys, and NL terms (multi-word supported).
- attribute_signals: mappings used when hoisting attributes from the section prelude.CLI commands update this file as needed (e.g.,
upml signal βFull Formβ ABBR).π Documentation
For detailed UPML syntax and advanced features, see documentation.md.
$3
- Add a new mapping:
upml signal "Full Term" ABBR
- Edit mappings: open src/settings.ultraprompt.json and modify signals or attribute_signals.π€ Contributing
Contributions are welcome! Please feel free to submit a Pull Request.
1. Fork the repository
2. Create your feature branch (
git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature`)This project is licensed under the MIT License - see the LICENSE file for details.
UltraPromptML is developed by TrinityAI Research, specializing in advanced LLM development.
- GitHub Repository
- npm Package
- Issue Tracker
- Full Documentation