@alisio/plugin-semantic 0.0.0-stage → 0.1.1

package/dist/engine/tar.js

@@ -0,0 +1,198 @@
+// Streaming, allow-list tar.gz extraction for the pinned ONNX Runtime npm tarballs (owner O9).
+// Rejects absolute paths, `..`, backslashes, links and device entries anywhere in the archive, caps
+// entry count and sizes, and writes only the files `select` keeps.
+import { createHash } from "node:crypto";
+import { createReadStream } from "node:fs";
+import { mkdir, open } from "node:fs/promises";
+import { dirname, join } from "node:path";
+import { createGunzip } from "node:zlib";
+import { isAbort, SemanticPluginError } from "../errors.js";
+const corrupt = () => {
+    throw new SemanticPluginError("model_corrupt");
+};
+/** Archive path checks applied to every entry, selected or not. */
+export function safeEntryPath(name) {
+    const path = name.replace(/\/+$/, "");
+    if (!path || path.includes("\0") || path.includes("\\") || path.startsWith("/"))
+        corrupt();
+    if (/^[A-Za-z]:/.test(path))
+        corrupt();
+    for (const part of path.split("/"))
+        if (part === "" || part === "." || part === "..")
+            corrupt();
+    return path;
+}
+const field = (block, start, length) => {
+    const raw = block.subarray(start, start + length);
+    const end = raw.indexOf(0);
+    return raw.subarray(0, end === -1 ? length : end).toString("utf8");
+};
+const octal = (block, start, length) => {
+    const text = field(block, start, length).trim();
+    if (!/^[0-7]*$/.test(text))
+        corrupt();
+    return text ? Number.parseInt(text, 8) : 0;
+};
+function checksumOk(block) {
+    let sum = 0;
+    for (let i = 0; i < 512; i++)
+        sum += i >= 148 && i < 156 ? 32 : block[i];
+    return sum === octal(block, 148, 8);
+}
+function paxPath(data) {
+    let at = 0;
+    let path;
+    while (at < data.length) {
+        const space = data.indexOf(0x20, at);
+        if (space === -1)
+            corrupt();
+        const length = Number(data.toString("utf8", at, space));
+        if (!Number.isInteger(length) || length <= 0 || at + length > data.length)
+            corrupt();
+        const record = data.toString("utf8", space + 1, at + length - 1);
+        const eq = record.indexOf("=");
+        if (record.slice(0, eq) === "path")
+            path = record.slice(eq + 1);
+        at += length;
+    }
+    return path;
+}
+/**
+ * Extracts `archive` (gzip + ustar/pax) into `dest`. `select` maps an archive path to the path to
+ * write under `dest`, or null to skip. Returns size and SHA-256 of every written file.
+ */
+export async function extractTarGz(archive, dest, select, limits) {
+    const maxEntries = limits.maxEntries ?? 5000;
+    const maxFileBytes = limits.maxFileBytes ?? 256 * 1024 * 1024;
+    const maxTotalBytes = limits.maxTotalBytes ?? 400 * 1024 * 1024;
+    const out = {};
+    let buffer = Buffer.alloc(0);
+    let entries = 0;
+    let total = 0;
+    let ended = false;
+    let nextPath;
+    // Current entry state.
+    let remaining = 0;
+    let padding = 0;
+    let meta;
+    let metaKind;
+    let target;
+    const finishEntry = async () => {
+        if (target) {
+            await target.handle.close();
+            out[target.path] = { bytes: target.bytes, sha256: target.hash.digest("hex") };
+            target = undefined;
+        }
+        if (meta) {
+            const data = Buffer.concat(meta);
+            nextPath = metaKind === "x" ? (paxPath(data) ?? nextPath) : field(data, 0, data.length);
+            meta = undefined;
+            metaKind = undefined;
+        }
+    };
+    const startEntry = async (block) => {
+        if (block.every((b) => b === 0)) {
+            ended = true;
+            return;
+        }
+        if (!checksumOk(block))
+            corrupt();
+        if (++entries > maxEntries)
+            corrupt();
+        const type = String.fromCharCode(block[156]) || "0";
+        const size = octal(block, 124, 12);
+        const prefix = field(block, 345, 155);
+        const headerName = prefix ? `${prefix}/${field(block, 0, 100)}` : field(block, 0, 100);
+        remaining = size;
+        padding = (512 - (size % 512)) % 512;
+        if (type === "x" || type === "L") {
+            if (size > 64 * 1024)
+                corrupt();
+            meta = [];
+            metaKind = type;
+            return;
+        }
+        if (type === "g")
+            return;
+        const name = nextPath ?? headerName;
+        nextPath = undefined;
+        const path = safeEntryPath(name);
+        if (type === "5")
+            return;
+        if (type !== "0" && type !== "\0" && type !== "7")
+            corrupt(); // links, devices, fifos
+        if (size > maxFileBytes)
+            corrupt();
+        const keep = select(path);
+        if (keep === null)
+            return;
+        const relative = safeEntryPath(keep);
+        total += size;
+        if (total > maxTotalBytes)
+            corrupt();
+        const file = join(dest, relative);
+        await mkdir(dirname(file), { recursive: true, mode: 0o700 });
+        target = {
+            handle: await open(file, "wx", 0o600),
+            hash: createHash("sha256"),
+            path: relative,
+            bytes: 0,
+        };
+    };
+    try {
+        limits.signal.throwIfAborted();
+        const source = createReadStream(archive).pipe(createGunzip());
+        for await (const chunk of source) {
+            limits.signal.throwIfAborted();
+            buffer = buffer.length ? Buffer.concat([buffer, chunk]) : chunk;
+            while (!ended) {
+                if (remaining > 0) {
+                    if (buffer.length === 0)
+                        break;
+                    const piece = buffer.subarray(0, Math.min(remaining, buffer.length));
+                    buffer = buffer.subarray(piece.length);
+                    remaining -= piece.length;
+                    if (meta)
+                        meta.push(Buffer.from(piece));
+                    else if (target) {
+                        target.hash.update(piece);
+                        target.bytes += piece.length;
+                        await target.handle.write(piece);
+                    }
+                    if (remaining > 0)
+                        break;
+                    await finishEntry();
+                    continue;
+                }
+                if (padding > 0) {
+                    if (buffer.length < padding)
+                        break;
+                    buffer = buffer.subarray(padding);
+                    padding = 0;
+                    continue;
+                }
+                if (buffer.length < 512)
+                    break;
+                const block = buffer.subarray(0, 512);
+                buffer = buffer.subarray(512);
+                await startEntry(block);
+                if (remaining === 0 && !ended)
+                    await finishEntry();
+            }
+            if (ended)
+                break;
+        }
+        if (!ended || remaining > 0)
+            corrupt();
+        return out;
+    }
+    catch (error) {
+        await target?.handle.close().catch(() => { });
+        if (isAbort(error) || limits.signal.aborted)
+            throw new SemanticPluginError("cancelled");
+        if (error instanceof SemanticPluginError)
+            throw error;
+        throw new SemanticPluginError("model_corrupt");
+    }
+}
+//# sourceMappingURL=tar.js.map

