No Clocks

Keep a Single Source of Truth Regardless of your design approach (design-first or code-first) always keep a single source of truth, i.e., information should not be duplicated in different places. It is really the same concept used in programming, where repeated code should be moved to a common function.

Otherwise, eventually one of the places will be updated while the other won’t, leading to headaches… in the best of cases. For instance, it is also commonplace to use code annotations to generate an OpenAPI description and then commit the latter to source control while the former still lingers in the code. As a result, newcomers to the project will not know which one is actually in use and mistakes will be made. Alternatively, you can use a Continuous Integration test to ensure that the two sources stay consistent.

Add OpenAPI Descriptions to Source Control OpenAPI Descriptions are not just a documentation artifact: they are first-class source files which can drive a great number of automated processes, including boilerplate generation, unit testing and documentation rendering. As such, OADs should be committed to source control, and, in fact, they should be among the first files to be committed. From there, they should also participate in Continuous Integration processes.

Make the OpenAPI Descriptions Available to the Users Beautifully-rendered documentation can be very useful for the users of an API, but sometimes they might want to access the source OAD. For instance, to use tools to generate client code for them, or to build automated bindings for some language. Therefore, making the OAD available to the users is an added bonus for them. The documents that make up the OAD can even be made available through the same API to allow runtime discovery.

There is Seldom Need to Write OpenAPI Descriptions by Hand Since OADs are plain text documents, in an easy-to-read format (be it JSON or YAML), API designers are usually tempted to write them by hand. While there is nothing stopping you from doing this, and, in fact, hand-written API descriptions are usually the most terse and efficient, approaching any big project by such method is highly impractical. Instead, you should try the other existing creation methods and choose the one that better suits you and your team (No YAML or JSON knowledge needed!):

OpenAPI Editors: Be it text editors or GUI editors they usually take care of repetitive tasks, allow you to keep a library of reusable components and provide real-time preview of the generated documentation.

Domain-Specific Languages: As its name indicates, DSL’s are API description languages tailored to specific development fields. A tool is then used to produce the OpenAPI Description. A new language has to be learned, but, in return, extremely concise descriptions can be achieved.

Code Annotations: Most programming languages allow you to annotate the code, be it with specific syntax or with general code comments. These annotations, for example, can be used to extend a method signature with information regarding the API endpoint and HTTP method that lead to it. A tool can then parse the code annotations and generate OADs automatically. This method fits very nicely with the code-first approach, so keep in mind the first advice given at the top of this page when using it (Use a Design-First Approach)…

A Mix of All the Above: It’s perfectly possible to create the bulk of an OpenAPI Description using an editor or DSL and then hand-tune the resulting file. Just be aware of the second advice above (Keep a Single Source of Truth): Once you modify a file it becomes the source of truth and the previous one should be discarded (maybe keep it as backup, but out of the sight and reach of children and newcomers to the project).

Describing Large APIs

Do not repeat yourself (The DRY principle). If the same piece of YAML or JSON appears more than once in the document, it’s time to move it to the components section and reference it from other places using $ref (See Reusing Descriptions. Not only will the resulting document be smaller but it will also be much easier to maintain). Components can be referenced from other documents, so you can even reuse them across different API descriptions!

Split the description into several documents: Smaller files are easier to navigate, but too many of them are equally taxing. The key lies somewhere in the middle. A good rule of thumb is to use the natural hierarchy present in URLs to build your directory structure. For example, put all routes starting with /users (like /users and /users/{id}) in the same file (think of it as a “sub-API”). Bear in mind that some tools might have issues with large files, whereas some other tools might not handle too many files gracefully. The solution will have to take your toolkit into account.

Use tags to keep things organized: Tags have not been described in the Specification chapter, but they can help you arrange your operations and find them faster. A tag is simply a piece of metadata (a unique name and an optional description) that you can attach to operations. Tools, specially GUI editors, can then sort all your API’s operation by their tags to help you keep them organized.

Links to External Best Practices There’s quite a bit of literature about how to organize your API more efficiently. Make sure you check out how other people solved the same issues you are facing now! For example: The API Stylebook contains internal API Design Guidelines shared with the community by some well known companies and government agencies.

Best Practices This page contains general pieces of advice which do not strictly belong to the Specification Explained chapter because they are not directly tied to the OpenAPI Specification (OAS). However, they greatly simplify creating and maintaining OpenAPI Descriptions (OADs), so they are worth keeping in mind.

Use a Design-First Approach Traditionally, two main approaches exist when creating OADs: Code-first and Design-first. In the Code-first approach, the API is first implemented in code, and then its description is created from it, using code comments, code annotations or simply written from scratch. This approach does not require developers to learn another language so it is usually regarded as the easiest one. Conversely, in Design-first, the API description is written first and then the code follows. The first obvious advantages are that the code already has a skeleton upon which to build, and that some tools can provide boilerplate code automatically. There have been a number of heated debates over the relative merits of these two approaches but, in the opinion of the OpenAPI Initiative (OAI), the importance of using Design-first cannot be stressed strongly enough.

