A key player in the emerging LLM tech stack offers dev-friendly Python and TypeScript SDKs to its huge user base

How Unstructured Makes Its Enterprise API More Accessible With Speakeasy

“You are the front-end of our first paid product.”

Austin Walker, Software Engineer at Unstructured

Unstructured (opens in a new tab) enables companies to talk to their data. They extract unstructured data from any file format and transform it into AI-friendly JSON files that can be passed to LLMs. They’re currently the only company doing this kind of AI pre-processing – and have seen tremendous success. In fact, since the summer of 2022, Unstructured’s OSS library has been downloaded 6.4M times and used by 27,000 organizations.

When the Unstructured team was working on building their enterprise API, they wanted to provide a seamless, high-class experience to their enterprise customers every step of the way. They understood the massive difference that SDKs can make for streamlining and elevating the API integration experience and started investigating how to offer them to their customers.

Choosing an SDK Partner

They decided against building SDKs in-house, as it would create a massive burden for the team and distract focus from other core development priorities.

By creating and maintaining their SDKs in the Speakeasy platform, the Unstructured dev team was able to save time and cognitive overhead. This has proven to be the right decision considering the rapid growth Unstructured has been enjoying.

Additionally, Unstructured wanted their SDK partner to be highly experienced with OpenAPI and advise them on spec management for a public API. They chose Speakeasy because it ticked all these boxes and provided the expertise they were looking for.

API Experience Plan

While planning their top-quality API experience, the Unstructured team wanted to focus on a few key points:

• Comprehensive documentation - With code snippets at the heart, story-driven, easily digestible, and always up-to-date. Unstructured’s API includes multiple parameters that make it challenging to maintain versioning in the API. With Speakeasy, Unstructured can automate the creation of high-quality docs with minimal team effort.
• Robust Error Handling - As the company deals with many file formats, occasional errors are bound to happen. They wanted the error-handling experience to be smooth and informative so that users could quickly unblock themselves. This also drives down the support burden for Unstructured’s team.
• Automated Retries - Since Unstructured’s users often have long-running file uploads, ensuring that the SDK has appropriate retry logic “out of the box” was critical. Speakeasy enables customers to offer features like retries, pagination, and auth in their SDKs with incredible ease – Unstructured just needed to add a few lines of configuration and a few minutes of work to add retries to their SDKs.

"We wanted to point Speakeasy to our API spec and just set it and forget it. As the new features make it into the library, we add them to the API, so it continues to grow. To be able to do all that and have Typescript and Python clients that stay up-to-date has been amazing."

Austin Walker, Software Engineer at Unstructured

With Speakeasy, Unstructured is able to maintain Typescript and Python SDKs that offer an amazing integration experience to their users. These SDKs are now the primary way developers are directed to integrate with Unstructured. Great API accessibility is paramount to Unstructured’s product usage, and having a partner like Speakeasy that they can fully rely on has enabled the team to continue developing their product while knowing that their users are equipped with all the necessary tools.