refactor: share responses input handling
This commit is contained in:
Peter Steinberger
385
src/media/input-files.ts
Normal file
385
src/media/input-files.ts
Normal file
@@ -0,0 +1,385 @@
|
||||
import { lookup } from "node:dns/promises";
|
||||
|
||||
import { createCanvas } from "@napi-rs/canvas";
|
||||
import { getDocument } from "pdfjs-dist/legacy/build/pdf.mjs";
|
||||
|
||||
export type InputImageContent = {
|
||||
type: "image";
|
||||
data: string;
|
||||
mimeType: string;
|
||||
};
|
||||
|
||||
export type InputFileExtractResult = {
|
||||
filename: string;
|
||||
text?: string;
|
||||
images?: InputImageContent[];
|
||||
};
|
||||
|
||||
export type InputPdfLimits = {
|
||||
maxPages: number;
|
||||
maxPixels: number;
|
||||
minTextChars: number;
|
||||
};
|
||||
|
||||
export type InputFileLimits = {
|
||||
allowUrl: boolean;
|
||||
allowedMimes: Set<string>;
|
||||
maxBytes: number;
|
||||
maxChars: number;
|
||||
maxRedirects: number;
|
||||
timeoutMs: number;
|
||||
pdf: InputPdfLimits;
|
||||
};
|
||||
|
||||
export type InputImageLimits = {
|
||||
allowUrl: boolean;
|
||||
allowedMimes: Set<string>;
|
||||
maxBytes: number;
|
||||
maxRedirects: number;
|
||||
timeoutMs: number;
|
||||
};
|
||||
|
||||
export type InputImageSource = {
|
||||
type: "base64" | "url";
|
||||
data?: string;
|
||||
url?: string;
|
||||
mediaType?: string;
|
||||
};
|
||||
|
||||
export type InputFileSource = {
|
||||
type: "base64" | "url";
|
||||
data?: string;
|
||||
url?: string;
|
||||
mediaType?: string;
|
||||
filename?: string;
|
||||
};
|
||||
|
||||
export type InputFetchResult = {
|
||||
buffer: Buffer;
|
||||
mimeType: string;
|
||||
contentType?: string;
|
||||
};
|
||||
|
||||
export const DEFAULT_INPUT_IMAGE_MIMES = ["image/jpeg", "image/png", "image/gif", "image/webp"];
|
||||
export const DEFAULT_INPUT_FILE_MIMES = [
|
||||
"text/plain",
|
||||
"text/markdown",
|
||||
"text/html",
|
||||
"text/csv",
|
||||
"application/json",
|
||||
"application/pdf",
|
||||
];
|
||||
export const DEFAULT_INPUT_IMAGE_MAX_BYTES = 10 * 1024 * 1024;
|
||||
export const DEFAULT_INPUT_FILE_MAX_BYTES = 5 * 1024 * 1024;
|
||||
export const DEFAULT_INPUT_FILE_MAX_CHARS = 200_000;
|
||||
export const DEFAULT_INPUT_MAX_REDIRECTS = 3;
|
||||
export const DEFAULT_INPUT_TIMEOUT_MS = 10_000;
|
||||
export const DEFAULT_INPUT_PDF_MAX_PAGES = 4;
|
||||
export const DEFAULT_INPUT_PDF_MAX_PIXELS = 4_000_000;
|
||||
export const DEFAULT_INPUT_PDF_MIN_TEXT_CHARS = 200;
|
||||
|
||||
const PRIVATE_IPV4_PATTERNS = [
|
||||
/^127\./,
|
||||
/^10\./,
|
||||
/^192\.168\./,
|
||||
/^172\.(1[6-9]|2[0-9]|3[0-1])\./,
|
||||
/^0\./,
|
||||
];
|
||||
const PRIVATE_IPV6_PREFIXES = ["::1", "fe80:", "fec0:", "fc", "fd"];
|
||||
|
||||
function isPrivateIpAddress(address: string): boolean {
|
||||
if (address.includes(":")) {
|
||||
const lower = address.toLowerCase();
|
||||
if (lower === "::1") return true;
|
||||
return PRIVATE_IPV6_PREFIXES.some((prefix) => lower.startsWith(prefix));
|
||||
}
|
||||
return PRIVATE_IPV4_PATTERNS.some((pattern) => pattern.test(address));
|
||||
}
|
||||
|
||||
function isBlockedHostname(hostname: string): boolean {
|
||||
const lower = hostname.toLowerCase();
|
||||
return (
|
||||
lower === "localhost" ||
|
||||
lower.endsWith(".localhost") ||
|
||||
lower.endsWith(".local") ||
|
||||
lower.endsWith(".internal")
|
||||
);
|
||||
}
|
||||
|
||||
async function assertPublicHostname(hostname: string): Promise<void> {
|
||||