package/dist/engine/tar.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"tar.js","sourceRoot":"","sources":["../../src/engine/tar.ts"],"names":[],"mappings":"AAAA,+FAA+F;AAC/F,oGAAoG;AACpG,mEAAmE;AACnE,OAAO,EAAE,UAAU,EAAa,MAAM,aAAa,CAAC;AACpD,OAAO,EAAE,gBAAgB,EAAE,MAAM,SAAS,CAAC;AAC3C,OAAO,EAAmB,KAAK,EAAE,IAAI,EAAE,MAAM,kBAAkB,CAAC;AAChE,OAAO,EAAE,OAAO,EAAE,IAAI,EAAE,MAAM,WAAW,CAAC;AAC1C,OAAO,EAAE,YAAY,EAAE,MAAM,WAAW,CAAC;AACzC,OAAO,EAAE,OAAO,EAAE,mBAAmB,EAAE,MAAM,cAAc,CAAC;AAU5D,MAAM,OAAO,GAAG,GAAU,EAAE;IAC1B,MAAM,IAAI,mBAAmB,CAAC,eAAe,CAAC,CAAC;AACjD,CAAC,CAAC;AAEF,mEAAmE;AACnE,MAAM,UAAU,aAAa,CAAC,IAAY;IACxC,MAAM,IAAI,GAAG,IAAI,CAAC,OAAO,CAAC,MAAM,EAAE,EAAE,CAAC,CAAC;IACtC,IAAI,CAAC,IAAI,IAAI,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC,IAAI,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC,IAAI,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC;QAAE,OAAO,EAAE,CAAC;IAC3F,IAAI,YAAY,CAAC,IAAI,CAAC,IAAI,CAAC;QAAE,OAAO,EAAE,CAAC;IACvC,KAAK,MAAM,IAAI,IAAI,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC;QAAE,IAAI,IAAI,KAAK,EAAE,IAAI,IAAI,KAAK,GAAG,IAAI,IAAI,KAAK,IAAI;YAAE,OAAO,EAAE,CAAC;IAChG,OAAO,IAAI,CAAC;AACd,CAAC;AAED,MAAM,KAAK,GAAG,CAAC,KAAa,EAAE,KAAa,EAAE,MAAc,EAAE,EAAE;IAC7D,MAAM,GAAG,GAAG,KAAK,CAAC,QAAQ,CAAC,KAAK,EAAE,KAAK,GAAG,MAAM,CAAC,CAAC;IAClD,MAAM,GAAG,GAAG,GAAG,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC;IAC3B,OAAO,GAAG,CAAC,QAAQ,CAAC,CAAC,EAAE,GAAG,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,QAAQ,CAAC,MAAM,CAAC,CAAC;AACrE,CAAC,CAAC;AACF,MAAM,KAAK,GAAG,CAAC,KAAa,EAAE,KAAa,EAAE,MAAc,EAAE,EAAE;IAC7D,MAAM,IAAI,GAAG,KAAK,CAAC,KAAK,EAAE,KAAK,EAAE,MAAM,CAAC,CAAC,IAAI,EAAE,CAAC;IAChD,IAAI,CAAC,UAAU,CAAC,IAAI,CAAC,IAAI,CAAC;QAAE,OAAO,EAAE,CAAC;IACtC,OAAO,IAAI,CAAC,CAAC,CAAC,MAAM,CAAC,QAAQ,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;AAC7C,CAAC,CAAC;AAEF,SAAS,UAAU,CAAC,KAAa;IAC/B,IAAI,GAAG,GAAG,CAAC,CAAC;IACZ,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,GAAG,EAAE,CAAC,EAAE;QAAE,GAAG,IAAI,CAAC,IAAI,GAAG,IAAI,CAAC,GAAG,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAE,KAAK,CAAC,CAAC,CAAY,CAAC;IACrF,OAAO,GAAG,KAAK,KAAK,CAAC,KAAK,EAAE,GAAG,EAAE,CAAC,CAAC,CAAC;AACtC,CAAC;AAED,SAAS,OAAO,CAAC,IAAY;IAC3B,IAAI,EAAE,GAAG,CAAC,CAAC;IACX,IAAI,IAAwB,CAAC;IAC7B,OAAO,EAAE,GAAG,IAAI,CAAC,MAAM,EAAE,CAAC;QACxB,MAAM,KAAK,GAAG,IAAI,CAAC,OAAO,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC;QACrC,IAAI,KAAK,KAAK,CAAC,CAAC;YAAE,OAAO,EAAE,CAAC;QAC5B,MAAM,MAAM,GAAG,MAAM,CAAC,IAAI,CAAC,QAAQ,CAAC,MAAM,EAAE,EAAE,EAAE,KAAK,CAAC,CAAC,CAAC;QACxD,IAAI,CAAC,MAAM,CAAC,SAAS,CAAC,MAAM,CAAC,IAAI,MAAM,IAAI,CAAC,IAAI,EAAE,GAAG,MAAM,GAAG,IAAI,CAAC,MAAM;YAAE,OAAO,EAAE,CAAC;QACrF,MAAM,MAAM,GAAG,IAAI,CAAC,QAAQ,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC,EAAE