Intent to Experiment: Rewriter API

[Domenic Denicola]

Contact emails

dom...@chromium.org, kenji...@chromium.org, m...@chromium.org, btr...@chromium.org, ds...@chromium.org, ay...@chromium.org

Explainer

https://github.com/explainers-by-googlers/writing-assistance-apis/blob/main/README.md

Specification

https://webmachinelearning.github.io/writing-assistance-apis/#rewriter-api

Summary

A JavaScript API for transforming and rephrasing input text in the requested ways, backed by an AI language model.

Blink component

Blink>AI>Rewrite

TAG review

https://github.com/w3ctag/design-reviews/issues/991

TAG review status

Pending

Risks

Interoperability and Compatibility

This feature has definite interoperability and compatibility risks, due to the likelihood that different implementations will use different language models, prompts, and fine-tunings, and even within a single implementation such as Chrome, these pieces will likely change over time. Additionally, not all browsers and operating systems will have a built-in language model to expose, and not all devices will be powerful enough to run one effectively. We are taking a variety of steps to attempt to mitigate these risks. For example, the specification is designed to allow the API to be backed by a cloud-based language model. This approach could extend the functionality to a wider range of devices and users. The API is designed to abstract away the specifics of the underlying language model, including prompts and fine-tuning. This prevents developers from relying on specific outputs, ensuring they receive rewritten text rather than structured data that might vary across implementations. Finally, the API surface is designed with many clear points of failure, that encourage the developer to probe for capabilities ahead of time and fall back to other techniques if a capability is not available. Nevertheless, interoperability and compatibility risk remains high for these sorts of APIs, and we'll be closely monitoring it during the experiment period.

Gecko: Defer (https://github.com/mozilla/standards-positions/issues/1067)

WebKit: No signal (https://github.com/WebKit/standards-positions/issues/393)

Web developers: Mixed signals (https://github.com/WICG/proposals/issues/163) Prototyping with partners behind a flag revealed enthusiasm and many prototypes built, from which we drew the discussion of potential use cases [1]. Feedback on the WICG thread was more mixed. Some themes we saw include: asking for more capabilities (e.g. full prompting of a language model instead of higher-level APIs (our response at [2]); multi-modal support); desire to make sure the API actually works robustly in many real-world use cases; removal of any safety/ethical safeguards; and confusion about client-side vs. cloud APIs. [1]: https://github.com/WICG/writing-assistance-apis/blob/main/README.md#summarizer-api [2]: https://github.com/WICG/writing-assistance-apis/blob/main/README.md#directly-exposing-a-prompt-api

Other signals:

Activation

This feature would definitely benefit from having polyfills, backed by any of: cloud services, lazily-loaded client-side models using WebGPU, or the web developer's own server. We anticipate seeing an ecosystem of such polyfills grow as more developers experiment with this API.

WebView application risks

Does this intent deprecate or change behavior of existing APIs, such that it has potentially high risk for Android WebView-based applications?

None

Goals for experimentation

Although developers have prototyped using the behind-a-flag implementation and given good feedback, several partners are interested in testing out the API with real users. We're looking forward to getting feedback from such testing, especially with regards to output quality, multilingual support, and what types of input data are handled well vs. poorly. We also want to understand whether we've found the right set of options to offer to control the output, and whether the resulting output reflects those controls to the extent that developers expect.

Ongoing technical constraints

None

Debuggability

It is possible that giving DevTools more insight into the nondeterministic states of the model, e.g. random seeds, could help with debugging. See related discussion at https://github.com/explainers-by-googlers/prompt-api/issues/9.

Will this feature be supported on all six Blink platforms (Windows, Mac, Linux, ChromeOS, Android, and Android WebView)?

No

Not all platforms will come with a language model. In particular, in the initial stages we are focusing on Windows, Mac, and Linux.

Is this feature fully tested by web-platform-tests?

No

We plan to expand the web platform test coverage of the API surface over the course of the origin trial. The core algorithm might be difficult to test, given the nondeterministic nature of the output. The explainer discusses this in https://github.com/WICG/writing-assistance-apis/blob/main/README.md#specifications-and-tests.

DevTrial instructions

https://docs.google.com/document/d/1v6-fOC13zS3S-bOLuqIRbzgmia_aPGJl-wzOx4ItSVE/edit?usp=sharing

Flag name on about://flags

rewriter-api-for-gemini-nano

Finch feature name

EnableAIRewriterAPI

Requires code in //chrome?

True

Tracking bug

https://issues.chromium.org/issues/358214322

Non-OSS dependencies

Does the feature depend on any code or APIs outside the Chromium open source repository and its open-source dependencies to function?

Yes: this feature depends on a language model, which is bridged to the open-source parts of the implementation via the interfaces in //services/on_device_model.

Estimated milestones

Origin trial desktop first	137
Origin trial desktop last	142
DevTrial on desktop	129

Anticipated spec changes

Open questions about a feature may be a source of future web compat or interop issues. Please list open issues (e.g. links to known github issues in the project for the feature specification) whose resolution may introduce web compat/interop risk (e.g., changing to naming or structure of the API in a non-backward-compatible way).

At this point all known proposed changes have been incorporated into the specification and implementation.

Link to entry on the Chrome Platform Status

https://chromestatus.com/feature/5112320150470656?gate=5159400139128832

Links to previous Intent discussions

Intent to Prototype: https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAM0wra9Jhe7o59mX5tJPD%3DcZQb2oL3mNi-T57wA86fPXn55OPw%40mail.gmail.com

This intent message was generated by Chrome Platform Status.

[Daniel Bratell]

LGTM

/Daniel

--
You received this message because you are subscribed to the Google Groups "blink-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to blink-dev+...@chromium.org.
To view this discussion visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAM0wra8dnNHXKM26MvQxZ6LBE16PUbvH6-GH5B3Z2WDv4uH0WQ%40mail.gmail.com.