feat: add media-library-auto-alt-text function example

[bramdoppen]

Description

This PR introduces an example function for the Media Library. Its purpose is to demonstrate Media Library functions, how to properly access image assets via the asset container, and how to generate localized alt-text.

The example intentionally focuses on clarity and guidance rather than prescribing a rigid pattern. It uses agent.action.prompt and enforces JSON-shaped output, but because AI models may not always return strict JSON, this should be viewed as a working example rather than a guaranteed contract. Everyone may also choose to add in their own functions that enforce JSON, the function is designed flexibly enough for that.

It also avoids using agent.action.translate because that action currently expects a full document and is not yet compatible with Media Library use cases. Once Media Library and agent actions evolve, we can update this example accordingly. Until then, this version provides valuable early guidance and unblocks everyone who need a functional starting point.

The PR also introduces an internationalized array structure for alt text (mirroring the Studio’s internationalized array plugin), so the frontend can query language-tagged items in a familiar format. This format is approved internally by the Media Library team.

At the end is a video demonstration that visually shows how the alt-text appears after upload.

---

What to review

• Review the example function with the understanding that it is inspirational,
• Confirm that dereferencing from the asset container to the image asset is correct and aligns with how Media Library stores keywords on the actual image.
• Ensure that the use of agent.action.prompt is appropriate given its current behavior and limitations.
• Check that explanations are clear enough for everyone who want a practical starting point.

---

Testing

Automated tests were not added.

The function was manually tested in a Media Library environment and works end-to-end:

• All text fields are generated for the provided languages
• JSON output is enforced through prompt instructions (with the understanding that AI may occasionally deviate)
• The structure was reviewed with the Media Library team to ensure consistency with expected behavior

This manual testing approach is sufficient for an inspirational example that everyone can adapt.

---

Notes for release

This adds an inspirational, fully working example function for the Media Library to help everyone understand:

• How to access the underlying image asset from the asset container
• How to generate multilingual values via agent.action.prompt
• How to structure internationalized text arrays consistently

Limitations:

• agent.action.prompt is instructed to produce JSON, but AI responses may still vary
• agent.action.translate cannot be used yet because it requires a full document input
• This is intended as an early example to help everyone get started; we can revise it as agent actions evolve

Video:

Screen.Recording.2025-11-28.at.14.46.14.mov

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Preview	Comments	Updated (UTC)
page-building-studio	Ready	Preview	Comment	Nov 28, 2025 3:22pm
test-studio	Ready	Preview	Comment	Nov 28, 2025 3:22pm

3 Skipped Deployments

Project	Deployment	Preview	Comments	Updated (UTC)
e2e-studio	Ignored			Nov 28, 2025 3:22pm
studio-workshop	Ignored	Preview		Nov 28, 2025 3:22pm
test-next-studio	Ignored			Nov 28, 2025 3:22pm

[github-actions]

🧪 E2E Preview environment

🔑 Environment Variables for Local Testing

This is the preview URL for the E2E tests: https://e2e-studio-7tfvw5ddb.sanity.dev

To run the E2E tests locally, you can use the following environment variables, then run pnpm test:e2e --ui to open the Playwright test runner.

💬 Remember to build the project first with pnpm build:e2e.

  SANITY_E2E_PROJECT_ID=ittbm412
  SANITY_E2E_BASE_URL=https://e2e-studio-7tfvw5ddb.sanity.dev
  SANITY_E2E_DATASET="update depending the project you want to test (pr-11337-chromium-19767755953 || pr-11337-firefox-19767755953 )"
  SANITY_E2E_DATASET_CHROMIUM=pr-11337-chromium-19767755953
  SANITY_E2E_DATASET_FIREFOX=pr-11337-firefox-19767755953

[github-actions]

📊 Playwright Test Report

Download Full E2E Report

This report contains test results, including videos of failing tests.