,EAAE,GAAG,MAAM,GAAG,CAAC,CAAC,CAAC;QACjE,MAAM,EAAE,GAAG,MAAM,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC;QAC/B,IAAI,MAAM,CAAC,KAAK,CAAC,CAAC,EAAE,EAAE,CAAC,KAAK,MAAM;YAAE,IAAI,GAAG,MAAM,CAAC,KAAK,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;QAChE,EAAE,IAAI,MAAM,CAAC;IACf,CAAC;IACD,OAAO,IAAI,CAAC;AACd,CAAC;AAED;;;GAGG;AACH,MAAM,CAAC,KAAK,UAAU,YAAY,CAChC,OAAe,EACf,IAAY,EACZ,MAAuC,EACvC,MAAqB;IAErB,MAAM,UAAU,GAAG,MAAM,CAAC,UAAU,IAAI,IAAI,CAAC;IAC7C,MAAM,YAAY,GAAG,MAAM,CAAC,YAAY,IAAI,GAAG,GAAG,IAAI,GAAG,IAAI,CAAC;IAC9D,MAAM,aAAa,GAAG,MAAM,CAAC,aAAa,IAAI,GAAG,GAAG,IAAI,GAAG,IAAI,CAAC;IAChE,MAAM,GAAG,GAAc,EAAE,CAAC;IAC1B,IAAI,MAAM,GAAW,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;IACrC,IAAI,OAAO,GAAG,CAAC,CAAC;IAChB,IAAI,KAAK,GAAG,CAAC,CAAC;IACd,IAAI,KAAK,GAAG,KAAK,CAAC;IAClB,IAAI,QAA4B,CAAC;IACjC,uBAAuB;IACvB,IAAI,SAAS,GAAG,CAAC,CAAC;IAClB,IAAI,OAAO,GAAG,CAAC,CAAC;IAChB,IAAI,IAA0B,CAAC;IAC/B,IAAI,QAA+B,CAAC;IACpC,IAAI,MAAmF,CAAC;IAExF,MAAM,WAAW,GAAG,KAAK,IAAI,EAAE;QAC7B,IAAI,MAAM,EAAE,CAAC;YACX,MAAM,MAAM,CAAC,MAAM,CAAC,KAAK,EAAE,CAAC;YAC5B,GAAG,CAAC,MAAM,CAAC,IAAI,CAAC,GAAG,EAAE,KAAK,EAAE,MAAM,CAAC,KAAK,EAAE,MAAM,EAAE,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,CAAC;YAC9E,MAAM,GAAG,SAAS,CAAC;QACrB,CAAC;QACD,IAAI,IAAI,EAAE,CAAC;YACT,MAAM,IAAI,GAAG,MAAM,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC;YACjC,QAAQ,GAAG,QAAQ,KAAK,GAAG,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,IAAI,CAAC,IAAI,QAAQ,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,IAAI,EAAE,CAAC,EAAE,IAAI,CAAC,MAAM,CAAC,CAAC;YACxF,IAAI,GAAG,SAAS,CAAC;YACjB,QAAQ,GAAG,SAAS,CAAC;QACvB,CAAC;IACH,CAAC,CAAC;IAEF,MAAM,UAAU,GAAG,KAAK,EAAE,KAAa,EAAE,EAAE;QACzC,IAAI,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC;YAChC,KAAK,GAAG,IAAI,CAAC;YACb,OAAO;QACT,CAAC;QACD,IAAI,CAAC,UAAU,CAAC,KAAK,CAAC;YAAE,OAAO,EAAE,CAAC;QAClC,IAAI,EAAE,OAAO,GAAG,UAAU;YAAE,OAAO,EAAE,CAAC;QACtC,MAAM,IAAI,GAAG,MAAM,CAAC,YAAY,CAAC,KAAK,CAAC,GAAG,CAAW,CAAC,IAAI,GAAG,CAAC;QAC9D,MAAM,IAAI,GAAG,KAAK,CAAC,KAAK,EAAE,GAAG,EAAE,EAAE,CAAC,CAAC;QACnC,MAAM,MAAM,GAAG,KAAK,CAAC,KAAK,EAAE,GAAG,EAAE,GAAG,CAAC,CAAC;QACtC,MAAM,UAAU,GAAG,MAAM,CAAC,CAAC,CAAC,GAAG,MAAM,IAAI,KAAK,CAAC,KAAK,EAAE,CAAC,EAAE,GAAG,CAAC,EAAE,CAAC,CAAC,CAAC,KAAK,CAAC,KAAK,EAAE,CAAC,EAAE,GAAG,CAAC,CAAC;QACvF,SAAS,GAAG,IAAI,CAAC;QACjB,OAAO,GAAG,CAAC,GAAG,GAAG,CAAC,IAAI,GAAG,GAAG,CAAC,CAAC,GAAG,GAAG,CAAC;QACrC,IAAI,IAAI,KAAK,GAAG,IAAI,IAAI,KAAK,GAAG,EAAE,CAAC;YACjC,IAAI,IAAI,GAAG,EAAE,GAAG,IAAI;gBAAE,OAAO,EAAE,CAAC;YAChC,IAAI,GAAG,EAAE,CAAC;YACV,QAAQ,GAAG,IAAI,CAAC;YAChB,OAAO;QACT,CAAC;QACD,IAAI,IAAI,KAAK,GAAG;YAAE,OAAO;QACzB,MAAM,IAAI,GAAG,QAAQ,IAAI,UAAU,CAAC;QACpC,QAAQ,GAAG,SAAS,CAAC;QACrB,MAAM,IAAI,GAAG,aAAa,CAAC,IAAI,CAAC,CAAC;QACjC,IAAI,IAAI,KAAK,GAAG;YAAE,OAAO;QACzB,