The reason is simple: The number of APIs that can be created in code is far superior to what can be described in OpenAPI. To emphasize: OpenAPI is not capable of describing every possible HTTP API, it has limitations. Therefore, unless these descriptive limitations are perfectly known and taken into account when coding the API, they will rear their ugly head later on when trying to create an OpenAPI description for it. At that point, the right fix will be to change the code so that it uses an API which can be actually described with OpenAPI (or switch to Design-first altogether). Sometimes, however, since it is late in the process, it will be preferred to twist the API description so that it matches more or less the actual API. It goes without saying that this leads to unintuitive and incomplete descriptions, that will rarely scale in the future. Finally, there exist a number of validation tools that can verify that the implemented code adheres to the OpenAPI description. Running these tools as part of a Continuous Integration process allows changing the OpenAPI Description with peace of mind, since deviations in the code behavior will be promptly detected. Bottom line: OpenAPI opens the door to a wealth of automated tools. Make sure you use them!

What’s an OpenAI Assistant? Think of it as a software glue that affords you to gel together agent-like capabilities in your applications to conduct tasks expressed as instructions in natural language to an Assistant. Able to understand instructions, it can leverage OpenAI’s SOTA models and tools to carry out tasks. With Assistants stateful API, you can create Assistants within your application, providing you access to three types of supported tools: Code Interpreter, Retrieval, and Function calling [5]. At the core it has few concepts and components that cogently interact together, to enable agent-like capabilities.

Assistants API, concepts, components, and tools Unfortunately, OpenAI documentation falls short in explaining or illustrating these components into finer details and showing how they work together. Randy Michak of Empowerment AI does a fine job of dissecting these core components and illustrating their flow and data interactions [7]. Inspired by Michak, I mildly modified Figure 4, showing dynamic interaction and data flowing among Assistants API components.

To get started with Assistants, the OpenAI guide stipulates four simple steps to use the Assistants API to glue together these core components for coordination [8]. Step 1: Create an Assistant, to declare a custom model and provide instructions for the Assistant. This helps the Assistant to elect the appropriate supported tool to employ. Step 2: Create a Thread, a stateful session for the Assistant to retrieve messages from and add Assistant messages to. Step 3: Use the Thread as a conversational session to add messages for the assistants to consume. Step 4: Run the Assistant on a newly added Thread message to trigger responses. This run is Assistant’s asynchronous runtime environment.

How does it all work together?

Let’s methodically walk through a simple example where we want to accomplish the following: Integrate Assistants API, using Retrieval tool, to a) upload a couple of pdf documents and b) use an Assistant to query the contents of the document. Consider this as a mini Retrieval Augmented Generation (RAG) application. Use Files objects to upload the pdf files so that the Assistant can access them. Create and employ the Assistant, Threads, Messages, and Run objects to query the uploaded pdf documents. Coordinate all these concrete objects to interact and interplay together as part of my application.

Step 1: Create File objects as our knowledge base Upload your PDFs in the retrievers’ database, using a File object. The Assistants API breaks them into parts, as chunks, and saves them, as indexes and vector embeddings. When you ask a question, Retrievers find the best matches and help the Assistant give you a detailed answer, just like a big RAG retriever.

Step 2: Create an Assistant object. To use an Assistant and conduct tasks, first, create an AI Assistant object. Supply the Assistant with a model, instructional behavior, tools to use, and file IDs to employ for its knowledge base, as parameters.

Both OpenAI programing guide and Anyscale Endpoints blog [7] distill down to simple steps: Call the model with the user query and a list of functions defined in the Chat Completions API parameter as tools. The model can choose to call one or more functions; if so, the content will be a stringified JSON object adhering to your custom schema. Parse the string into JSON in your code, and call your function with the provided arguments if they exist. Call the model again by appending the function response as a new message, and let the model summarize the results back to the user. Following the above simple steps, our user_content to the LLM generates three required parameters (location, latitude, longitude) as a JSON object in its response.

Examples and Use Cases of Function Calling in LLM

Apart from the above use cases mentioned in the Open AI programming guide [10], Ben Lorica visually and comprehensively captures use cases of general function calling in LLMs, including the OpenAI Assistant Tools API [11]. Lorica succinctly states that early use cases include applications such as customer service chatbots, data analysis assistants, and code generation tools. Other examples extend to creative, logistical, and operational domains: writing assistants, scheduling agents, summarizing news., etc.

inurl:”/api/v1" site:microsoft.com