if (isBlockedHostname(hostname)) {
|
||||
throw new Error(`Blocked hostname: ${hostname}`);
|
||||
}
|
||||
|
||||
const results = await lookup(hostname, { all: true });
|
||||
if (results.length === 0) {
|
||||
throw new Error(`Unable to resolve hostname: ${hostname}`);
|
||||
}
|
||||
for (const entry of results) {
|
||||
if (isPrivateIpAddress(entry.address)) {
|
||||
throw new Error(`Private IP addresses are not allowed: ${entry.address}`);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
function isRedirectStatus(status: number): boolean {
|
||||
return status === 301 || status === 302 || status === 303 || status === 307 || status === 308;
|
||||
}
|
||||
|
||||
export function normalizeMimeType(value: string | undefined): string | undefined {
|
||||
if (!value) return undefined;
|
||||
const [raw] = value.split(";");
|
||||
const normalized = raw?.trim().toLowerCase();
|
||||
return normalized || undefined;
|
||||
}
|
||||
|
||||
export function parseContentType(value: string | undefined): {
|
||||
mimeType?: string;
|
||||
charset?: string;
|
||||
} {
|
||||
if (!value) return {};
|
||||
const parts = value.split(";").map((part) => part.trim());
|
||||
const mimeType = normalizeMimeType(parts[0]);
|
||||
const charset = parts
|
||||
.map((part) => part.match(/^charset=(.+)$/i)?.[1]?.trim())
|
||||
.find((part) => part && part.length > 0);
|
||||
return { mimeType, charset };
|
||||
}
|
||||
|
||||
export function normalizeMimeList(values: string[] | undefined, fallback: string[]): Set<string> {
|
||||
const input = values && values.length > 0 ? values : fallback;
|
||||
return new Set(input.map((value) => normalizeMimeType(value)).filter(Boolean) as string[]);
|
||||
}
|
||||
|
||||
export async function fetchWithGuard(params: {
|
||||
url: string;
|
||||
maxBytes: number;
|
||||
timeoutMs: number;
|
||||
maxRedirects: number;
|
||||
}): Promise<InputFetchResult> {
|
||||
let currentUrl = params.url;
|
||||
let redirectCount = 0;
|
||||
|
||||
const controller = new AbortController();
|
||||
const timeoutId = setTimeout(() => controller.abort(), params.timeoutMs);
|
||||
|
||||
try {
|
||||
while (true) {
|
||||
const parsedUrl = new URL(currentUrl);
|
||||
if (!["http:", "https:"].includes(parsedUrl.protocol)) {
|
||||
throw new Error(`Invalid URL protocol: ${parsedUrl.protocol}. Only HTTP/HTTPS allowed.`);
|
||||
}
|
||||
await assertPublicHostname(parsedUrl.hostname);
|
||||
|
||||
const response = await fetch(parsedUrl, {
|
||||
signal: controller.signal,
|
||||
headers: { "User-Agent": "Clawdbot-Gateway/1.0" },
|
||||
redirect: "manual",
|
||||
});
|
||||
|
||||
if (isRedirectStatus(response.status)) {
|
||||
const location = response.headers.get("location");
|
||||
if (!location) {
|
||||
throw new Error(`Redirect missing location header (${response.status})`);
|
||||
}
|
||||
redirectCount += 1;
|
||||
if (redirectCount > params.maxRedirects) {
|
||||
throw new Error(`Too many redirects (limit: ${params.maxRedirects})`);
|
||||
}
|
||||
currentUrl = new URL(location, parsedUrl).toString();
|
||||
continue;
|
||||
}
|
||||
|
||||
if (!response.ok) {
|
||||
throw new Error(`Failed to fetch: ${response.status} ${response.statusText}`);
|
||||
}
|
||||
|
||||
const contentLength = response.headers.get("content-length");
|
||||
if (contentLength) {
|
||||
const size = parseInt(contentLength, 10);
|
||||
if (size > params.maxBytes) {
|
||||
throw new Error(`Content too large: ${size} bytes (limit: ${params.maxBytes} bytes)`);
