@llmtrim/js 0.2.1

package/README.md

@@ -0,0 +1,67 @@
+# llmtrim-wasm
+
+WebAssembly/JS bindings for the [`llmtrim-core`](../llmtrim-core) compression engine, for
+running the static prompt/payload compressor in a browser, Node, Bun, Deno, or a Cloudflare
+Worker. No network or filesystem access.
+
+Published to npm as **`@llmtrim/js`** (with **`@llmtrim/wasm`** as an alias that re-exports
+the same package).
+
+## API
+
+```ts
+import { compress } from "@llmtrim/js";
+
+const out = compress(requestBodyJson, "openai", "agent");
+// out: CompressOutput { request_json, provider, model, tokenizer_label,
+//   tokenizer_exact, input_tokens_before, input_tokens_after,
+//   frozen_input_tokens, output_shaped, stages: StageReport[] }
+```
+
+- `provider`: `"openai" | "anthropic" | "google"`, or `undefined`/`null` to auto-detect from
+  the body.
+- `preset`: a named workload preset (`aggressive`, `agent`, `code`, `rag`, `safe`, …), or
+  `undefined`/`null` for the built-in defaults. This binding never reads the environment or a
+  config file.
+
+TypeScript types for `CompressOutput` and `StageReport` are generated via `tsify`, so the
+`.d.ts` is fully typed (no `any`).
+
+This build links `llmtrim-core` with `default-features = false`: the estimate tokenizer is
+used (counts are approximate; savings percentages are unchanged), and code skeletonization
+and image downscaling are no-ops. That keeps the bundle small (~1 MB gzipped, under the
+Cloudflare Workers 3 MB free-tier cap).
+
+## Building
+
+The npm package is built in CI with `wasm-pack build --target bundler` (it manages a
+matching `wasm-bindgen` internally). The manual `cargo` + `wasm-bindgen` recipe below is the
+equivalent for local development:
+
+```sh
+rustup target add wasm32-unknown-unknown
+cargo install wasm-bindgen-cli
+
+# The JS-backed getrandom backend needs a rustc cfg in the environment (it cannot live in a
+# repo .cargo/config.toml, which would break `cargo publish`).
+RUSTFLAGS='--cfg getrandom_backend="wasm_js"' \
+  cargo build -p llmtrim-wasm --release --target wasm32-unknown-unknown
+
+wasm-bindgen target/wasm32-unknown-unknown/release/llmtrim_wasm.wasm \
+  --out-dir pkg --target bundler   # or: nodejs | web
+
+# optional size pass
+wasm-opt -O3 --enable-reference-types --enable-bulk-memory --enable-mutable-globals \
+  --enable-nontrapping-float-to-int --enable-sign-ext --enable-multivalue \
+  pkg/llmtrim_wasm_bg.wasm -o pkg/llmtrim_wasm_bg.wasm
+```
+
+## Smoke test
+
+After building with `--target nodejs` into `pkg/`, run `node smoke.mjs` (it imports
+`./pkg/llmtrim_wasm.js`). It exercises OpenAI and Anthropic requests, a named preset, an
+error path, and non-ASCII (CJK) input.
+
+## License
+
+MPL-2.0.

