veryfront 0.1.867 → 0.1.869

package/esm/src/agent/react/use-chat/browser-inference/worker-client.js

@@ -1,66 +0,0 @@
-/**
- * BrowserInferenceClient — manages Web Worker lifecycle for browser-side inference.
- *
- * Singleton per session. Lazily creates Worker on first generate() call.
- * Uses inline Blob URL approach — no separate build entry point needed.
- */
-import { WORKER_SCRIPT } from "./worker-script.js";
-let instance = null;
-export class BrowserInferenceClient {
-    worker = null;
-    blobUrl = null;
-    static getInstance() {
-        if (!instance)
-            instance = new BrowserInferenceClient();
-        return instance;
-    }
-    ensureWorker() {
-        if (this.worker)
-            return this.worker;
-        const blob = new Blob([WORKER_SCRIPT], { type: "application/javascript" });
-        this.blobUrl = URL.createObjectURL(blob);
-        this.worker = new Worker(this.blobUrl, { type: "module" });
-        return this.worker;
-    }
-    generate(id, messages, options, callbacks) {
-        const worker = this.ensureWorker();
-        worker.onmessage = (event) => {
-            const msg = event.data;
-            switch (msg.type) {
-                case "status":
-                    callbacks.onStatus?.(msg.status);
-                    break;
-                case "download-progress":
-                    callbacks.onDownloadProgress?.(msg.progress, msg.file);
-                    break;
-                case "token":
-                    if (msg.id === id)
-                        callbacks.onToken?.(msg.token);
-                    break;
-                case "done":
-                    if (msg.id === id)
-                        callbacks.onDone?.(msg.text);
-                    break;
-                case "error":
-                    if (msg.id === id)
-                        callbacks.onError?.(msg.error);
-                    break;
-            }
-        };
-        worker.onerror = (event) => {
-            callbacks.onError?.(event.message || "Worker error");
-        };
-        const request = { type: "generate", id, messages, options };
-        worker.postMessage(request);
-    }
-    stop() {
-        if (this.worker) {
-            this.worker.terminate();
-            this.worker = null;
-        }
-        if (this.blobUrl) {
-            URL.revokeObjectURL(this.blobUrl);
-            this.blobUrl = null;
-        }
-    }
-}

package/esm/src/agent/react/use-chat/browser-inference/worker-script.d.ts

@@ -1,8 +0,0 @@
-/**
- * Inline Worker script as a string.
- *
- * Loaded via Blob URL — no separate build entry point needed.
- * Dynamically imports @huggingface/transformers from CDN inside the Worker.
- */
-export declare const WORKER_SCRIPT = "\nlet pipeline = null;\nlet generating = false;\n\nasync function loadPipeline(callbacks) {\n  if (pipeline) return pipeline;\n\n  callbacks.onStatus(\"loading-runtime\");\n\n  const { pipeline: createPipeline, env } =\n    await import(\"https://cdn.jsdelivr.net/npm/@huggingface/transformers@3.4.2\");\n\n  env.useBrowserCache = true;\n  env.allowLocalModels = false;\n\n  callbacks.onStatus(\"downloading-model\");\n\n  pipeline = await createPipeline(\n    \"text-generation\",\n    \"HuggingFaceTB/SmolLM2-135M-Instruct\",\n    {\n      dtype: \"q4\",\n      device: \"wasm\",\n      progress_callback: (progress) => {\n        if (progress.status === \"progress\" && progress.total) {\n          callbacks.onProgress(Math.round((progress.loaded / progress.total) * 100), progress.file);\n        }\n      },\n    },\n  );\n\n  callbacks.onStatus(\"ready\");\n  return pipeline;\n}\n\nself.onmessage = async (event) => {\n  const request = event.data;\n  if (request.type !== \"generate\") return;\n\n  const { id, messages, options } = request;\n\n  try {\n    const pipe = await loadPipeline({\n      onStatus: (status) => self.postMessage({ type: \"status\", status }),\n      onProgress: (progress, file) => self.postMessage({ type: \"download-progress\", progress, file }),\n    });\n\n    self.postMessage({ type: \"status\", status: \"generating\" });\n    generating = true;\n\n    const chatMessages = [];\n    if (options?.systemPrompt) {\n      chatMessages.push({ role: \"system\", content: options.systemPrompt });\n    }\n    chatMessages.push(...messages);\n\n    // Helper: generated_text is a plain string for raw prompts but an array\n    // of {role, content} message objects when using chat format. Extract the\n    // last assistant message's content in either case.\n    function extractText(generated) {\n      if (typeof generated === \"string\") return generated;\n      if (Array.isArray(generated)) {\n        const last = generated[generated.length - 1];\n        return last?.content ?? \"\";\n      }\n      return \"\";\n    }\n\n    const result = await pipe(chatMessages, {\n      max_new_tokens: options?.maxNewTokens ?? 512,\n      temperature: options?.temperature ?? 0.7,\n      do_sample: true,\n      return_full_text: false,\n      callback_function: (output) => {\n        if (!generating) return;\n        const text = extractText(output?.[0]?.generated_text);\n        if (text) {\n          self.postMessage({ type: \"token\", id, token: text });\n        }\n      },\n    });\n\n    generating = false;\n    const finalText = extractText(result?.[0]?.generated_text);\n    self.postMessage({ type: \"done\", id, text: finalText });\n  } catch (error) {\n    generating = false;\n    self.postMessage({ type: \"error\", id, error: error?.message ?? String(error) });\n  }\n};\n";
-//# sourceMappingURL=worker-script.d.ts.map

