Arthur Kaza, QuickView GenC - Open Framework.pdf
0 likes50 views
Genc is an open-source framework designed for building generative AI applications, enabling the integration of diverse capabilities across multiple languages, platforms, and devices. It enhances development efficiency through composability, portability, and customizability, allowing users to create and deploy applications seamlessly across environments like Android and cloud services. Unlike langchain, which is Python-centric, genc offers a cross-language and cross-platform solution for developers, complementing existing tools rather than competing with them.
Arthur Kaza, QuickView GenC - Open Framework.pdf
- 1. Quick View around GenC An open source framework for building generative AI applications December, 2024
- 2. Meet your speaker ● Engineering Manager, AppDev & AI-Data Analytics, Standard Bank Group ● Ex-Data Scientist, Lead At IAG for BNP PariBas ● AI Researcher, Woxsen University ● Google Developer Group Organizer, Kinshasa @Arthur(Bahati)Kaza @KazaArthur
- 3. ● An open-source framework for building GenAI-powered applications ● A glue framework that enables developers to compose capabilities across different languages, SDKs, platforms, devices & cloud, and connect them together What is GenC
- 4. ● GenAI ecosystem fragmentation: many GenAI capabilities - LLMs, chains, agent logic, custom functions, RAG, ReAct, model cascades and routers, etc. evolve in siloed domains, and do not compose or interoperate with one-another ● GenC connects them together through an abstraction in interfaces, greatly enhance the development efficiency and flexibility Why GenC
- 6. Code example: Combining LangChain as the developer surface with deployment in a Java app on Android, and that combines the use of both on-device and cloud LLMs within the same application: The above yields intermediate representation (IR) that represents application logic in a form that’s portable across platform (e.g., Linux to Android), languages (e.g., Python to Java), from prototyping to production, across cloud and devices. Here’s how one can load and call it in an Android app without changing the code:
- 7. ● Composability & flexibility ○ Ability to freely compose building blocks to express your own application logic, and integrate diverse SDKs, platforms, and ecosystems ● Portability & platform independence ○ Across programming languages (e.g., from Python to Java and C++). ○ Across prototyping and production deployments (e.g., from Colab notebooks, to cloud servers, to mobile apps) ○ Across on-device and cloud platforms Benefits of GenC
- 8. ● Customizability ○ Create your own custom runtimes, and incorporate your own libraries & services (models, databases) as building blocks, and mix and match them ● Development velocity ○ Allow logic to be expressed at high level, and easily modifiable ○ Portability enables prototype code to be deployed with ease ● Performance & security ○ Faster and more secure C++ runtime, asynchronous programming model ○ Application logic can be statically analyzed, verified, optimized, and executed on scalable distributed platforms Benefits of GenC (cont’d)
- 9. GenC FAQ ● Who can be the users of GenC? ○ GenC audience is any GenAI developers, especially those seeking benefits such as composability, portability, and deep customizability ● What features does GenC offer? ○ Building blocks to express a variety of genAI workflows, and the ability to deploy them across a range of platforms, including Android), Cloud, and hybrid on-device + Cloud ● What is the developer experience in using GenC? ○ Write code in LangChain in Python or similar SDK of your choice, one-liner call to translate it into a portable form, and another one-liner call to run it, e.g., in Java on Android (or another platform)
- 10. GenC FAQ (cont’d) ● How does GenC compared to LangChain? ○ GenC has a similar goal of enabling developers to build GenAI applications using high level abstractions and in a modular manner. Unlike Langchain which is a Python-centric SDK, GenC is a cross-language and cross-platform composition framework that users can combine diverse frontend and backend capabilities ● Does GenC aims to compete with LangChain? ○ No, GenC is complementary to Langchain users by enabling them to target deployment environments where Python is not feasible, but it also enable Android developers to use SDKs such as Langchain for more opportunities ● Is using Langchain a must for using GenC? ○ No, one can just use GenC, especially where C++ is strongly preferred
- 11. Thanks for your time !!! Open for Q&A