[Vertex AI] Refactor generateContentStream to fix Swift 6 warning

[paulb777]

Refactored GenerativeModel.generateContentStream(...) to use init(_:bufferingPolicy:_:) (AsyncStream.Continuation approach) instead of init(unfolding:) to resolve the Swift 6 warning "Capture of responseIterator with non-Sendable type AsyncThrowingStream<GenerateContentResponse, Error>" discussed in #14503.

#no-changelog

[gemini-code-assist]

Hello @paulb777, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

Summary of Changes

This pull request addresses a Swift 6 warning and aims to improve the readability and maintainability of the code. The core change involves refactoring the generateContentStream function in GenerativeModel.swift to use an AsyncThrowingStream with a for try await loop instead of returning an iterator. This allows for more idiomatic stream handling. Additionally, a similar change was made to GenerativeAIService.swift to handle the stream using stream.lines and a for try await loop.

Highlights

• Refactor: The generateContentStream function in GenerativeModel.swift was refactored to use AsyncThrowingStream and a for try await loop for better stream handling.
• Swift 6 Warning: This pull request fixes a Swift 6 warning, as discussed in issue [VertexAI] Add Swift 6 testing to CI #14503.
• Idiomatic Stream Usage: The changes promote more idiomatic use of streams, enhancing code readability and maintainability.

Changelog

Click here to see the changelog

• FirebaseVertexAI/Sources/GenerativeAIService.swift
  • Refactored the stream processing logic to use stream.lines and a for try await loop.
  • The code now iterates through the lines of the stream, parses JSON data, and yields content to the continuation.
  • Error handling is included within the loop to finish the continuation with an error if parsing fails.
  • The continuation.finish(throwing: nil) is now called inside the loop and after the loop.
• FirebaseVertexAI/Sources/GenerativeModel.swift
  • Refactored generateContentStream to return an AsyncThrowingStream directly.
  • The stream is now processed within a Task using a for try await loop.
  • Error handling and prompt feedback checks are performed within the loop.
  • The continuation.yield(response) is called inside the loop to provide valid content.
  • The continuation.finish(throwing: nil) is now called after the loop.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in issue comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist is currently in preview and may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments to provide feedback.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

---

Did you know?

The concept of streams in programming dates back to the early days of Unix, where they were used to connect different programs together.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

The pull request refactors the generateContentStream function to address a Swift 6 warning and improves code readability. The changes are generally well-structured and address the issue described in the pull request description. The use of AsyncThrowingStream and await makes the code more modern and easier to follow.

Summary of Findings

• Error Handling in Streaming: The error handling within the stream.lines loop in GenerativeAIService.swift could be improved to ensure that all potential errors are properly handled and propagated.
• Redundant finish(throwing: nil) calls: The finish(throwing: nil) calls in GenerativeAIService.swift and GenerativeModel.swift appear to be redundant and could be simplified.
• Task Management in GenerativeModel.swift: Consider adding error handling within the Task block in GenerativeModel.swift to ensure that any unexpected errors are caught and handled appropriately.

Assessment

The pull request refactors the generateContentStream function to address a Swift 6 warning and improves code readability. The changes appear to be well-structured and address the issue described in the pull request description. However, there are a few areas where the code could be improved for clarity and error handling. I recommend addressing these comments before merging, and users should have others review and approve this code before merging.

[ncooke3]

[andrewheard]

LGTM to me now. (I just pushed a minor change to rename responseIterator to responseStream for clarity and updated the PR name/description)

One thing I'm not certain about is whether this affects the ability for the Task to get cancelled if the caller stops reading with for try await but I think getting the Swift 6 build in good shape is more important.

Not relevant after Nick's changes