add model loading layer for Transformers.js WebGPU pipelines

Singleton loaders for embedding (embeddinggemma-300m), reranker
(Qwen3-Reranker-0.6B), and query expansion (1.7B Q4) models using
@huggingface/transformers v4 with WebGPU backend. Includes progress
reporting, WebGPU detection, and parallel loading via loadAllModels.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

src/pipeline/models.ts

@@ -0,0 +1,201 @@
+import {
+  pipeline,
+  type FeatureExtractionPipeline,
+  type TextGenerationPipeline,
+  type TextClassificationPipeline,
+  type ProgressInfo,
+} from "@huggingface/transformers";
+import { MODEL_EMBEDDING, MODEL_RERANKER, MODEL_EXPANSION } from "../constants";
+import type { ModelState } from "../types";
+
+type ProgressCallback = (state: ModelState) => void;
+
+// Singleton model instances
+let embeddingPipeline: FeatureExtractionPipeline | null = null;
+let rerankerPipeline: TextClassificationPipeline | null = null;
+let generationPipeline: TextGenerationPipeline | null = null;
+
+/** Check whether WebGPU is available in this browser. */
+export async function checkWebGPU(): Promise<boolean> {
+  if (!navigator.gpu) return false;
+  try {
+    const adapter = await navigator.gpu.requestAdapter();
+    return adapter !== null;
+  } catch {
+    return false;
+  }
+}
+
+// ---------------------------------------------------------------------------
+// Internal: translate Transformers.js ProgressInfo → our ModelState
+// ---------------------------------------------------------------------------
+
+function makeProgressHandler(
+  modelName: string,
+  onProgress?: ProgressCallback,
+): ((info: ProgressInfo) => void) | undefined {
+  if (!onProgress) return undefined;
+
+  return (info: ProgressInfo) => {
+    switch (info.status) {
+      case "initiate":
+      case "download":
+        onProgress({
+          name: modelName,
+          status: "downloading",
+          progress: 0,
+        });
+        break;
+      case "progress":
+        onProgress({
+          name: modelName,
+          status: "downloading",
+          progress: (info as { progress: number }).progress / 100,
+        });
+        break;
+      case "done":
+        onProgress({
+          name: modelName,
+          status: "loading",
+          progress: 1,
+        });
+        break;
+      case "ready":
+        onProgress({
+          name: modelName,
+          status: "ready",
+          progress: 1,
+        });
+        break;
+    }
+  };
+}
+
+// ---------------------------------------------------------------------------
+// Individual model loaders
+// ---------------------------------------------------------------------------
+
+export async function loadEmbeddingModel(
+  onProgress?: ProgressCallback,
+): Promise<void> {
+  if (embeddingPipeline) return;
+  const name = "embedding";
+  onProgress?.({ name, status: "pending", progress: 0 });
+  try {
+    embeddingPipeline = await pipeline("feature-extraction", MODEL_EMBEDDING, {
+      device: "webgpu",
+      progress_callback: makeProgressHandler(name, onProgress),
+    });
+    onProgress?.({ name, status: "ready", progress: 1 });
+  } catch (err) {
+    onProgress?.({
+      name,
+      status: "error",
+      progress: 0,
+      error: err instanceof Error ? err.message : String(err),
+    });
+    throw err;
+  }
+}
+
+export async function loadRerankerModel(
+  onProgress?: ProgressCallback,
+): Promise<void> {
+  if (rerankerPipeline) return;
+  const name = "reranker";
+  onProgress?.({ name, status: "pending", progress: 0 });
+  try {
+    rerankerPipeline = await pipeline(
+      "text-classification",
+      MODEL_RERANKER,
+      {
+        device: "webgpu",
+        progress_callback: makeProgressHandler(name, onProgress),
+      },
+    );
+    onProgress?.({ name, status: "ready", progress: 1 });
+  } catch (err) {
+    onProgress?.({
+      name,
+      status: "error",
+      progress: 0,
+      error: err instanceof Error ? err.message : String(err),
+    });
+    throw err;
+  }
+}
+
+export async function loadExpansionModel(
+  onProgress?: ProgressCallback,
+): Promise<void> {
+  if (generationPipeline) return;
+  const name = "expansion";
+  onProgress?.({ name, status: "pending", progress: 0 });
+  try {
+    generationPipeline = await pipeline(
+      "text-generation",
+      MODEL_EXPANSION,
+      {
+        dtype: "q4",
+        device: "webgpu",
+        progress_callback: makeProgressHandler(name, onProgress),
+      },
+    );
+    onProgress?.({ name, status: "ready", progress: 1 });
+  } catch (err) {
+    onProgress?.({
+      name,
+      status: "error",
+      progress: 0,
+      error: err instanceof Error ? err.message : String(err),
+    });
+    throw err;
+  }
+}
+
+// ---------------------------------------------------------------------------
+// Load all models in parallel
+// ---------------------------------------------------------------------------
+
+export async function loadAllModels(
+  onProgress?: ProgressCallback,
+): Promise<void> {
+  const hasWebGPU = await checkWebGPU();
+  if (!hasWebGPU) {
+    const err = "WebGPU is not available in this browser";
+    for (const name of ["embedding", "reranker", "expansion"]) {
+      onProgress?.({ name, status: "error", progress: 0, error: err });
+    }
+    throw new Error(err);
+  }
+
+  await Promise.all([
+    loadEmbeddingModel(onProgress),
+    loadRerankerModel(onProgress),
+    loadExpansionModel(onProgress),
+  ]);
+}
+
+// ---------------------------------------------------------------------------
+// Getters
+// ---------------------------------------------------------------------------
+
+export function getEmbeddingPipeline(): FeatureExtractionPipeline | null {
+  return embeddingPipeline;
+}
+
+export function getRerankerPipeline(): TextClassificationPipeline | null {
+  return rerankerPipeline;
+}
+
+export function getExpansionPipeline(): TextGenerationPipeline | null {
+  return generationPipeline;
+}
+
+export function isAllModelsReady(): boolean {
+  return (
+    embeddingPipeline !== null &&
+    rerankerPipeline !== null &&
+    generationPipeline !== null
+  );
+}