Search

Low-latency voice AI inference runtime (STT→LLM→TTS <500ms)

生图 / 视频提示词约束引擎 — 把模糊需求变成可执行的提示词。An Agent Skill for constrained image & video prompt engineering (Nano Banana / Seedance / Veo / Sora / Kling / Midjourney / FLUX).

Rules for an AI coding agent to filter out generic AI-generated UI designs, text, and code.

The storage block family for dirigent: copy, read, write, and wait for an object at a URI.

Sverka storage — file and SQLite snapshot stores for run persistence

The code-coverage backend module for the scorecard plugin.

Local zero-telemetry secret sanitiser and LLM prompt injection firewall

LLM-based semantic chunking library for RAG pipelines

Self-hosted filesystem storage adapter and backup foundation for Payload CMS.

Deterministic, local-first context optimization for LLM prompts: clean, dedupe, structure, and measure token and cost savings before calling a large model.

Placeholder for the sandboxio project. No code in this release.

Boruna: a deterministic, capability-safe programming language and framework for LLM-native applications