開発スタッフがアプリ関連の最新技術やニュースをご紹介します。

In the intricate world of AI development, managing diverse models has been a persistent challenge. Portkey’s AI Gateway, an open-source solution, tackles this complexity head-on, offering a streamlined approach for developers working with over 100 models. Let’s take a look.

Unified API for Simplified Interactions

Portkey’s AI Gateway serves as the bridge between applications and Language Model Models (LLMs). It excels in simplifying API requests to various providers like OpenAI, Anthropic, Mistral, and more, providing a unified API that eliminates the hassle of dealing with different signatures.

Smart Load Balancing for Optimal Performance

A standout feature is its ability to distribute requests across multiple API keys and providers seamlessly, ensuring optimal performance and a smoother workflow without bottlenecks.

Robust Error Handling for Enhanced Reliability

Portkey’s AI Gateway gracefully handles errors through fallbacks and automatic retries. In case of failure with a specific provider or model, it seamlessly shifts to alternatives, enhancing overall system resilience.

Customizable Middleware Functions for Tailored Solutions

For added flexibility, developers can incorporate custom middleware functions, tailoring adjustments to meet specific application requirements.

Proven Reliability in Real-World Scenarios

Tested rigorously with over 100 billion tokens in real-world scenarios, Portkey’s AI Gateway brings practical reliability to large-scale AI applications.

Broad Language, Provider, and Framework Support

Currently supported languages include Javascript, Python, Go, Java, Rust, and Ruby. Providers range from OpenAI, Anthropic, Azure OpenAI to Google Palm, Together AI, and more. Portkey’s AI Gateway also supports frameworks like Langchain and Llamaindex, providing a comprehensive environment for diverse development needs.

Conclusion

Portkey’s AI Gateway emerges as an efficient solution for developers navigating the complexities of diverse AI models. With its universal API, load balancing, error handling, and customization options, it sets the stage for a streamlined and resilient AI development process. Explore Portkey’s AI Gateway features on GitHub and let Portkey’s AI Gateway guide you into the future of large-scale AI development.

Hope you enjoy that.
Asahi

Tweet

Microsoft’s AI-powered chatbot, Copilot, has joined forces with the GenAI music app Suno to enable users to compose songs effortlessly. By inputting prompts such as “Create a pop song about adventures with your family” into Copilot, users can utilize the Suno plug-in to transform their ideas into complete songs, complete with lyrics, instrumentals, and singing voices. This integration is accessible through Microsoft Edge by visiting Copilot.Microsoft.com, logging in with a Microsoft account, and enabling the Suno plug-in or clicking on the Suno logo labeled “Make music with Suno.” The collaboration aims to democratize music creation, with the experience rolling out to users progressively.

The partnership between Microsoft Copilot and Suno reflects the growing trend among tech giants and startups to invest in GenAI-driven music creation technology. In November, Google AI lab DeepMind and YouTube introduced Lyria, a GenAI model for music, and Dream Track, a tool for building AI tunes in YouTube Shorts. Meta has also explored AI music generation experiments. Despite this surge in innovation, ethical and legal issues surrounding AI-synthesized music persist. Questions of consent, compensation for artists, and the potential exploitation of creators remain unresolved, leading to debates about fair use and intellectual property rights.

Suno, while contributing to the evolving landscape of AI-generated music, faces scrutiny for not disclosing its AI training data sources on its website. Unlike some competitors, Suno does not restrict users from entering prompts like “in the style of [artist],” raising concerns about the potential misuse of copyrighted material. As debates over usage rights unfold in legal forums, viral homemade tracks utilizing GenAI to replicate authentic sounds have surfaced, prompting music labels to address intellectual property concerns with streaming platforms. While legal clarity on the status of GenAI music may emerge through court decisions, a newly introduced Senate bill seeks to provide artists with recourse when their digital likenesses, including musical styles, are used without permission.

You can read the full detail article from microsoft here.

Yuuma

Tweet

In the world of software development, commit messages play a crucial role in tracking changes, understanding the evolution of code, and collaborating effectively within a team. However, writing meaningful commit messages can be a tedious task, often resulting in generic or uninspiring messages. This is where OpenCommit comes into play, aiming to eliminate lame commits with the power of artificial intelligence.

Auto-generate Meaningful Commits in 1 Second

OpenCommit is an innovative tool that leverages AI to automatically generate meaningful commit messages for your code changes. The primary goal is to enhance the quality of commit messages, making them more descriptive and relevant. By doing so, OpenCommit contributes to better code documentation and collaboration.