package/esm/src/agent/react/use-chat/browser-inference/worker-script.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"worker-script.d.ts","sourceRoot":"","sources":["../../../../../../src/src/agent/react/use-chat/browser-inference/worker-script.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AAEH,eAAO,MAAM,aAAa,+wFA0FzB,CAAC"}

package/esm/src/agent/react/use-chat/browser-inference/worker-script.js

@@ -1,97 +0,0 @@
-/**
- * Inline Worker script as a string.
- *
- * Loaded via Blob URL — no separate build entry point needed.
- * Dynamically imports @huggingface/transformers from CDN inside the Worker.
- */
-export const WORKER_SCRIPT = /* js */ `
-let pipeline = null;
-let generating = false;
-
-async function loadPipeline(callbacks) {
-  if (pipeline) return pipeline;
-
-  callbacks.onStatus("loading-runtime");
-
-  const { pipeline: createPipeline, env } =
-    await import("https://cdn.jsdelivr.net/npm/@huggingface/transformers@3.4.2");
-
-  env.useBrowserCache = true;
-  env.allowLocalModels = false;
-
-  callbacks.onStatus("downloading-model");
-
-  pipeline = await createPipeline(
-    "text-generation",
-    "HuggingFaceTB/SmolLM2-135M-Instruct",
-    {
-      dtype: "q4",
-      device: "wasm",
-      progress_callback: (progress) => {
-        if (progress.status === "progress" && progress.total) {
-          callbacks.onProgress(Math.round((progress.loaded / progress.total) * 100), progress.file);
-        }
-      },
-    },
-  );
-
-  callbacks.onStatus("ready");
-  return pipeline;
-}
-
-self.onmessage = async (event) => {
-  const request = event.data;
-  if (request.type !== "generate") return;
-
-  const { id, messages, options } = request;
-
-  try {
-    const pipe = await loadPipeline({
-      onStatus: (status) => self.postMessage({ type: "status", status }),
-      onProgress: (progress, file) => self.postMessage({ type: "download-progress", progress, file }),
-    });
-
-    self.postMessage({ type: "status", status: "generating" });
-    generating = true;
-
-    const chatMessages = [];
-    if (options?.systemPrompt) {
-      chatMessages.push({ role: "system", content: options.systemPrompt });
-    }
-    chatMessages.push(...messages);
-
-    // Helper: generated_text is a plain string for raw prompts but an array
-    // of {role, content} message objects when using chat format. Extract the
-    // last assistant message's content in either case.
-    function extractText(generated) {
-      if (typeof generated === "string") return generated;
-      if (Array.isArray(generated)) {
-        const last = generated[generated.length - 1];
-        return last?.content ?? "";
-      }
-      return "";
-    }
-
-    const result = await pipe(chatMessages, {
-      max_new_tokens: options?.maxNewTokens ?? 512,
-      temperature: options?.temperature ?? 0.7,
-      do_sample: true,
-      return_full_text: false,
-      callback_function: (output) => {
-        if (!generating) return;
-        const text = extractText(output?.[0]?.generated_text);
-        if (text) {
-          self.postMessage({ type: "token", id, token: text });
-        }
-      },
-    });
-
-    generating = false;
-    const finalText = extractText(result?.[0]?.generated_text);
-    self.postMessage({ type: "done", id, text: finalText });
-  } catch (error) {
-    generating = false;
-    self.postMessage({ type: "error", id, error: error?.message ?? String(error) });
-  }
-};
-`;