[github-actions]

⚡️ Editor Performance Report

Updated Fri, 28 Nov 2025 15:35:27 GMT

Benchmark	reference latency of sanity@latest	experiment latency of this branch	Δ (%) latency difference
article (title)	27.4 efps (37ms)	29.4 efps (34ms)	-3ms (-6.8%)	✅
article (body)	39.5 efps (25ms)	38.7 efps (26ms)	+1ms (+2.2%)	✅
article (string inside object)	28.6 efps (35ms)	29.4 efps (34ms)	-1ms (-2.9%)	✅
article (string inside array)	26.3 efps (38ms)	24.4 efps (41ms)	+3ms (+7.9%)	✅
recipe (name)	43.5 efps (23ms)	47.6 efps (21ms)	-2ms (-8.7%)	✅
recipe (description)	71.4 efps (14ms)	71.4 efps (14ms)	+0ms (-/-%)	✅
recipe (instructions)	99.9+ efps (6ms)	99.9+ efps (5ms)	-1ms (-/-%)	✅
singleString (stringField)	58.8 efps (17ms)	62.5 efps (16ms)	-1ms (-5.9%)	✅
synthetic (title)	17.2 efps (58ms)	17.5 efps (57ms)	-1ms (-1.7%)	✅
synthetic (string inside object)	18.5 efps (54ms)	17.7 efps (57ms)	+3ms (+4.6%)	✅

efps — editor "frames per second". The number of updates assumed to be possible within a second.

Derived from input latency. efps = 1000 / input_latency

Detailed information

🏠 Reference result

The performance result of sanity@latest

Benchmark	latency	p75	p90	p99	blocking time	test duration
article (title)	37ms	42ms	64ms	89ms	15ms	9.5s
article (body)	25ms	32ms	64ms	121ms	124ms	5.9s
article (string inside object)	35ms	39ms	50ms	84ms	9ms	5.8s
article (string inside array)	38ms	40ms	48ms	90ms	0ms	6.0s
recipe (name)	23ms	26ms	33ms	54ms	0ms	7.3s
recipe (description)	14ms	18ms	20ms	22ms	0ms	4.0s
recipe (instructions)	6ms	10ms	11ms	26ms	0ms	3.0s
singleString (stringField)	17ms	19ms	22ms	35ms	0ms	6.7s
synthetic (title)	58ms	61ms	66ms	114ms	216ms	16.0s
synthetic (string inside object)	54ms	57ms	62ms	261ms	445ms	7.5s

🧪 Experiment result

The performance result of this branch

Benchmark	latency	p75	p90	p99	blocking time	test duration
article (title)	34ms	37ms	39ms	77ms	24ms	9.0s
article (body)	26ms	36ms	78ms	122ms	134ms	6.2s
article (string inside object)	34ms	37ms	55ms	100ms	13ms	5.8s
article (string inside array)	41ms	46ms	62ms	96ms	3ms	6.2s
recipe (name)	21ms	24ms	26ms	44ms	0ms	7.2s
recipe (description)	14ms	17ms	20ms	31ms	0ms	4.0s
recipe (instructions)	5ms	9ms	11ms	14ms	0ms	3.0s
singleString (stringField)	16ms	19ms	21ms	26ms	0ms	7.1s
synthetic (title)	57ms	58ms	60ms	88ms	197ms	14.3s
synthetic (string inside object)	57ms	59ms	71ms	264ms	476ms	7.5s

📚 Glossary

column definitions

• benchmark — the name of the test, e.g. "article", followed by the label of the field being measured, e.g. "(title)".
• latency — the time between when a key was pressed and when it was rendered. derived from a set of samples. the median (p50) is shown to show the most common latency.
• p75 — the 75th percentile of the input latency in the test run. 75% of the sampled inputs in this benchmark were processed faster than this value. this provides insight into the upper range of typical performance.
• p90 — the 90th percentile of the input latency in the test run. 90% of the sampled inputs were faster than this. this metric helps identify slower interactions that occurred less frequently during the benchmark.
• p99 — the 99th percentile of the input latency in the test run. only 1% of sampled inputs were slower than this. this represents the worst-case scenarios encountered during the benchmark, useful for identifying potential performance outliers.
• blocking time — the total time during which the main thread was blocked, preventing user input and UI updates. this metric helps identify performance bottlenecks that may cause the interface to feel unresponsive.
• test duration — how long the test run took to complete.