IAAI,IAAI,KAAK,GAAG,IAAI,IAAI,KAAK,IAAI,IAAI,IAAI,KAAK,GAAG;YAAE,OAAO,EAAE,CAAC,CAAC,wBAAwB;QACtF,IAAI,IAAI,GAAG,YAAY;YAAE,OAAO,EAAE,CAAC;QACnC,MAAM,IAAI,GAAG,MAAM,CAAC,IAAI,CAAC,CAAC;QAC1B,IAAI,IAAI,KAAK,IAAI;YAAE,OAAO;QAC1B,MAAM,QAAQ,GAAG,aAAa,CAAC,IAAI,CAAC,CAAC;QACrC,KAAK,IAAI,IAAI,CAAC;QACd,IAAI,KAAK,GAAG,aAAa;YAAE,OAAO,EAAE,CAAC;QACrC,MAAM,IAAI,GAAG,IAAI,CAAC,IAAI,EAAE,QAAQ,CAAC,CAAC;QAClC,MAAM,KAAK,CAAC,OAAO,CAAC,IAAI,CAAC,EAAE,EAAE,SAAS,EAAE,IAAI,EAAE,IAAI,EAAE,KAAK,EAAE,CAAC,CAAC;QAC7D,MAAM,GAAG;YACP,MAAM,EAAE,MAAM,IAAI,CAAC,IAAI,EAAE,IAAI,EAAE,KAAK,CAAC;YACrC,IAAI,EAAE,UAAU,CAAC,QAAQ,CAAC;YAC1B,IAAI,EAAE,QAAQ;YACd,KAAK,EAAE,CAAC;SACT,CAAC;IACJ,CAAC,CAAC;IAEF,IAAI,CAAC;QACH,MAAM,CAAC,MAAM,CAAC,cAAc,EAAE,CAAC;QAC/B,MAAM,MAAM,GAAG,gBAAgB,CAAC,OAAO,CAAC,CAAC,IAAI,CAAC,YAAY,EAAE,CAAC,CAAC;QAC9D,IAAI,KAAK,EAAE,MAAM,KAAK,IAAI,MAA+B,EAAE,CAAC;YAC1D,MAAM,CAAC,MAAM,CAAC,cAAc,EAAE,CAAC;YAC/B,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC,CAAC,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC,MAAM,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC;YAChE,OAAO,CAAC,KAAK,EAAE,CAAC;gBACd,IAAI,SAAS,GAAG,CAAC,EAAE,CAAC;oBAClB,IAAI,MAAM,CAAC,MAAM,KAAK,CAAC;wBAAE,MAAM;oBAC/B,MAAM,KAAK,GAAG,MAAM,CAAC,QAAQ,CAAC,CAAC,EAAE,IAAI,CAAC,GAAG,CAAC,SAAS,EAAE,MAAM,CAAC,MAAM,CAAC,CAAC,CAAC;oBACrE,MAAM,GAAG,MAAM,CAAC,QAAQ,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC;oBACvC,SAAS,IAAI,KAAK,CAAC,MAAM,CAAC;oBAC1B,IAAI,IAAI;wBAAE,IAAI,CAAC,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC;yBACnC,IAAI,MAAM,EAAE,CAAC;wBAChB,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC;wBAC1B,MAAM,CAAC,KAAK,IAAI,KAAK,CAAC,MAAM,CAAC;wBAC7B,MAAM,MAAM,CAAC,MAAM,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC;oBACnC,CAAC;oBACD,IAAI,SAAS,GAAG,CAAC;wBAAE,MAAM;oBACzB,MAAM,WAAW,EAAE,CAAC;oBACpB,SAAS;gBACX,CAAC;gBACD,IAAI,OAAO,GAAG,CAAC,EAAE,CAAC;oBAChB,IAAI,MAAM,CAAC,MAAM,GAAG,OAAO;wBAAE,MAAM;oBACnC,MAAM,GAAG,MAAM,CAAC,QAAQ,CAAC,OAAO,CAAC,CAAC;oBAClC,OAAO,GAAG,CAAC,CAAC;oBACZ,SAAS;gBACX,CAAC;gBACD,IAAI,MAAM,CAAC,MAAM,GAAG,GAAG;oBAAE,MAAM;gBAC/B,MAAM,KAAK,GAAG,MAAM,CAAC,QAAQ,CAAC,CAAC,EAAE,GAAG,CAAC,CAAC;gBACtC,MAAM,GAAG,MAAM,CAAC,QAAQ,CAAC,GAAG,CAAC,CAAC;gBAC9B,MAAM,UAAU,CAAC,KAAK,CAAC,CAAC;gBACxB,IAAI,SAAS,KAAK,CAAC,IAAI,CAAC,KAAK;oBAAE,MAAM,WAAW,EAAE,CAAC;YACrD,CAAC;YACD,IAAI,KAAK;gBAAE,MAAM;QACnB,CAAC;QACD,IAAI,CAAC,KAAK,IAAI,SAAS,GAAG,CAAC;YAAE,OAAO,EAAE,CAAC;QACvC,OAAO,GAAG,CAAC;IACb,CAAC;IAAC,OAAO,KAAK,EAAE,CAAC;QACf,MAAM,MAAM,EAAE,MAAM,CAAC,KAAK,EAAE,CAAC,KAAK,CAAC,GAAG,EAAE,GAAE,CAAC,CAAC,CAAC;QAC7C,IAAI,OAAO,CAAC,KAAK,CAAC,IAAI,MAAM,CAAC,MAAM,CAAC,OAAO;YAAE,MAAM,IAAI,mBAAmB,CAAC,WAAW,CAAC,CAAC;QACxF,IAAI,KAAK,YAAY,mBAAmB;YAAE,MAAM,KAAK,CAAC;QACtD,MAAM,IAAI,mBAAmB,CAAC,eAAe,CAAC,CAAC;IACjD,CAAC;AACH,CAAC"}