|
||||
}
|
||||
}
|
||||
|
||||
const buffer = Buffer.from(await response.arrayBuffer());
|
||||
if (buffer.byteLength > params.maxBytes) {
|
||||
throw new Error(
|
||||
`Content too large: ${buffer.byteLength} bytes (limit: ${params.maxBytes} bytes)`,
|
||||
);
|
||||
}
|
||||
|
||||
const contentType = response.headers.get("content-type") || undefined;
|
||||
const parsed = parseContentType(contentType);
|
||||
const mimeType = parsed.mimeType ?? "application/octet-stream";
|
||||
return { buffer, mimeType, contentType };
|
||||
}
|
||||
} finally {
|
||||
clearTimeout(timeoutId);
|
||||
}
|
||||
}
|
||||
|
||||
function decodeTextContent(buffer: Buffer, charset: string | undefined): string {
|
||||
const encoding = charset?.trim().toLowerCase() || "utf-8";
|
||||
try {
|
||||
return new TextDecoder(encoding).decode(buffer);
|
||||
} catch {
|
||||
return new TextDecoder("utf-8").decode(buffer);
|
||||
}
|
||||
}
|
||||
|
||||
function clampText(text: string, maxChars: number): string {
|
||||
if (text.length <= maxChars) return text;
|
||||
return text.slice(0, maxChars);
|
||||
}
|
||||
|
||||
async function extractPdfContent(params: {
|
||||
buffer: Buffer;
|
||||
limits: InputFileLimits;
|
||||
}): Promise<{ text: string; images: InputImageContent[] }> {
|
||||
const { buffer, limits } = params;
|
||||
const pdf = await getDocument({
|
||||
data: new Uint8Array(buffer),
|
||||
// @ts-expect-error pdfjs-dist legacy option not in current type defs.
|
||||
disableWorker: true,
|
||||
}).promise;
|
||||
const maxPages = Math.min(pdf.numPages, limits.pdf.maxPages);
|
||||
const textParts: string[] = [];
|
||||
|
||||
for (let pageNum = 1; pageNum <= maxPages; pageNum += 1) {
|
||||
const page = await pdf.getPage(pageNum);
|
||||
const textContent = await page.getTextContent();
|
||||
const pageText = textContent.items
|
||||
.map((item) => ("str" in item ? String(item.str) : ""))
|
||||
.filter(Boolean)
|
||||
.join(" ");
|
||||
if (pageText) textParts.push(pageText);
|
||||
}
|
||||
|
||||
const text = textParts.join("\n\n");
|
||||
if (text.trim().length >= limits.pdf.minTextChars) {
|
||||
return { text, images: [] };
|
||||
}
|
||||
|
||||
const images: InputImageContent[] = [];
|
||||
for (let pageNum = 1; pageNum <= maxPages; pageNum += 1) {
|
||||
const page = await pdf.getPage(pageNum);
|
||||
const viewport = page.getViewport({ scale: 1 });
|
||||
const maxPixels = limits.pdf.maxPixels;
|
||||
const pixelBudget = Math.max(1, maxPixels);
|
||||
const pagePixels = viewport.width * viewport.height;
|
||||
const scale = Math.min(1, Math.sqrt(pixelBudget / pagePixels));
|
||||
const scaled = page.getViewport({ scale: Math.max(0.1, scale) });
|
||||
const canvas = createCanvas(Math.ceil(scaled.width), Math.ceil(scaled.height));
|
||||
await page.render({
|
||||
canvas: canvas as unknown as HTMLCanvasElement,
|
||||
viewport: scaled,
|
||||
}).promise;
|
||||
const png = canvas.toBuffer("image/png");
|
||||
images.push({ type: "image", data: png.toString("base64"), mimeType: "image/png" });
|
||||
}
|
||||
|
||||
return { text, images };
|
||||
}
|
||||
|
||||
export async function extractImageContentFromSource(
|
||||
source: InputImageSource,
|
||||
limits: InputImageLimits,
|
||||
): Promise<InputImageContent> {
|
||||
if (source.type === "base64") {
|
||||
if (!source.data) {
|
||||
throw new Error("input_image base64 source missing 'data' field");
|
||||
}
|
||||
const mimeType = normalizeMimeType(source.mediaType) ?? "image/png";
