features, as well as some new features, bug fixes and more.
Lanarky has shifted its focus to becoming the web framework to building LLM-based microservices.
LLMs are here to stay and there's technology stack gap for building applications with LLMs
that are not limited to chatbots. Lanarky aims to fill that gap.
Let's take a look at what's changed.
What's Changed
Features
* **Adapters API**: allows Lanarky to support multiple LLM tooling frameworks. The support is
currently limited to OpenAI and LangChain, but more will be added in future releases.
* **`EventSource` Protocol**: Lanarky now follows the `EventSource` protocol for building streaming
microservices using Server-sent events.
* **LLM-based Dependency Injection**: This is an exciting new feature where Lanarky extends FastAPI's
dependency injection to use LLM functions as dependencies.
* **FastAPI Backport**: If you want to build complex microservices which are currently not supported
by Lanarky routers, you can use the low-level modules to build your microservice with FastAPI.
Documentation
Lanarky documentation has a new home at https://lanarky.ajndkr.com.
Breaking Changes
* `LangChainRouter` has been refactored into the LangChain Adapter API
* Registry for LangChain callback handlers has been removed (including the auto-detect feature
in the `StreamingResponse` class). The decision was made based on user feedback.
* Gradio-based playground UI has been removed. The current implementation was not agnostic to
microservice logic and this feature needs to be reworked.
Changelog
* build: v0.8 release by ajndkr in https://github.com/ajndkr/lanarky/pull/160
**Full Changelog**: https://github.com/ajndkr/lanarky/compare/v0.7.17...v0.8.0