A functional-inspired programming language designed for AI/LLM-first development, featuring context-sensitive keywords, quaternary operators, and schema-driven development.
A concurrent, multi-model programming language designed for synthetic data generation, benchmarking, and evaluation at scale.
Loquora helps orchestrate complex AI conversations, generate high-quality synthetic datasets, and evaluate model performance across multiple providers—all in a type-safe, composable language.
- Parallel dataset creation
- Multi-model orchestration
- Schema-driven generation
- Conversation simulation
- Multi-provider evaluation
- Custom metrics (still designing this)
- Analysis
- A/B testing
- Async-first
- Rate limiting
- Streaming support
- Batch processing
.loq- Main source files
loq(pronounced "lock") - Main CLI and interpreterloqi(pronounced "low-key") - Package managerloqi.toml- Package configuration file
- DESIGN.md - Complete language specification and examples (rough)
- loquora.ebnf - Formal grammar definition (this is NOT being read and parsed dynamically, I'm manually implementing the lexer and parser)
The Loquora lexer and parser are currently implemented with a basic interpreter. The language can be tokenized, parsed into an Abstract Syntax Tree (AST), and execute simple expressions. The cargo run command provides a REPL for interactive development.
Not yet implemented: LLM integration, concurrent execution, synthetic data generation, evaluation frameworks, multi-model orchestration.
To start the interactive REPL, simply run:
cargo runYou can then type Loquora code line by line. The REPL supports multiline input; it will prompt with ...> until a complete statement (ending with a semicolon or a closing brace) is entered. To exit the REPL, type :q, :quit, quit, or exit on an empty prompt line.