mirror of
https://github.com/moltbot/moltbot.git
synced 2026-03-09 15:35:17 +00:00
revert(agents): revert base64 image validation (#19221)
This commit is contained in:
@@ -2,153 +2,6 @@ import sharp from "sharp";
|
||||
import { describe, expect, it } from "vitest";
|
||||
import { sanitizeContentBlocksImages, sanitizeImageBlocks } from "./tool-images.js";
|
||||
|
||||
describe("base64 validation", () => {
|
||||
it("rejects invalid base64 characters and replaces with error text", async () => {
|
||||
const blocks = [
|
||||
{
|
||||
type: "image" as const,
|
||||
data: "not-valid-base64!!!@#$%",
|
||||
mimeType: "image/png",
|
||||
},
|
||||
];
|
||||
|
||||
const out = await sanitizeContentBlocksImages(blocks, "test");
|
||||
expect(out.length).toBe(1);
|
||||
expect(out[0].type).toBe("text");
|
||||
if (out[0].type === "text") {
|
||||
expect(out[0].text).toContain("omitted image payload");
|
||||
expect(out[0].text).toContain("invalid");
|
||||
}
|
||||
});
|
||||
|
||||
it("strips data URL prefix and processes valid base64", async () => {
|
||||
// Create a small valid image
|
||||
const jpeg = await sharp({
|
||||
create: {
|
||||
width: 10,
|
||||
height: 10,
|
||||
channels: 3,
|
||||
background: { r: 255, g: 0, b: 0 },
|
||||
},
|
||||
})
|
||||
.jpeg()
|
||||
.toBuffer();
|
||||
|
||||
const base64 = jpeg.toString("base64");
|
||||
const dataUrl = `data:image/jpeg;base64,${base64}`;
|
||||
|
||||
const blocks = [
|
||||
{
|
||||
type: "image" as const,
|
||||
data: dataUrl,
|
||||
mimeType: "image/jpeg",
|
||||
},
|
||||
];
|
||||
|
||||
const out = await sanitizeContentBlocksImages(blocks, "test");
|
||||
expect(out.length).toBe(1);
|
||||
expect(out[0].type).toBe("image");
|
||||
});
|
||||
|
||||
it("rejects base64 with invalid padding", async () => {
|
||||
const blocks = [
|
||||
{
|
||||
type: "image" as const,
|
||||
data: "SGVsbG8===", // too many padding chars
|
||||
mimeType: "image/png",
|
||||
},
|
||||
];
|
||||
|
||||
const out = await sanitizeContentBlocksImages(blocks, "test");
|
||||
expect(out.length).toBe(1);
|
||||
expect(out[0].type).toBe("text");
|
||||
if (out[0].type === "text") {
|
||||
expect(out[0].text).toContain("omitted image payload");
|
||||
}
|
||||
});
|
||||
|
||||
it("rejects base64 with padding in wrong position", async () => {
|
||||
const blocks = [
|
||||
{
|
||||
type: "image" as const,
|
||||
data: "SGVs=bG8=", // = in middle is invalid
|
||||
mimeType: "image/png",
|
||||
},
|
||||
];
|
||||
|
||||
const out = await sanitizeContentBlocksImages(blocks, "test");
|
||||
expect(out.length).toBe(1);
|
||||
expect(out[0].type).toBe("text");
|
||||
if (out[0].type === "text") {
|
||||
expect(out[0].text).toContain("omitted image payload");
|
||||
}
|
||||
});
|
||||
|
||||
it("normalizes URL-safe base64 to standard base64", async () => {
|
||||
// Create a small valid image
|
||||
const jpeg = await sharp({
|
||||
create: {
|
||||
width: 10,
|
||||
height: 10,
|
||||
channels: 3,
|
||||
background: { r: 255, g: 0, b: 0 },
|
||||
},
|
||||
})
|
||||
.jpeg()
|
||||
.toBuffer();
|
||||
|
||||
// Convert to URL-safe base64 (replace + with -, / with _)
|
||||
const standardBase64 = jpeg.toString("base64");
|
||||
const urlSafeBase64 = standardBase64.replace(/\+/g, "-").replace(/\//g, "_");
|
||||
|
||||
const blocks = [
|
||||
{
|
||||
type: "image" as const,
|
||||
data: urlSafeBase64,
|
||||
mimeType: "image/jpeg",
|
||||
},
|
||||
];
|
||||
|
||||
const out = await sanitizeContentBlocksImages(blocks, "test");
|
||||
expect(out.length).toBe(1);
|
||||
expect(out[0].type).toBe("image");
|
||||
});
|
||||
|
||||
it("rejects base64 with invalid length", async () => {
|
||||
const blocks = [
|
||||
{
|
||||
type: "image" as const,
|
||||
data: "AAAAA", // length 5 without padding is invalid (remainder 1)
|
||||
mimeType: "image/png",
|
||||
},
|
||||
];
|
||||
|
||||
const out = await sanitizeContentBlocksImages(blocks, "test");
|
||||
expect(out.length).toBe(1);
|
||||
expect(out[0].type).toBe("text");
|
||||
if (out[0].type === "text") {
|