Getting Started with OpenCommit

Setting up OpenCommit is a breeze, and you can start using it as a command-line interface (CLI) tool. The process involves a few simple steps:

1. Install OpenCommit Globally

   npm install -g opencommit

2. Obtain Your OpenAI API Key

   Get your API key from OpenAI and ensure that you add your payment details to enable API functionality.

3. Set API Key in OpenCommit Config

   oco config set OCO_OPENAI_API_KEY=<your_api_key>

   Your API key is stored locally in the ~/.opencommit config file.

Usage Example

Once OpenCommit is set up, you can use it to generate commit messages for your staged changes. Here’s a quick example:

Stage your changes

git add <files...>

Generate a commit message with OpenCommit

opencommit

Alternatively, you can use the `oco` shortcut:

Stage your changes

git add <files...>

Generate a commit message with OpenCommit using the shortcut

oco

Configuration Options

OpenCommit provides various configuration options to tailor the generated commit messages according to your preferences. These configurations include:

Local per repo configuration

Create a `.env` file in your repository and set OpenCommit config variables such as API key, max response tokens, base path to OpenAI API, description, emoji usage, model selection, language, message template, and prompt module.

OCO_OPENAI_API_KEY=<your OpenAI API token>
OCO_OPENAI_MAX_TOKENS=<max response tokens from OpenAI API>
OCO_OPENAI_BASE_PATH=<may be used to set proxy path to OpenAI api>
OCO_DESCRIPTION=<postface a message with ~3 sentences description of the changes>
OCO_EMOJI=<boolean, add GitMoji>
OCO_MODEL=<either 'gpt-4', 'gpt-3.5-turbo-16k' (default), 'gpt-3.5-turbo-0613' or 'gpt-3.5-turbo'>
OCO_LANGUAGE=<locale, scroll to the bottom to see options>
OCO_MESSAGE_TEMPLATE_PLACEHOLDER=<message template placeholder, default: '$msg'>
OCO_PROMPT_MODULE=<either conventional-commit or @commitlint, default: conventional-commit>

Global Config for All Repos

Set global configurations for OpenCommit, such as the default model and language. But local configurations still take priority, allowing flexibility for specific repository needs.

oco config set OCO_MODEL=gpt-4

Conclusion

OpenCommit is a game-changer in the world of version control. By automating the process of crafting meaningful commit messages, it not only saves time for developers but also contributes to the overall readability and understanding of code repositories. Embrace the power of AI with OpenCommit, and say goodbye to mundane commit messages.

Hope you enjoy that.

Asahi

Tweet

Google’s generative AI chatbot, Bard, is set to receive a substantial upgrade with the integration of Gemini, Google’s latest and most advanced AI model. This update is expected to significantly enhance Bard’s capabilities, providing it with improved reasoning, planning, and understanding, among other features. Gemini, available in Ultra, Pro, and Nano sizes, is designed to run seamlessly across a range of devices, from mobile phones to data centers.

The deployment of Gemini to Bard will occur in two phases. Initially, Bard will be upgraded with a tailored version of Gemini Pro, available in English in over 170 countries. Subsequently, Google plans to introduce Bard Advanced, granting users access to Gemini Ultra, the most powerful AI model. These enhancements mark the most significant quality improvement for Bard since its initial launch, promising heightened proficiency in tasks such as understanding and summarizing content, reasoning, brainstorming, writing, and planning.

Gemini Pro underwent rigorous testing against industry-standard benchmarks, showcasing its superiority over GPT-3.5 in six out of eight tests, including MMLU and GSM8K. However, it’s worth noting that GPT-3.5 is over a year old, framing this launch as more of a catch-up move than a revolutionary leap. Despite this, the improvements aim to position Bard as a more capable and versatile AI, offering users an enhanced and refined experience.

Gemini Pro will initially power text-based prompts in Bard, with plans for future expansion to include multimodal support, incorporating both text and images. Looking ahead to 2024, Bard Advanced will make its debut, featuring Gemini Ultra. This version promises a comprehensive AI experience capable of understanding and acting on various types of information, such as text, images, audio, video, and code. Google intends to initiate a trusted tester program for Bard Advanced before a broader release, emphasizing the importance of safety checks in the development process. The update builds upon several recent improvements to Bard, positioning it as a robust and evolving AI chatbot.

Tweet

text-generation-webui を使用してローカルでLLMを動かし、
AIとチャットを行ってみました。

Tweet