|
||||
if (!limits.allowedMimes.has(mimeType)) {
|
||||
throw new Error(`Unsupported image MIME type: ${mimeType}`);
|
||||
}
|
||||
const buffer = Buffer.from(source.data, "base64");
|
||||
if (buffer.byteLength > limits.maxBytes) {
|
||||
throw new Error(
|
||||
`Image too large: ${buffer.byteLength} bytes (limit: ${limits.maxBytes} bytes)`,
|
||||
);
|
||||
}
|
||||
return { type: "image", data: source.data, mimeType };
|
||||
}
|
||||
|
||||
if (source.type === "url" && source.url) {
|
||||
if (!limits.allowUrl) {
|
||||
throw new Error("input_image URL sources are disabled by config");
|
||||
}
|
||||
const result = await fetchWithGuard({
|
||||
url: source.url,
|
||||
maxBytes: limits.maxBytes,
|
||||
timeoutMs: limits.timeoutMs,
|
||||
maxRedirects: limits.maxRedirects,
|
||||
});
|
||||
if (!limits.allowedMimes.has(result.mimeType)) {
|
||||
throw new Error(`Unsupported image MIME type from URL: ${result.mimeType}`);
|
||||
}
|
||||
return { type: "image", data: result.buffer.toString("base64"), mimeType: result.mimeType };
|
||||
}
|
||||
|
||||
throw new Error("input_image must have 'source.url' or 'source.data'");
|
||||
}
|
||||
|
||||
export async function extractFileContentFromSource(params: {
|
||||
source: InputFileSource;
|
||||
limits: InputFileLimits;
|
||||
}): Promise<InputFileExtractResult> {
|
||||
const { source, limits } = params;
|
||||
const filename = source.filename || "file";
|
||||
|
||||
let buffer: Buffer;
|
||||
let mimeType: string | undefined;
|
||||
let charset: string | undefined;
|
||||
|
||||
if (source.type === "base64") {
|
||||
if (!source.data) {
|
||||
throw new Error("input_file base64 source missing 'data' field");
|
||||
}
|
||||
const parsed = parseContentType(source.mediaType);
|
||||
mimeType = parsed.mimeType;
|
||||
charset = parsed.charset;
|
||||
buffer = Buffer.from(source.data, "base64");
|
||||
} else if (source.type === "url" && source.url) {
|
||||
if (!limits.allowUrl) {
|
||||
throw new Error("input_file URL sources are disabled by config");
|
||||
}
|
||||
const result = await fetchWithGuard({
|
||||
url: source.url,
|
||||
maxBytes: limits.maxBytes,
|
||||
timeoutMs: limits.timeoutMs,
|
||||
maxRedirects: limits.maxRedirects,
|
||||
});
|
||||
const parsed = parseContentType(result.contentType);
|
||||
mimeType = parsed.mimeType ?? normalizeMimeType(result.mimeType);
|
||||
charset = parsed.charset;
|
||||
buffer = result.buffer;
|
||||
} else {
|
||||
throw new Error("input_file must have 'source.url' or 'source.data'");
|
||||
}
|
||||
|
||||
if (buffer.byteLength > limits.maxBytes) {
|
||||
throw new Error(`File too large: ${buffer.byteLength} bytes (limit: ${limits.maxBytes} bytes)`);
|
||||
}
|
||||
|
||||
if (!mimeType) {
|
||||
throw new Error("input_file missing media type");
|
||||
}
|
||||
if (!limits.allowedMimes.has(mimeType)) {
|
||||
throw new Error(`Unsupported file MIME type: ${mimeType}`);
|
||||
}
|
||||
|
||||
if (mimeType === "application/pdf") {
|
||||
const extracted = await extractPdfContent({ buffer, limits });
|
||||
const text = extracted.text ? clampText(extracted.text, limits.maxChars) : "";
|
||||
return {
|
||||
filename,
|
||||
text,
|
||||
images: extracted.images.length > 0 ? extracted.images : undefined,
|
||||
};
|
||||
}
|
||||
|
||||
const text = clampText(decodeTextContent(buffer, charset), limits.maxChars);
|
||||
return { filename, text };
|
||||
}
|
||||
Reference in New Issue
Block a user