||||
expect(out[0].text).toContain("omitted image payload");
|
||||
}
|
||||
});
|
||||
|
||||
it("handles empty base64 data gracefully", async () => {
|
||||
const blocks = [
|
||||
{
|
||||
type: "image" as const,
|
||||
data: " ",
|
||||
mimeType: "image/png",
|
||||
},
|
||||
];
|
||||
|
||||
const out = await sanitizeContentBlocksImages(blocks, "test");
|
||||
expect(out.length).toBe(1);
|
||||
expect(out[0].type).toBe("text");
|
||||
if (out[0].type === "text") {
|
||||
expect(out[0].text).toContain("omitted empty image payload");
|
||||
}
|
||||
});
|
||||
});
|
||||
|
||||
describe("tool image sanitizing", () => {
|
||||
it("shrinks oversized images to <=5MB", async () => {
|
||||
const width = 2800;
|
||||
|
||||
@@ -17,55 +17,6 @@ const MAX_IMAGE_DIMENSION_PX = 2000;
|
||||
const MAX_IMAGE_BYTES = 5 * 1024 * 1024;
|
||||
const log = createSubsystemLogger("agents/tool-images");
|
||||
|
||||
// Valid base64: alphanumeric, +, /, with 0-2 trailing = padding only
|
||||
// This regex ensures = only appears at the end as valid padding
|
||||
const BASE64_REGEX = /^[A-Za-z0-9+/]*={0,2}$/;
|
||||
|
||||
/**
|
||||
* Validates and normalizes base64 image data before processing.
|
||||
* - Strips data URL prefixes (e.g., "data:image/png;base64,")
|
||||
* - Converts URL-safe base64 to standard base64 (- → +, _ → /)
|
||||
* - Validates base64 character set and structure
|
||||
* - Ensures the string is not empty after trimming
|
||||
*
|
||||
* Returns the cleaned base64 string or throws an error if invalid.
|
||||
*/
|
||||
function validateAndNormalizeBase64(base64: string): string {
|
||||
let data = base64.trim();
|
||||
|
||||
// Strip data URL prefix if present (e.g., "data:image/png;base64,...")
|
||||
const dataUrlMatch = data.match(/^data:[^;]+;base64,(.*)$/i);
|
||||
if (dataUrlMatch) {
|
||||
data = dataUrlMatch[1].trim();
|
||||
}
|
||||
|
||||
if (!data) {
|
||||
throw new Error("Base64 data is empty");
|
||||
}
|
||||
|
||||
// Normalize URL-safe base64 to standard base64
|
||||
// URL-safe uses - instead of + and _ instead of /
|
||||
data = data.replace(/-/g, "+").replace(/_/g, "/");
|
||||
|
||||
// Check for valid base64 characters and structure
|
||||
// The regex ensures = only appears as 0-2 trailing padding chars
|
||||
// Node's Buffer.from silently ignores invalid chars, but Anthropic API rejects them
|
||||
if (!BASE64_REGEX.test(data)) {
|
||||
throw new Error("Base64 data contains invalid characters or malformed padding");
|
||||
}
|
||||
|
||||
// Check that length is valid for base64 (must be multiple of 4 when padded)
|
||||
// Remove padding for length check, then verify
|
||||
const withoutPadding = data.replace(/=+$/, "");
|
||||
const remainder = withoutPadding.length % 4;
|
||||
if (remainder === 1) {
|
||||
// A single char remainder is always invalid in base64
|
||||
throw new Error("Base64 data has invalid length");
|
||||
}
|
||||
|
||||
return data;
|
||||
}
|
||||
|
||||
function isImageBlock(block: unknown): block is ImageContentBlock {
|
||||
if (!block || typeof block !== "object") {
|
||||
return false;
|
||||
@@ -209,8 +160,8 @@ export async function sanitizeContentBlocksImages(
|
||||
continue;
|
||||
}
|
||||
|
||||
const rawData = block.data.trim();
|
||||
if (!rawData) {
|
||||
const data = block.data.trim();
|
||||
if (!data) {
|
||||
out.push({
|
||||
type: "text",
|
||||
text: `[${label}] omitted empty image payload`,
|
||||
@@ -219,11 +170,6 @@ export async function sanitizeContentBlocksImages(
|
||||
}
|
||||
|
||||
try {
|
||||
// Validate and normalize base64 before processing
|
||||
// This catches invalid base64 that Buffer.from() would silently accept
|
||||
// but Anthropic's API would reject, preventing permanent session corruption
|
||||
const data = validateAndNormalizeBase64(rawData);
|
||||
|
||||
const inferredMimeType = inferMimeTypeFromBase64(data);
|
||||
const mimeType = inferredMimeType ?? block.mimeType;
|
||||
const resized = await resizeImageBase64IfNeeded({
|
||||
|
||||
Reference in New Issue
Block a user