fastkafka VS pydantic

Our journey with FastStream started when we needed to integrate our machine learning models into a customer's Apache Kafka environment. To streamline this process, we created FastKafka using AIOKafka, AsyncAPI, and asyncio. It was our first step in making message queue management easier.

After a short discussion, we concluded we were just too spoiled to use low-level libraries that were nothing more than just tiny wrappers around C++ libs and that we could just build our own. So, we shamelessly made one by reusing beloved paradigms from FastAPI and we shamelessly named it FastKafka. The point was to set the expectations right - you get pretty much what you would expect: function decorators for consumers and producers with type hints specifying Pydantic classes for JSON encoding/decoding, automatic message routing to Kafka brokers and documentation generation.

FastStream simplifies the process of writing producers and consumers for message queues, handling all the parsing, networking and documentation generation automatically. It is a new package based on the ideas and experiences gained from FastKafka and Propan. By joining our forces, we picked up the best from both packages and created a unified way to write services capable of processing streamed data regardless of the underlying protocol. We'll continue to maintain both packages, but new development will be in this project.

FastStream (https://github.com/airtai/faststream) is a new Python framework, born from Propan and FastKafka teams' collaboration (both are deprecated now). It extremely simplifies event-driven system development, handling all the parsing, networking, and documentation generation automatically. Now FastStream supports RabbitMQ and Kafka, but supported brokers are constantly growing (wait for NATS and Redis a bit). FastStream itself is a really great tool to build event-driven services. Also, it has a native FastAPI integration. Just create a StreamRouter (very close to APIRouter) and register event handlers the same with the regular HTTP-endpoints way:

Inspired by FastAPI, FastKafka uses the same paradigms for routing, validation, and documentation, making it easy to learn and integrate into your existing streaming data projects. Please check out the latest version adds supporting the newly released Pydantic v2.0, making it significantly faster. https://github.com/airtai/fastkafka

First, note the method prefix_allowed_tokens_fn. This method applies a Pydantic model to constrain/guide how the LLM generates tokens. Next, see how that constrain can be applied to txtai's LLM pipeline.

utype is a concise alternative of pydantic with simplified parameters and usages, supporting both sync/async functions and generators parsing, and capable of using native logic operators to define logical types like AND/OR/NOT, also provides custom type parsing by register mechanism that supports libraries like pydantic, attrs and dataclasses

I like some of the changes from v1 to v2. But then you have something like this [0] removed from the library without proper documentation or replacement, resulting in ugly workarounds in the link that wont' work properly.

[0]: https://github.com/pydantic/pydantic/discussions/6337

The easiest implementation I found was to use a PyDantic class for my target schema — and use that as a parameter for the method call to “ChatCompletion.create()”. Here’s a fragment of the GinAI Python classes used.

Also, FastStream uses Pydantic to parse input JSON-encoded data into Python objects, making it easy to work with structured data in your applications, so you can serialize your input messages just using type annotations.

Pydantic Validation: Leverage Pydantic's validation capabilities to serialize and validate incoming messages

Not sure if it is exactly related, but there is an open issue on Github for that exact message.

Well the performance increase is so huge because pydantic1 is really really slow. And for using rust, I'd have expected more tbh…

I've been benchmarking pydantic v2 against typedload (which I write) and despite the rust, it still manages to be slower than pure python in some benchmarks.

The ones on the website are still about comparing to v1 because v2 was not out yet at the time of the last release.

pydantic's author will refuse to benchmark any library that is faster (https://github.com/pydantic/pydantic/pull/3264 https://github.com/pydantic/pydantic/pull/1525 https://github.com/pydantic/pydantic/pull/1810) and keep boasting about amazing performances.

On pypy, v2 beta was really really really slow.