package/llmtrim_wasm.d.ts

@@ -0,0 +1,47 @@
+/* tslint:disable */
+/* eslint-disable */
+/**
+ * The result of compressing one request body.
+ */
+export interface CompressOutput {
+    request_json: string;
+    provider: string;
+    model: string | undefined;
+    tokenizer_label: string;
+    tokenizer_exact: boolean;
+    input_tokens_before: number;
+    input_tokens_after: number;
+    frozen_input_tokens: number;
+    output_shaped: boolean;
+    stages: StageReport[];
+}
+
+/**
+ * What one pipeline stage did to the request, projected from
+ * [`llmtrim_core::pipeline::StageReport`]. `tokens_before - tokens_after` is that stage\'s
+ * own contribution to the input reduction.
+ */
+export interface StageReport {
+    name: string;
+    applied: boolean;
+    tokens_before: number;
+    tokens_after: number;
+    note: string | undefined;
+}
+
+
+/**
+ * Compress an LLM API request body (a JSON string).
+ *
+ * - `provider`: `"openai"`, `"anthropic"`, or `"google"`; omit (`null`/`undefined`) to
+ *   auto-detect from the body shape.
+ * - `preset`: a named workload preset (`aggressive`, `agent`, `code`, `rag`, `safe`, …);
+ *   omit to use the built-in defaults. Unlike the native crate, this binding never reads
+ *   the environment or a config file (there is none in a Worker), so the configuration
+ *   comes only from the preset or the defaults.
+ *
+ * Returns the [`CompressOutput`] as a typed JS object (TypeScript types are generated for
+ * it), or throws on invalid JSON, an undetectable provider, or an unknown preset/provider
+ * name.
+ */
+export function compress(input: string, provider?: string | null, preset?: string | null): CompressOutput;

package/llmtrim_wasm.js

@@ -0,0 +1,9 @@
+/* @ts-self-types="./llmtrim_wasm.d.ts" */
+import * as wasm from "./llmtrim_wasm_bg.wasm";
+import { __wbg_set_wasm } from "./llmtrim_wasm_bg.js";
+
+__wbg_set_wasm(wasm);
+
+export {
+    compress
+} from "./llmtrim_wasm_bg.js";