package/dist/engine/worker.d.ts

@@ -0,0 +1,2 @@
+export {};
+//# sourceMappingURL=worker.d.ts.map

package/dist/engine/worker.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"worker.d.ts","sourceRoot":"","sources":["../../src/engine/worker.ts"],"names":[],"mappings":""}

package/dist/engine/worker.js

@@ -0,0 +1,64 @@
+// Worker thread entry (D5): loads the cached ONNX Runtime (downloaded on first use, owner O9), the
+// pinned ONNX model (CPU) and the tokenizer (@huggingface/tokenizers), then embeds batches posted by
+// `embedder.ts`.
+import { readFile } from "node:fs/promises";
+import { availableParallelism } from "node:os";
+import { join } from "node:path";
+import { parentPort, workerData } from "node:worker_threads";
+import { batchInputs, encodeBatch, meanPoolNormalize } from "./infer.js";
+import { loadRuntime } from "./runtime.js";
+const { modelDir, runtimeDir, maxTokens } = workerData;
+const port = parentPort;
+async function start() {
+    if (!port)
+        return;
+    try {
+        const ort = loadRuntime(runtimeDir);
+        const { Tokenizer } = await import("@huggingface/tokenizers");
+        const readJson = async (name) => JSON.parse(await readFile(join(modelDir, name), "utf8"));
+        const [tokenizerJson, tokenizerConfig, config] = await Promise.all([
+            readJson("tokenizer.json"),
+            readJson("tokenizer_config.json"),
+            readJson("config.json"),
+        ]);
+        const tokenizer = new Tokenizer(tokenizerJson, tokenizerConfig);
+        const padToken = typeof tokenizerConfig.pad_token === "string" ? tokenizerConfig.pad_token : "<pad>";
+        const padId = tokenizer.token_to_id(padToken) ?? (Number(config.pad_token_id) || 0);
+        const session = await ort.InferenceSession.create(join(modelDir, "onnx/model_quantized.onnx"), {
+            executionProviders: ["cpu"],
+            intraOpNumThreads: Math.max(1, Math.min(4, availableParallelism() - 1)),
+            interOpNumThreads: 1,
+        });
+        const output = session.outputNames[0];
+        port.on("message", async (message) => {
+            if (message?.type !== "embed")
+                return;
+            try {
+                const encoded = encodeBatch(tokenizer, message.texts, maxTokens);
+                const input = batchInputs(encoded, padId);
+                const shape = [input.batch, input.length];
+                const feeds = {
+                    input_ids: new ort.Tensor("int64", input.inputIds, shape),
+                    attention_mask: new ort.Tensor("int64", input.attention, shape),
+                };
+                if (session.inputNames.includes("token_type_ids"))
+                    feeds.token_type_ids = new ort.Tensor("int64", input.tokenTypes, shape);
+                const result = await session.run(feeds);
+                const hidden = result[output];
+                if (!hidden)
+                    throw new Error("no output");
+                const vectors = meanPoolNormalize(hidden.data, hidden.dims, input.lengths);
+                port.postMessage({ type: "vectors", id: message.id, vectors }, vectors.map((v) => v.buffer));
+            }
+            catch {
+                port.postMessage({ type: "error", id: message.id, code: "inference_failed" });
+            }
+        });
+        port.postMessage({ type: "ready", dimensions: Number(config.hidden_size) });
+    }
+    catch {
+        port.postMessage({ type: "error", code: "load_failed" });
+    }
+}
+void start();
+//# sourceMappingURL=worker.js.map