package/esm/src/react/components/chat/chat/components/upgrade-cta.d.ts

@@ -1,9 +0,0 @@
-import * as React from "../../../../../../react/react.js";
-import type { InferenceMode } from "../../../../../agent/react/index.js";
-/** Props accepted by the upgrade CTA. */
-export interface UpgradeCTAProps {
-    inferenceMode: InferenceMode;
-}
-/** Render upgrade CTA. */
-export declare function UpgradeCTA({ inferenceMode }: UpgradeCTAProps): React.ReactElement | null;
-//# sourceMappingURL=upgrade-cta.d.ts.map

package/esm/src/react/components/chat/chat/components/upgrade-cta.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"upgrade-cta.d.ts","sourceRoot":"","sources":["../../../../../../../src/src/react/components/chat/chat/components/upgrade-cta.tsx"],"names":[],"mappings":"AAAA,OAAO,KAAK,KAAK,MAAM,kCAAkC,CAAC;AAC1D,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,qCAAqC,CAAC;AAIzE,yCAAyC;AACzC,MAAM,WAAW,eAAe;IAC9B,aAAa,EAAE,aAAa,CAAC;CAC9B;AAED,0BAA0B;AAC1B,wBAAgB,UAAU,CAAC,EAAE,aAAa,EAAE,EAAE,eAAe,GAAG,KAAK,CAAC,YAAY,GAAG,IAAI,CA8CxF"}

package/esm/src/react/components/chat/chat/components/upgrade-cta.js

@@ -1,35 +0,0 @@
-import * as React from "../../../../../../react/react.js";
-const DISMISS_KEY = "vf-upgrade-cta-dismissed";
-/** Render upgrade CTA. */
-export function UpgradeCTA({ inferenceMode }) {
-    const [dismissed, setDismissed] = React.useState(() => {
-        try {
-            return localStorage.getItem(DISMISS_KEY) === "1";
-        }
-        catch (_) {
-            /* expected: localStorage may be unavailable */
-            return false;
-        }
-    });
-    if (inferenceMode === "cloud" || dismissed)
-        return null;
-    const handleDismiss = () => {
-        setDismissed(true);
-        try {
-            localStorage.setItem(DISMISS_KEY, "1");
-        }
-        catch (_) {
-            /* expected: localStorage may be unavailable */
-        }
-    };
-    return (React.createElement("div", { className: "w-full max-w-2xl mx-auto mt-4 px-4 py-3 bg-[var(--accent)] rounded-xl text-sm text-[var(--foreground)] flex items-start gap-3" },
-        React.createElement("span", { className: "flex-1" },
-            "Using a lightweight local model. Add an API key to your",
-            " ",
-            React.createElement("code", { className: "px-1 py-0.5 bg-[var(--border)] rounded text-xs" }, ".env"),
-            " ",
-            "for GPT-4o or Claude."),
-        React.createElement("button", { type: "button", onClick: handleDismiss, className: "text-[var(--muted-foreground)] hover:text-[var(--foreground)] transition-all flex-shrink-0 rounded-full focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-[var(--ring)] focus-visible:ring-offset-2", "aria-label": "Dismiss" },
-            React.createElement("svg", { className: "size-4", viewBox: "0 0 16 16", fill: "none", stroke: "currentColor", strokeWidth: "2" },
-                React.createElement("path", { d: "M4 4l8 8M12 4l-8 8" })))));
-}