@try-works/dsh-model-as-tool 0.1.3 → 0.4.0

package/lib/adapters/voz.js

@@ -1,102 +0,0 @@
-/**
- * Voz adapter — on-device speech recognition with word timestamps.
- *
- * Voz has no prebuilt Node core; it runs the same WebAssembly/ONNX Runtime
- * pipeline everywhere and requires the caller to hand it an ONNX Runtime
- * module. `onnxruntime-node` ships win32-x64 binaries, which is what makes Voz
- * the one audio model that works on Windows.
- *
- * Loading is expensive (the weights are ~349 MB and ONNX Runtime compiles three
- * graphs), so the instance is cached for the process and reused.
- *
- * @module @try-works/dsh-model-as-tool/adapters/voz
- */
-var __rewriteRelativeImportExtension = (this && this.__rewriteRelativeImportExtension) || function (path, preserveJsx) {
-    if (typeof path === "string" && /^\.\.?\//.test(path)) {
-        return path.replace(/\.(tsx)$|((?:\.d)?)((?:\.[^./]+?)?)\.([cm]?)ts$/i, function (m, tsx, d, ext, cm) {
-            return tsx ? preserveJsx ? ".jsx" : ".js" : d && (!ext || !cm) ? m : (d + ext + "." + cm.toLowerCase() + "js");
-        });
-    }
-    return path;
-};
-import { optionalBoolean, requireString, throwIfAborted, } from "./types.js";
-/** The npm package this adapter knows how to call. */
-export const VOZ_PACKAGE = '@desert-ant-labs/voz';
-/** Cap on the words returned inline, so a long recording cannot flood the context. */
-const MAX_WORDS = 2000;
-/**
- * Build the Voz adapter.
- *
- * Keyed on the package rather than the catalog id, for the same reason as the
- * Tongue adapter: the adapter knows an SDK, so any catalog row bound to it can be
- * served, which is what lets a second implementation join the contract.
- * @param entry - the catalog entry this adapter serves.
- * @returns the adapter, or undefined if the entry does not use this SDK.
- */
-export function createVozAdapter(entry) {
-    if (entry.npmPackage !== VOZ_PACKAGE)
-        return undefined;
-    let loaded;
-    const load = async () => {
-        loaded ??= (async () => {
-            try {
-                // Both specifiers are variables for the same reason as Tongue: each is
-                // an optional dependency that may legitimately be absent.
-                const runtimeSpecifier = entry.requiresRuntime[0] ?? 'onnxruntime-node';
-                const ort = await import(__rewriteRelativeImportExtension(runtimeSpecifier));
-                const module = await import(__rewriteRelativeImportExtension(entry.npmPackage));
-                return await module.Voz.load({ ort });
-            }
-            catch (error) {
-                loaded = undefined;
-                throw new Error(`Voz could not be loaded: ${error instanceof Error ? error.message : String(error)}. `
-                    + 'Voz needs onnxruntime-node installed beside the plugin.');
-            }
-        })();
-        return loaded;
-    };
-    return {
-        modelId: entry.id,
-        async invoke({ args, signal, authority }) {
-            throwIfAborted(signal);
-            const requested = requireString(args, 'audio_path');
-            const includeWords = optionalBoolean(args, 'include_words', true);
-            // SEC-06: a caller-supplied path is authorized before it is opened, and the
-            // CANONICAL path is what gets opened. Validating the string and then opening
-            // it would leave a check-then-open window a symlink could win.
-            const audioPath = authority === undefined
-                ? requested
-                : authority.resolvePath(requested);
-            const voz = await load();
-            throwIfAborted(signal);
-            const result = await voz.transcribe(audioPath);
-            throwIfAborted(signal);
-            const value = {
-                text: result.text,
-                wordCount: result.words?.length ?? 0,
-            };
-            if (typeof result.realtimeFactor === 'number') {
-                value.realtimeFactor = result.realtimeFactor;
-            }
-            if (includeWords) {
-                const words = result.words ?? [];
-                value.words = words.slice(0, MAX_WORDS).map(word => ({
-                    text: word.text,
-                    start: word.start,
-                    end: word.end,
-                }));
-                if (words.length > MAX_WORDS)
-                    value.wordsTruncated = true;
-            }
-            return value;
-        },
-        async dispose() {
-            if (loaded === undefined)
-                return;
-            const instance = await loaded.catch(() => undefined);
-            instance?.dispose?.();
-            loaded = undefined;
-        },
-    };
-}
-//# sourceMappingURL=voz.js.map