package/dist/engine/worker.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"worker.js","sourceRoot":"","sources":["../../src/engine/worker.ts"],"names":[],"mappings":"AAAA,mGAAmG;AACnG,qGAAqG;AACrG,iBAAiB;AAEjB,OAAO,EAAE,QAAQ,EAAE,MAAM,kBAAkB,CAAC;AAC5C,OAAO,EAAE,oBAAoB,EAAE,MAAM,SAAS,CAAC;AAC/C,OAAO,EAAE,IAAI,EAAE,MAAM,WAAW,CAAC;AACjC,OAAO,EAAE,UAAU,EAAE,UAAU,EAAE,MAAM,qBAAqB,CAAC;AAC7D,OAAO,EAAE,WAAW,EAAE,WAAW,EAAE,iBAAiB,EAAE,MAAM,YAAY,CAAC;AACzE,OAAO,EAAE,WAAW,EAAE,MAAM,cAAc,CAAC;AAyB3C,MAAM,EAAE,QAAQ,EAAE,UAAU,EAAE,SAAS,EAAE,GAAG,UAI3C,CAAC;AACF,MAAM,IAAI,GAAG,UAAU,CAAC;AAExB,KAAK,UAAU,KAAK;IAClB,IAAI,CAAC,IAAI;QAAE,OAAO;IAClB,IAAI,CAAC;QACH,MAAM,GAAG,GAAG,WAAW,CAAY,UAAU,CAAC,CAAC;QAC/C,MAAM,EAAE,SAAS,EAAE,GAAG,MAAM,MAAM,CAAC,yBAAyB,CAAC,CAAC;QAC9D,MAAM,QAAQ,GAAG,KAAK,EAAE,IAAY,EAAE,EAAE,CACtC,IAAI,CAAC,KAAK,CAAC,MAAM,QAAQ,CAAC,IAAI,CAAC,QAAQ,EAAE,IAAI,CAAC,EAAE,MAAM,CAAC,CAAC,CAAC;QAC3D,MAAM,CAAC,aAAa,EAAE,eAAe,EAAE,MAAM,CAAC,GAAG,MAAM,OAAO,CAAC,GAAG,CAAC;YACjE,QAAQ,CAAC,gBAAgB,CAAC;YAC1B,QAAQ,CAAC,uBAAuB,CAAC;YACjC,QAAQ,CAAC,aAAa,CAAC;SACxB,CAAC,CAAC;QACH,MAAM,SAAS,GAAG,IAAI,SAAS,CAAC,aAAa,EAAE,eAAe,CAAC,CAAC;QAChE,MAAM,QAAQ,GACZ,OAAO,eAAe,CAAC,SAAS,KAAK,QAAQ,CAAC,CAAC,CAAC,eAAe,CAAC,SAAS,CAAC,CAAC,CAAC,OAAO,CAAC;QACtF,MAAM,KAAK,GAAG,SAAS,CAAC,WAAW,CAAC,QAAQ,CAAC,IAAI,CAAC,MAAM,CAAC,MAAM,CAAC,YAAY,CAAC,IAAI,CAAC,CAAC,CAAC;QACpF,MAAM,OAAO,GAAG,MAAM,GAAG,CAAC,gBAAgB,CAAC,MAAM,CAAC,IAAI,CAAC,QAAQ,EAAE,2BAA2B,CAAC,EAAE;YAC7F,kBAAkB,EAAE,CAAC,KAAK,CAAC;YAC3B,iBAAiB,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,oBAAoB,EAAE,GAAG,CAAC,CAAC,CAAC;YACvE,iBAAiB,EAAE,CAAC;SACrB,CAAC,CAAC;QACH,MAAM,MAAM,GAAG,OAAO,CAAC,WAAW,CAAC,CAAC,CAAW,CAAC;QAChD,IAAI,CAAC,EAAE,CAAC,SAAS,EAAE,KAAK,EAAE,OAAsD,EAAE,EAAE;YAClF,IAAI,OAAO,EAAE,IAAI,KAAK,OAAO;gBAAE,OAAO;YACtC,IAAI,CAAC;gBACH,MAAM,OAAO,GAAG,WAAW,CAAC,SAAS,EAAE,OAAO,CAAC,KAAK,EAAE,SAAS,CAAC,CAAC;gBACjE,MAAM,KAAK,GAAG,WAAW,CAAC,OAAO,EAAE,KAAK,CAAC,CAAC;gBAC1C,MAAM,KAAK,GAAG,CAAC,KAAK,CAAC,KAAK,EAAE,KAAK,CAAC,MAAM,CAAC,CAAC;gBAC1C,MAAM,KAAK,GAA8B;oBACvC,SAAS,EAAE,IAAI,GAAG,CAAC,MAAM,CAAC,OAAO,EAAE,KAAK,CAAC,QAAQ,EAAE,KAAK,CAAC;oBACzD,cAAc,EAAE,IAAI,GAAG,CAAC,MAAM,CAAC,OAAO,EAAE,KAAK,CAAC,SAAS,EAAE,KAAK,CAAC;iBAChE,CAAC;gBACF,IAAI,OAAO,CAAC,UAAU,CAAC,QAAQ,CAAC,gBAAgB,CAAC;oBAC/C,KAAK,CAAC,cAAc,GAAG,IAAI,GAAG,CAAC,MAAM,CAAC,OAAO,EAAE,KAAK,CAAC,UAAU,EAAE,KAAK,CAAC,CAAC;gBAC1E,MAAM,MAAM,GAAG,MAAM,OAAO,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;gBACxC,MAAM,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC,CAAC;gBAC9B,IAAI,CAAC,MAAM;oBAAE,MAAM,IAAI,KAAK,CAAC,WAAW,CAAC,CAAC;gBAC1C,MAAM,OAAO,GAAG,iBAAiB,CAAC,MAAM,CAAC,IAAoB,EAAE,MAAM,CAAC,IAAI,EAAE,KAAK,CAAC,OAAO,CAAC,CAAC;gBAC3F,IAAI,CAAC,WAAW,CACd,EAAE,IAAI,EAAE,SAAS,EAAE,EAAE,EAAE,OAAO,CAAC,EAAE,EAAE,OAAO,EAAE,EAC5C,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,MAAqB,CAAC,CAC5C,CAAC;YACJ,CAAC;YAAC,MAAM,CAAC;gBACP,IAAI,CAAC,WAAW,CAAC,EAAE,IAAI,EAAE,OAAO,EAAE,EAAE,EAAE,OAAO,CAAC,EAAE,EAAE,IAAI,EAAE,kBAAkB,EAAE,CAAC,CAAC;YAChF,CAAC;QACH,CAAC,CAAC,CAAC;QACH,IAAI,CAAC,WAAW,CAAC,EAAE,IAAI,EAAE,OAAO,EAAE,UAAU,EAAE,MAAM,CAAC,MAAM,CAAC,WAAW,CAAC,EAAE,CAAC,CAAC;IAC9E,CAAC;IAAC,MAAM,CAAC;QACP,IAAI,CAAC,WAAW,CAAC,EAAE,IAAI,EAAE,OAAO,EAAE,IAAI,EAAE,aAAa,EAAE,CAAC,CAAC;IAC3D,CAAC;AACH,CAAC;AAED,KAAK,KAAK,EAAE,CAAC"}

