can1357/oh-my-pi · error

Artifact ${artifact.id} is ${artifact.size} bytes; full inte

Error message

Artifact ${artifact.id} is ${artifact.size} bytes; full internal resolution is blocked. Use read selectors such as artifact://${artifact.id}:1-3000 or artifact://${artifact.id}:raw:1-3000, and use the artifact file path for search/copy workflows: ${artifact.path}

What it means

Full-text resolution of an artifact through artifact:// is capped at MAX_INLINE_ARTIFACT_BYTES. Larger artifacts are refused so a single read doesn't dump megabytes of tool output into the model context; the error tells you to use line/range selectors or the on-disk path for bulk operations.

Source

Thrown at packages/coding-agent/src/internal-urls/artifact-protocol.ts:119

	async resolve(url: InternalUrl, context?: ResolveContext): Promise<InternalResource> {
		const artifact = await resolveArtifactFile(url, context);

		// Path-only callers (search/grep, bash URL expansion) never touch the
		// artifact bytes. Return the resource shape so those flows keep working
		// on artifacts of any size — only content materialization is gated.
		if (context?.pathOnly) {
			return {
				url: url.href,
				content: "",
				contentType: "text/plain",
				size: artifact.size,
				sourcePath: artifact.path,
			};
		}

		if (artifact.size > MAX_INLINE_ARTIFACT_BYTES) {
			throw new Error(
				`Artifact ${artifact.id} is ${artifact.size} bytes; full internal resolution is blocked. Use read selectors such as artifact://${artifact.id}:1-3000 or artifact://${artifact.id}:raw:1-3000, and use the artifact file path for search/copy workflows: ${artifact.path}`,
			);
		}

		const content = await Bun.file(artifact.path).text();
		return {
			url: url.href,
			content,
			contentType: "text/plain",
			size: artifact.size,
			sourcePath: artifact.path,
		};
	}

	async complete(): Promise<UrlCompletion[]> {
		const ids = new Set<string>();
		for (const dir of artifactsDirsFromRegistry()) {
			let files: string[];

View on GitHub (pinned to 9690622007)

Solutions

1. Read a slice with a selector: artifact://<id>:1-3000
2. Use raw byte ranges: artifact://<id>:raw:1-3000
3. For search/copy/paging workflows, operate on artifact.path directly with grep/read tools
4. Iterate ranges (1-3000, 3001-6000, ...) if you truly need the whole content

Example fix

// before
const all = await resolve('artifact://17')
// after
const first = await resolve('artifact://17:1-3000')
const rest  = await resolve('artifact://17:raw:3001-6000')

const MAX = 256 * 1024 // match MAX_INLINE_ARTIFACT_BYTES
const size = (await Bun.file(artifactPath).stat()).size
const url = size > MAX ? `artifact://${id}:1-3000` : `artifact://${id}`

try {
  return await resolve(`artifact://${id}`)
} catch (err) {
  if (String(err).includes('full internal resolution is blocked')) {
    return resolve(`artifact://${id}:1-3000`)
  }
  throw err
}

When it happens

Trigger: Calling artifact://<id> (no selector) or a selector that resolves the full body when the artifact's stat().size exceeds MAX_INLINE_ARTIFACT_BYTES.

Common situations: Reading a large build log, test output, or generated dump artifact in full; agent tried to ingest a big command-output artifact wholesale instead of slicing it.

Related errors

• Managed skill is ${bytes} bytes; the limit is ${MAX_MANAGED_
• Session too large to share: ${sealed.byteLength} bytes seale
• Artifact ${id} not found. Available: ${availableStr}
• Artifact ${id} resolved to a directory, not a file
• ${name} exceeds the ${SHARPSHOOTER_MAX_FILE_LINES}-line limi

AI-assisted analysis of can1357/oh-my-pi@9690622007 (2026-08-31). Data as JSON: /api/errors/52ebcb9c40d1caaf. Report an issue: GitHub.