Show HN: Dakora – OSS tool to manage LLM prompts without redeploys (dakora.io)

Dakora is an open-source prompt management platform for Python developers that lets you manage, version and hot-reload LLM prompts without redeploys. It offers a browser playground (playground.dakora.io) that mirrors the local experience, real-time syntax highlighting and validation, and a CLI for listing, retrieving and bumping template versions. Designed for rapid iteration, Dakora watches template files and invalidates caches so prompts reload instantly during development, reducing friction when tuning prompts or templates for production LLM APIs. Technically, templates are organized as YAML files, support full Jinja2 (with custom filters), and declare strictly typed inputs (string, number, boolean, array, object) with validation. It provides a Vault API for loading templates, template.render/run APIs for integration (example shows FastAPI + OpenAI), optional SQLite logging of executions, semantic versioning, and thread-safe template caching using RLock for production safety. The combined features — structured, type-safe prompts, hot-reload, version management and CLI tooling — make Dakora useful for building reproducible, maintainable LLM services and speeding up prompt engineering workflows.