package/dist/errors.d.ts

@@ -0,0 +1,15 @@
+/** Error codes the plugin returns (§8). Never carry inputs, paths or stacks. */
+export declare const ERROR_CODES: readonly ["invalid_input", "not_found", "limit_exceeded", "unsupported_format", "invalid_encoding", "no_text", "path_outside_workspace", "file_too_large", "zip_invalid", "engine_unavailable", "model_missing", "model_corrupt", "unknown_model", "model_mismatch", "download_failed", "cancelled", "answer_failed", "no_model", "operation_failed"];
+export type SemanticPluginErrorCode = (typeof ERROR_CODES)[number];
+export declare class SemanticPluginError extends Error {
+    readonly code: SemanticPluginErrorCode;
+    constructor(code: SemanticPluginErrorCode, message?: string);
+}
+/** True for an abort raised by an `AbortSignal` (DOMException or a plain AbortError). */
+export declare function isAbort(error: unknown): boolean;
+/**
+ * Maps any error to a plugin code. The host's `SemanticError` is matched by `name`/`code`, not
+ * `instanceof`, because the host and the plugin may load different copies of the SDK.
+ */
+export declare function errorCode(error: unknown): SemanticPluginErrorCode;
+//# sourceMappingURL=errors.d.ts.map

package/dist/errors.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"errors.d.ts","sourceRoot":"","sources":["../src/errors.ts"],"names":[],"mappings":"AAAA,gFAAgF;AAChF,eAAO,MAAM,WAAW,YACtB,eAAe,EACf,WAAW,EACX,gBAAgB,EAChB,oBAAoB,EACpB,kBAAkB,EAClB,SAAS,EACT,wBAAwB,EACxB,gBAAgB,EAChB,aAAa,EACb,oBAAoB,EACpB,eAAe,EACf,eAAe,EACf,eAAe,EACf,gBAAgB,EAChB,iBAAiB,EACjB,WAAW,EACX,eAAe,EACf,UAAU,EACV,kBAAkB,CACV,CAAC;AACX,MAAM,MAAM,uBAAuB,GAAG,CAAC,OAAO,WAAW,CAAC,CAAC,MAAM,CAAC,CAAC;AAEnE,qBAAa,mBAAoB,SAAQ,KAAK;IAC5C,QAAQ,CAAC,IAAI,EAAE,uBAAuB,CAAC;IACvC,YAAY,IAAI,EAAE,uBAAuB,EAAE,OAAO,CAAC,EAAE,MAAM,EAI1D;CACF;AAED,yFAAyF;AACzF,wBAAgB,OAAO,CAAC,KAAK,EAAE,OAAO,GAAG,OAAO,CAO/C;AAED;;;GAGG;AACH,wBAAgB,SAAS,CAAC,KAAK,EAAE,OAAO,GAAG,uBAAuB,CAcjE"}