package/llmtrim_wasm_bg.js

@@ -0,0 +1,228 @@
+/**
+ * Compress an LLM API request body (a JSON string).
+ *
+ * - `provider`: `"openai"`, `"anthropic"`, or `"google"`; omit (`null`/`undefined`) to
+ *   auto-detect from the body shape.
+ * - `preset`: a named workload preset (`aggressive`, `agent`, `code`, `rag`, `safe`, …);
+ *   omit to use the built-in defaults. Unlike the native crate, this binding never reads
+ *   the environment or a config file (there is none in a Worker), so the configuration
+ *   comes only from the preset or the defaults.
+ *
+ * Returns the [`CompressOutput`] as a typed JS object (TypeScript types are generated for
+ * it), or throws on invalid JSON, an undetectable provider, or an unknown preset/provider
+ * name.
+ * @param {string} input
+ * @param {string | null} [provider]
+ * @param {string | null} [preset]
+ * @returns {CompressOutput}
+ */
+export function compress(input, provider, preset) {
+    try {
+        const retptr = wasm.__wbindgen_add_to_stack_pointer(-16);
+        const ptr0 = passStringToWasm0(input, wasm.__wbindgen_export, wasm.__wbindgen_export2);
+        const len0 = WASM_VECTOR_LEN;
+        var ptr1 = isLikeNone(provider) ? 0 : passStringToWasm0(provider, wasm.__wbindgen_export, wasm.__wbindgen_export2);
+        var len1 = WASM_VECTOR_LEN;
+        var ptr2 = isLikeNone(preset) ? 0 : passStringToWasm0(preset, wasm.__wbindgen_export, wasm.__wbindgen_export2);
+        var len2 = WASM_VECTOR_LEN;
+        wasm.compress(retptr, ptr0, len0, ptr1, len1, ptr2, len2);
+        var r0 = getDataViewMemory0().getInt32(retptr + 4 * 0, true);
+        var r1 = getDataViewMemory0().getInt32(retptr + 4 * 1, true);
+        var r2 = getDataViewMemory0().getInt32(retptr + 4 * 2, true);
+        if (r2) {
+            throw takeObject(r1);
+        }
+        return takeObject(r0);
+    } finally {
+        wasm.__wbindgen_add_to_stack_pointer(16);
+    }
+}
+export function __wbg_Error_ef53bc310eb298a0(arg0, arg1) {
+    const ret = Error(getStringFromWasm0(arg0, arg1));
+    return addHeapObject(ret);
+}
+export function __wbg_String_8564e559799eccda(arg0, arg1) {
+    const ret = String(getObject(arg1));
+    const ptr1 = passStringToWasm0(ret, wasm.__wbindgen_export, wasm.__wbindgen_export2);
+    const len1 = WASM_VECTOR_LEN;
+    getDataViewMemory0().setInt32(arg0 + 4 * 1, len1, true);
+    getDataViewMemory0().setInt32(arg0 + 4 * 0, ptr1, true);
+}
+export function __wbg___wbindgen_throw_1506f2235d1bdba0(arg0, arg1) {
+    throw new Error(getStringFromWasm0(arg0, arg1));
+}
+export function __wbg_getRandomValues_3f44b700395062e5() { return handleError(function (arg0, arg1) {
+    globalThis.crypto.getRandomValues(getArrayU8FromWasm0(arg0, arg1));
+}, arguments); }
+export function __wbg_new_ce1ab61c1c2b300d() {
+    const ret = new Object();
+    return addHeapObject(ret);
+}
+export function __wbg_new_d90091b82fdf5b91() {
+    const ret = new Array();
+    return addHeapObject(ret);
+}
+export function __wbg_set_6be42768c690e380(arg0, arg1, arg2) {
+    getObject(arg0)[takeObject(arg1)] = takeObject(arg2);
+}
+export function __wbg_set_dca99999bba88a9a(arg0, arg1, arg2) {
+    getObject(arg0)[arg1 >>> 0] = takeObject(arg2);
+}
+export function __wbindgen_cast_0000000000000001(arg0) {
+    // Cast intrinsic for `F64 -> Externref`.
+    const ret = arg0;
+    return addHeapObject(ret);
+}
+export function __wbindgen_cast_0000000000000002(arg0, arg1) {
+    // Cast intrinsic for `Ref(String) -> Externref`.
+    const ret = getStringFromWasm0(arg0, arg1);
+    return addHeapObject(ret);
+}
+export function __wbindgen_cast_0000000000000003(arg0) {
+    // Cast intrinsic for `U64 -> Externref`.
+    const ret = BigInt.asUintN(64, arg0);
+    return addHeapObject(ret);
+}
+export function __wbindgen_object_clone_ref(arg0) {
+    const ret = getObject(arg0);
+    return addHeapObject(ret);
+}
+export function __wbindgen_object_drop_ref(arg0) {
+    takeObject(arg0);
+}
+function addHeapObject(obj) {
+    if (heap_next === heap.length) heap.push(heap.length + 1);
+    const idx = heap_next;
+    heap_next = heap[idx];
+
+    heap[idx] = obj;
+    return idx;
+}
+
+function dropObject(idx) {
