[lxsmnsyc]

With enableRichArguments(), a server function can now take files and blobs anywhere in its arguments. This ports solidjs/solid-start#2359 to the core transport.

Problem

A file only reached a server function as the only argument, or as the last argument of a bound call. Anywhere else, the call failed before it was sent:

await upload(file, { tags: ["a"] }); // The value [object File] ... cannot be parsed/serialized
await upload({ file });
await upload(formData, 1);

The codec's default plugin list has no Blob or File plugin.

Changes

• The rich-args encoder finds each file with a seroval plugin made for the call. It looks wherever the codec meets one: arguments, arrays, objects, maps, sets and FormData entries.
• A call with files sends multipart form data. The codec output goes in the args field, and each file goes as it is in its own file<index> part, without base64.
• The new BodyFormat.SerializedFormData ("10") marks the request. The Start PR uses "8", which is Json here.
• The server decodes args with a plugin made for that request, which turns each index back into its part. Any other body decodes without that plugin, so an index in a plain codec body fails to decode.
• A missing args field, a bad index, or a part that is not a file is refused with 400. bodySizeLimit bounds the whole body, which answers the "no total size limit" note on the Start PR.
• A file used twice is sent once and arrives as one value. A blob from another frame is found by its shape and copied. A Blob arrives as a File named blob, with its type kept.
• Response and Request arguments are still refused before sending.
• Default JSON mode is unchanged. Without enableRichArguments(), these calls still throw, and the message now names files.
• serializeArgs may return FormData. The core client only checks for it and picks the format.
• The upload code lives in uploads.ts and ships only in the rich-args entry: +930 B minified there. The core client grows by 97 B minified.
• The version skew spec moves its unknown tag from "10" to "11".
• RFC 10 documents the behavior, and a changeset is included.

The Start PR is still open. If its design changes before it lands, this port should follow.

Testing

• New server-functions-uploads.spec.tsx with 12 cases:
  • a file next to metadata, with raw bytes in the part;
  • files in arrays, objects, maps, sets and FormData;
  • a file used twice;
  • a blob's type;
  • a blob from another frame;
  • a call without files staying as codec text;
  • a refused Response;
  • a 413 over bodySizeLimit;
  • four malformed bodies answered with 400.
• @solidjs/web unit (1395), server (1746) and hydrate (520) suites pass with JSX_COMPILER=babel.
• test-types, the types build and Prettier pass.
• Locally, my native compiler.node predates recent compiler changes, so about 50 unrelated tests fail with it. The same tests fail without this change and pass with Babel.
• I didn't run the size tool locally, because it needs the native compiler. The sizes above come from minifying the built bundles directly.

🤖 Generated with Claude Code

[changeset-bot]

🦋 Changeset detected

Latest commit: 3a27300

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 14 packages

Name	Type
@solidjs/web	Patch
@solidjs/babel-plugin	Patch
@solidjs/diagnostics	Patch
@solidjs/element	Patch
@solidjs/h	Patch
@solidjs/html	Patch
test-integration	Patch
board-example	Patch
server-dbmon-example	Patch
todos-server-example	Patch
@solidjs/compiler	Patch
@solidjs/signals	Patch
solid-js	Patch
@solidjs/universal	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[codspeed]

Merging this PR will degrade performance by 27.21%

2 warnings

⚠️ 4 benchmarks are dominated by system calls

System calls cannot be simulated consistently and are left out of the measure. The reported value understates the real cost. Switch to the Walltime instrument to measure them.

Learn more about system calls

See affected benchmarks

⚠️ Base and head ran on different runner types

Some benchmarks with significant performance changes were compared across different runner types, so the changes might not come from the code. Run base and head on the same runner type, or use CodSpeed Macro Runners for more stable results.

Learn more about Macro Runners

Open the report in CodSpeed to investigate

❌ 2 regressed benchmarks
✅ 194 untouched benchmarks
⏩ 5 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
❌	projection derive: write one NESTED field (reference)	881.6 µs	1,486.5 µs	-40.69%
❌	build + consume	693.9 µs	776.6 µs	-10.65%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing lxsmnsyc:feat/server-function-uploads (b745f10) with next (9f64ca9)}

Footnotes

1. 5 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[ryansolid]

Thanks! This landed as #4208, which carries your commit (credited as co-author) plus one change on top: uploads keep the Serialized tag and the handler spots them by their multipart/form-data content type. That way the core client's request path has no upload branch, and apps that don't use rich arguments pay 0 B (−21 B minified on the server-function scenarios). Closing in favor of it.

— Claude via Claude Code

---

Generated by Claude Code