package/dist/errors.js

@@ -0,0 +1,59 @@
+/** Error codes the plugin returns (§8). Never carry inputs, paths or stacks. */
+export const ERROR_CODES = [
+    "invalid_input",
+    "not_found",
+    "limit_exceeded",
+    "unsupported_format",
+    "invalid_encoding",
+    "no_text",
+    "path_outside_workspace",
+    "file_too_large",
+    "zip_invalid",
+    "engine_unavailable",
+    "model_missing",
+    "model_corrupt",
+    "unknown_model",
+    "model_mismatch",
+    "download_failed",
+    "cancelled",
+    "answer_failed",
+    "no_model",
+    "operation_failed",
+];
+export class SemanticPluginError extends Error {
+    code;
+    constructor(code, message) {
+        super(message ?? code);
+        this.name = "SemanticPluginError";
+        this.code = code;
+    }
+}
+/** True for an abort raised by an `AbortSignal` (DOMException or a plain AbortError). */
+export function isAbort(error) {
+    return (!!error &&
+        typeof error === "object" &&
+        (error.name === "AbortError" ||
+            error.name === "TimeoutError"));
+}
+/**
+ * Maps any error to a plugin code. The host's `SemanticError` is matched by `name`/`code`, not
+ * `instanceof`, because the host and the plugin may load different copies of the SDK.
+ */
+export function errorCode(error) {
+    if (error instanceof SemanticPluginError)
+        return error.code;
+    if (isAbort(error))
+        return "cancelled";
+    if (error &&
+        typeof error === "object" &&
+        error.name === "SemanticError") {
+        const code = error.code;
+        if (code === "model_mismatch" || code === "dimension_mismatch")
+            return "model_mismatch";
+        if (code === "invalid_request")
+            return "invalid_input";
+        return "engine_unavailable";
+    }
+    return "operation_failed";
+}
+//# sourceMappingURL=errors.js.map

package/dist/errors.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"errors.js","sourceRoot":"","sources":["../src/errors.ts"],"names":[],"mappings":"AAAA,gFAAgF;AAChF,MAAM,CAAC,MAAM,WAAW,GAAG;IACzB,eAAe;IACf,WAAW;IACX,gBAAgB;IAChB,oBAAoB;IACpB,kBAAkB;IAClB,SAAS;IACT,wBAAwB;IACxB,gBAAgB;IAChB,aAAa;IACb,oBAAoB;IACpB,eAAe;IACf,eAAe;IACf,eAAe;IACf,gBAAgB;IAChB,iBAAiB;IACjB,WAAW;IACX,eAAe;IACf,UAAU;IACV,kBAAkB;CACV,CAAC;AAGX,MAAM,OAAO,mBAAoB,SAAQ,KAAK;IACnC,IAAI,CAA0B;IACvC,YAAY,IAA6B,EAAE,OAAgB;QACzD,KAAK,CAAC,OAAO,IAAI,IAAI,CAAC,CAAC;QACvB,IAAI,CAAC,IAAI,GAAG,qBAAqB,CAAC;QAClC,IAAI,CAAC,IAAI,GAAG,IAAI,CAAC;IACnB,CAAC;CACF;AAED,yFAAyF;AACzF,MAAM,UAAU,OAAO,CAAC,KAAc;IACpC,OAAO,CACL,CAAC,CAAC,KAAK;QACP,OAAO,KAAK,KAAK,QAAQ;QACzB,CAAE,KAA4B,CAAC,IAAI,KAAK,YAAY;YACjD,KAA4B,CAAC,IAAI,KAAK,cAAc,CAAC,CACzD,CAAC;AACJ,CAAC;AAED;;;GAGG;AACH,MAAM,UAAU,SAAS,CAAC,KAAc;IACtC,IAAI,KAAK,YAAY,mBAAmB;QAAE,OAAO,KAAK,CAAC,IAAI,CAAC;IAC5D,IAAI,OAAO,CAAC,KAAK,CAAC;QAAE,OAAO,WAAW,CAAC;IACvC,IACE,KAAK;QACL,OAAO,KAAK,KAAK,QAAQ;QACxB,KAA4B,CAAC,IAAI,KAAK,eAAe,EACtD,CAAC;QACD,MAAM,IAAI,GAAI,KAA4B,CAAC,IAAI,CAAC;QAChD,IAAI,IAAI,KAAK,gBAAgB,IAAI,IAAI,KAAK,oBAAoB;YAAE,OAAO,gBAAgB,CAAC;QACxF,IAAI,IAAI,KAAK,iBAAiB;YAAE,OAAO,eAAe,CAAC;QACvD,OAAO,oBAAoB,CAAC;IAC9B,CAAC;IACD,OAAO,kBAAkB,CAAC;AAC5B,CAAC"}

package/dist/index.d.ts

@@ -0,0 +1,20 @@
+import { type PluginAPI } from "@alisio/sdk";
+export declare const MIN_HOST = "0.11.7";
+declare const _default: {
+    id: string;
+    name: string;
+    version: string;
+    apiVersion: 1;
+    description: string;
+    categories: ("search" | "storage")[];
+    setup(api: PluginAPI): void;
+    /**
+     * The host gives no activation here, so this must not touch an activation that is still in use
+     * by another workspace (0.1.0 did, which unregistered the live Notebook app: O13). Activations
+     * with an app are released by their app's `dispose`; idle ones without an app (hosts without
+     * `api.apps`) only hold resources that are reopened on demand.
+     */
+    dispose(): void;
+};
+export default _default;
+//# sourceMappingURL=index.d.ts.map

package/dist/index.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAEA,OAAO,EAAgB,KAAK,SAAS,EAAc,MAAM,aAAa,CAAC;AAavE,eAAO,MAAM,QAAQ,WAAW,CAAC;;;;;;;;;IA6Q/B;;;;;OAKG"}