+    if (idx < 1028) return;
+    heap[idx] = heap_next;
+    heap_next = idx;
+}
+
+function getArrayU8FromWasm0(ptr, len) {
+    ptr = ptr >>> 0;
+    return getUint8ArrayMemory0().subarray(ptr / 1, ptr / 1 + len);
+}
+
+let cachedDataViewMemory0 = null;
+function getDataViewMemory0() {
+    if (cachedDataViewMemory0 === null || cachedDataViewMemory0.buffer.detached === true || (cachedDataViewMemory0.buffer.detached === undefined && cachedDataViewMemory0.buffer !== wasm.memory.buffer)) {
+        cachedDataViewMemory0 = new DataView(wasm.memory.buffer);
+    }
+    return cachedDataViewMemory0;
+}
+
+function getStringFromWasm0(ptr, len) {
+    return decodeText(ptr >>> 0, len);
+}
+
+let cachedUint8ArrayMemory0 = null;
+function getUint8ArrayMemory0() {
+    if (cachedUint8ArrayMemory0 === null || cachedUint8ArrayMemory0.byteLength === 0) {
+        cachedUint8ArrayMemory0 = new Uint8Array(wasm.memory.buffer);
+    }
+    return cachedUint8ArrayMemory0;
+}
+
+function getObject(idx) { return heap[idx]; }
+
+function handleError(f, args) {
+    try {
+        return f.apply(this, args);
+    } catch (e) {
+        wasm.__wbindgen_export3(addHeapObject(e));
+    }
+}
+
+let heap = new Array(1024).fill(undefined);
+heap.push(undefined, null, true, false);
+
+let heap_next = heap.length;
+
+function isLikeNone(x) {
+    return x === undefined || x === null;
+}
+
+function passStringToWasm0(arg, malloc, realloc) {
+    if (realloc === undefined) {
+        const buf = cachedTextEncoder.encode(arg);
+        const ptr = malloc(buf.length, 1) >>> 0;
+        getUint8ArrayMemory0().subarray(ptr, ptr + buf.length).set(buf);
+        WASM_VECTOR_LEN = buf.length;
+        return ptr;
+    }
+
+    let len = arg.length;
+    let ptr = malloc(len, 1) >>> 0;
+
+    const mem = getUint8ArrayMemory0();
+
+    let offset = 0;
+
+    for (; offset < len; offset++) {
+        const code = arg.charCodeAt(offset);
+        if (code > 0x7F) break;
+        mem[ptr + offset] = code;
+    }
+    if (offset !== len) {
+        if (offset !== 0) {
+            arg = arg.slice(offset);
+        }
+        ptr = realloc(ptr, len, len = offset + arg.length * 3, 1) >>> 0;
+        const view = getUint8ArrayMemory0().subarray(ptr + offset, ptr + len);
+        const ret = cachedTextEncoder.encodeInto(arg, view);
+
+        offset += ret.written;
+        ptr = realloc(ptr, len, offset, 1) >>> 0;
+    }
+
+    WASM_VECTOR_LEN = offset;
+    return ptr;
+}
+
+function takeObject(idx) {
+    const ret = getObject(idx);
+    dropObject(idx);
+    return ret;
+}
+
+let cachedTextDecoder = new TextDecoder('utf-8', { ignoreBOM: true, fatal: true });
+cachedTextDecoder.decode();
+const MAX_SAFARI_DECODE_BYTES = 2146435072;
+let numBytesDecoded = 0;
+function decodeText(ptr, len) {
+    numBytesDecoded += len;
+    if (numBytesDecoded >= MAX_SAFARI_DECODE_BYTES) {
+        cachedTextDecoder = new TextDecoder('utf-8', { ignoreBOM: true, fatal: true });
+        cachedTextDecoder.decode();
+        numBytesDecoded = len;
+    }
+    return cachedTextDecoder.decode(getUint8ArrayMemory0().subarray(ptr, ptr + len));
+}
+
+const cachedTextEncoder = new TextEncoder();
+
+if (!('encodeInto' in cachedTextEncoder)) {
+    cachedTextEncoder.encodeInto = function (arg, view) {
+        const buf = cachedTextEncoder.encode(arg);
+        view.set(buf);
+        return {
+            read: arg.length,
+            written: buf.length
+        };
+    };
+}
+
+let WASM_VECTOR_LEN = 0;
+
+
+let wasm;
+export function __wbg_set_wasm(val) {
+    wasm = val;
+}

package/package.json

@@ -0,0 +1,24 @@
+{
+  "name": "@llmtrim/js",
+  "type": "module",
+  "description": "WebAssembly/JS bindings for the llmtrim-core compression engine (wasm-bindgen).",
+  "version": "0.2.1",
+  "license": "MPL-2.0",
+  "repository": {
+    "type": "git",
+    "url": "https://github.com/fkiene/llmtrim"
+  },
+  "files": [
+    "llmtrim_wasm_bg.wasm",
+    "llmtrim_wasm.js",
+    "llmtrim_wasm_bg.js",
+    "llmtrim_wasm.d.ts"
+  ],
+  "main": "llmtrim_wasm.js",
+  "homepage": "https://github.com/fkiene/llmtrim",
+  "types": "llmtrim_wasm.d.ts",
+  "sideEffects": [
+    "./llmtrim_wasm.js",
+    "./snippets/*"
+  ]
+}