July 2024 Summaries

Pinecone Connect is now generally available, offering a seamless integration for developers to manage Pinecone resources from other platforms through a straightforward authentication process. This integration is already being utilized by industry leaders like Twilio and Matillion to streamline AI workflows, enabling developers to build sophisticated AI applications with ease. Pinecone Connect's widget allows developers to sign up, log in, and manage projects directly within their platform, improving efficiency by reducing the need to switch between multiple environments. It supports developers in leveraging Pinecone's vector database for creating accurate and secure AI applications while managing infrastructure on their behalf. The integration enables users to quickly import data and embark on their Generative AI journeys, with Pinecone managing the operational aspects. Industry leaders like Twilio and Matillion are using Pinecone Connect to enhance the speed and accuracy of AI applications, offering features like customer-aware autonomous agents and simplified RAG processes, respectively.

Pinecone has introduced a new approach to API versioning, balancing the need for innovation with the stability required by existing integrations. The company views an API as a crucial contract with users, and versioning is key to maintaining predictability and reducing integration disruptions. Pinecone has adopted a date-based versioning schema, communicated via a custom header, to provide clarity and transparency regarding the age and stage of each API iteration. This method allows for more frequent incremental improvements and simplifies the upgrade process for developers. Pinecone plans to release new stable API versions quarterly, with each stable version supported for at least 12 months, while release candidates preview upcoming changes three months in advance. This strategy aims to ensure a seamless and productive experience for developers by minimizing unexpected disruptions and offering clear communication about API evolution.

In a recent RAG Brag episode, Nick Gomez, Co-founder and CEO of Inkeep, discussed the inspiration and evolution of his company, which aims to improve self-help experiences for technical products using AI and machine learning. Drawing on his experiences at Microsoft, he identified a gap in the support tools available for developers, noting that traditional chatbots are often perceived as marketing tools rather than helpful resources. Inkeep distinguishes itself by offering a hands-free, high-quality solution that automatically ingests content from various sources and integrates seamlessly with platforms like Slack or Discord. The company focuses on refining AI capabilities through iterative feedback and overcoming challenges such as content ingestion from diverse formats. Inkeep uses Pinecone for flexible vector storage and retrieval, optimizing both performance and cost efficiency. Looking forward, Gomez envisions generative AI evolving through advanced search, action-based workflows, and eventually autonomous AI agents, emphasizing the importance of addressing real customer needs and adapting to the market.

Pinecone has introduced Inference, an API designed to streamline AI workflows by providing easy access to embedding and reranking models hosted on its infrastructure. This tool complements Pinecone's vector database, making it a comprehensive solution for embedding, managing, and retrieving vector data through a single API, thereby reducing complexity and tool management for developers. Inference is in public preview and initially supports the multilingual-e5-large model, chosen for its open-source nature, multilingual capabilities, and performance across languages. The API simplifies the process of choosing and implementing embedding models, allowing developers to focus on prototyping and deploying AI applications. While Inference aims to accelerate AI development, Pinecone also offers support for dense embeddings with up to 20k dimensions from any model provider. The service is available to all users, with pricing tiers based on usage, and plans to expand with additional models and features in the near future.

Pinecone's June 2024 product update introduces several new features, including the beta launch of Pinecone Assistant, an API service designed to answer complex questions about proprietary data within applications. Developers can join the waitlist to access and provide feedback on the beta, which emphasizes simplicity, high-quality results, and control over data. The update also includes a doubled serverless namespace quota, enhancing support for multitenancy by allowing up to 100,000 namespaces per serverless index. This enables faster responses and reduced costs by limiting operations to a single namespace. Additionally, Pinecone has revamped its Model Gallery for easier selection of models and introduced new sample applications to streamline the development of GenAI apps, including a legal semantic search sample app. Users can try Pinecone for free and access detailed release notes for more information about all product and feature updates.