@phi-code-admin/browser 1.0.3 → 1.0.4

package/dist/index.d.ts

@@ -108,12 +108,6 @@ export declare function screenshot(options: {
     tabId: string;
     userId?: string;
     fullPage?: boolean;
-    clip?: {
-        x: number;
-        y: number;
-        width: number;
-        height: number;
-    };
 }): Promise<ScreenshotResult>;
 /**
  * High-level search macro: opens a new tab, navigates to a search engine

package/dist/index.js

@@ -222,6 +222,27 @@ async function request(pathname, options = {}) {
             body: options.body !== undefined ? JSON.stringify(options.body) : undefined,
             signal: controller.signal,
         });
+        if (options.responseType === "binary") {
+            if (!res.ok) {
+                // Even on error, the body is JSON — fall through to error parsing.
+                const text = await res.text();
+                let parsed = undefined;
+                if (text) {
+                    try {
+                        parsed = JSON.parse(text);
+                    }
+                    catch {
+                        parsed = text;
+                    }
+                }
+                const message = typeof parsed === "object" && parsed && "error" in parsed
+                    ? String(parsed.error)
+                    : `HTTP ${res.status}`;
+                throw new Error(`${method} ${pathname} → ${message}`);
+            }
+            const buffer = new Uint8Array(await res.arrayBuffer());
+            return buffer;
+        }
         const text = await res.text();
         let parsed = undefined;
         if (text) {
@@ -332,29 +353,60 @@ export async function extract(options) {
             sessionKey: options.sessionKey,
         });
     }
-    const res = await request(`/tabs/${encodeURIComponent(tabId)}/extract`, {
+    // The camofox-browser POST /tabs/:tabId/extract endpoint is a
+    // *deterministic* extractor that requires a structured `schema` of
+    // refs from a prior snapshot — it's not a Readability extractor. Phi
+    // callers expect a plain `{title, content, textContent}` blob, so we
+    // achieve that via /evaluate, running a small Readability-style script
+    // inside the page. This keeps the public API stable regardless of how
+    // the camofox-browser server evolves.
+    const mode = options.mode ?? "readability";
+    const expression = `(() => {
+		const limit = 50000;
+		const title = document.title || "";
+		const url = window.location.href || "";
+		if (${JSON.stringify(mode)} === "html") {
+			return { title, url, content: document.documentElement.outerHTML.slice(0, limit) };
+		}
+		if (${JSON.stringify(mode)} === "text") {
+			return { title, url, textContent: (document.body && document.body.innerText || "").slice(0, limit) };
+		}
+		// readability-light: strip nav/footer/header/aside, keep <main>/<article>/body.
+		const clone = document.cloneNode(true);
+		clone.querySelectorAll("script,style,noscript,iframe,nav,footer,header,aside,svg,form").forEach((el) => el.remove());
+		const root = clone.querySelector("main") || clone.querySelector("article") || clone.body || clone;
+		const text = (root.innerText || root.textContent || "").replace(/\\n{3,}/g, "\\n\\n").trim();
+		const excerpt = text.slice(0, 240);
+		return {
+			title,
+			url,
+			content: root.innerHTML ? root.innerHTML.slice(0, limit) : undefined,
+			textContent: text.slice(0, limit),
+			excerpt,
+			length: text.length,
+		};
+	})()`;
+    const evalRes = await request(`/tabs/${encodeURIComponent(tabId)}/evaluate`, {
         method: "POST",
-        body: {
-            userId: options.userId ?? DEFAULT_USER_ID,
-            sessionKey: options.sessionKey ?? DEFAULT_SESSION_KEY,
-            mode: options.mode ?? "readability",
-        },
+        body: { userId: options.userId ?? DEFAULT_USER_ID, expression },
     });
-    return res;
+    return evalRes.result ?? {};
 }
 /** Capture a screenshot of the given tab as a base64-encoded PNG. */
 export async function screenshot(options) {
     const query = new URLSearchParams();
     query.set("userId", options.userId ?? DEFAULT_USER_ID);
+    // The server expects `fullPage=true` (string match), not `=1`.
     if (options.fullPage)
-        query.set("fullPage", "1");
-    if (options.clip)
-        query.set("clip", JSON.stringify(options.clip));
-    const res = await request(`/tabs/${encodeURIComponent(options.tabId)}/screenshot?${query.toString()}`);
+        query.set("fullPage", "true");
+    // The camofox-browser screenshot endpoint streams a raw `image/png`
+    // body, not a JSON envelope. Pull it as a Uint8Array and base64-encode
+    // here so the result is JSON-safe for the tool result channel.
+    const bytes = await request(`/tabs/${encodeURIComponent(options.tabId)}/screenshot?${query.toString()}`, { responseType: "binary" });
     return {
         tabId: options.tabId,
-        mimeType: res.mimeType ?? "image/png",
-        bytesBase64: res.image ?? "",
+        mimeType: "image/png",
+        bytesBase64: Buffer.from(bytes).toString("base64"),
     };
 }
 /**

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@phi-code-admin/browser",
-	"version": "1.0.3",
+	"version": "1.0.4",
 	"description": "Phi-code browser automation API: lazy-start the bundled Camoufox + camofox-browser server and expose 10 high-level tools (navigate, extract, screenshot, click, type, scroll, snapshot, search, close_tab, list_tabs) as plain ES module functions. Zero external dependencies at runtime.",
 	"type": "module",
 	"main": "dist/index.js",