Deepgram Utilities
The tactus.deepgram module provides helpers for working with Deepgram JSON transcripts. It can flatten transcripts into text, expose structured segments, and extract quotes with timestamps.
11 specs · 0 classes
Standard Library
Built-in Modules for Common Tasks
Tactus includes a standard library of reusable modules for classification, extraction, generation, and more. Each module is written in Tactus itself with comprehensive BDD specifications.
Available Modules
The standard library provides high-quality, tested implementations for common AI workflow tasks. Each module follows the same patterns and can be extended for custom use cases.
Generate Module
The tactus.generate module provides flexible text generation built on DSPy's modular architecture.
4 specs · 2 classes
Retriever Modules
Tactus exposes Biblicus-backed retrievers as Lua modules so you can select a retriever explicitly at the top of your .tac file.
5 classes
Web Research
tactus.web provides reusable web research primitives for agents and host applications. The active first provider is OpenAI Responses web_search.
3 specs · 0 classes
Design Philosophy
Written in Tactus
The stdlib is dogfooding - it's written in Tactus itself. This ensures the language is expressive enough for real-world use and surfaces any pain points.
BDD Specifications
Every module has comprehensive Gherkin specifications that serve as tests, documentation, and contract. If the specs pass, the module works correctly.
Extensible Classes
All modules use a class hierarchy. Extend BaseClassifier, BaseExtractor, or BaseGenerator to create your own implementations with the same interface.
Want to contribute?
The standard library is open source. Improve existing modules or add new ones.