@herbertgao/pi-extensions 2026.10.0 → 2026.10.2

package/node_modules/pi-web-access/exa.ts

@@ -43,6 +43,8 @@ export type ExaSearchResult = SearchResponse | null;
 
 export interface ExaSearchOptions extends SearchOptions {
 	includeContent?: boolean;
+	/** Restrict results to an Exa content category (e.g. "research paper", "news", "pdf"). */
+	category?: string;
 }
 
 type McpParsedResult = { title: string; url: string; content: string };
@@ -132,6 +134,7 @@ function exaSearchArgs(query: string, options: ExaSearchOptions): Record<string,
 		type: "auto",
 		numResults: options.numResults ?? 5,
 		...mapDomainFilter(options.domainFilter),
+		...(options.category ? { category: options.category } : {}),
 		...(startDate ? { startPublishedDate: startDate } : {}),
 	};
 }
@@ -345,6 +348,7 @@ function buildMcpQuery(query: string, options: ExaSearchOptions): string {
 			case "year": parts.push(String(now.getFullYear())); break;
 		}
 	}
+	if (options.category) parts.push(options.category);
 	return parts.join(" ");
 }
 
@@ -414,7 +418,7 @@ async function searchWithFilteredExaMcp(
 async function searchWithExaMcp(query: string, options: ExaSearchOptions = {}): Promise<SearchResponse | null> {
 	const activityId = activityMonitor.logStart({ type: "api", query });
 	const basicArgs = { query: buildMcpQuery(query, options), numResults: options.numResults ?? 5 };
-	const filtered = !!options.includeContent || !!options.recencyFilter || !!options.domainFilter?.length;
+	const filtered = !!options.includeContent || !!options.recencyFilter || !!options.domainFilter?.length || !!options.category;
 
 	try {
 		const response = filtered

package/node_modules/pi-web-access/extract.ts

@@ -149,6 +149,12 @@ function isRedirectPolicyError(message: string): boolean {
 		message.startsWith("Failed to resolve ");
 }
 
+// Errors that end routing: no other provider may be tried after them.
+function isNonRecoverableHttpError(error: string | null): boolean {
+	return !!error && (NON_RECOVERABLE_ERRORS.some(prefix => error.startsWith(prefix)) ||
+		isRedirectPolicyError(error) || isConfigParseError(error));
+}
+
 function imageGateError(): string | null {
 	try {
 		return isImageEnabled() ? null : "Image fetching is disabled by image.enabled";
@@ -360,6 +366,9 @@ export interface ExtractOptions {
 	toolNames?: RegisteredToolNames;
 	/** Optional HTTP(S) or SOCKS proxy URL; routed through the curl-backed transport. */
 	proxy?: string;
+	/** Reject direct image URLs with this error instead of decoding them with Pi's
+	 * resizeImage. Hosts without the Pi runtime set it; Pi leaves it unset. */
+	rejectDirectImages?: string;
 	/** Custom DNS resolver used for SSRF validation. Primarily a test seam. */
 	lookup?: Lookup;
 }
@@ -811,7 +820,7 @@ export async function extractContent(
 		declaredLinks = discoveredLinks;
 		if (signal?.aborted) return abortedResult(url);
 		if (!httpResult.error) return httpResult;
-		if (NON_RECOVERABLE_ERRORS.some(prefix => httpResult!.error!.startsWith(prefix)) || isRedirectPolicyError(httpResult.error) || isConfigParseError(httpResult.error)) {
+		if (isNonRecoverableHttpError(httpResult.error) || httpResult.error === options?.rejectDirectImages) {
 			return httpResult;
 		}
 		return null;
@@ -1175,11 +1184,19 @@ function responseSizeLimitError(maxBytes: number): Error {
 	return new Error(`Response too large (${Math.round(maxBytes / 1024 / 1024)}MB)`);
 }
 
+const BROWSER_ACCEPT = "text/html,application/xhtml+xml,application/xml;q=0.9,image/avif,image/webp,image/apng,*/*;q=0.8";
+// Servers that support markdown content negotiation (Cloudflare Markdown for
+// Agents, Mintlify, Vercel) return markdown directly; others see the same
+// relative preferences as BROWSER_ACCEPT. Raw mode keeps BROWSER_ACCEPT so it
+// returns the server's normal representation.
+const MARKDOWN_FIRST_ACCEPT = "text/markdown,text/html;q=0.9,application/xhtml+xml;q=0.9,application/xml;q=0.8,image/avif;q=0.9,image/webp;q=0.9,image/apng;q=0.9,*/*;q=0.7";
+
 async function extractViaHttp(
 	url: string,
 	timeoutMs: number,
 	signal?: AbortSignal,
 	options?: ExtractOptions,
+	preferMarkdown = options?.mode !== "raw",
 ): Promise<HttpExtractedContent> {
 	const activityId = activityMonitor.logStart({ type: "fetch", url });
 
@@ -1199,7 +1216,7 @@ async function extractViaHttp(
 			__proxy: options?.proxy,
 			headers: {
 				"User-Agent": "OpenAI File Downloader, XaiImageApiFetch/1.0",
-				"Accept": "text/html,application/xhtml+xml,application/xml;q=0.9,image/avif,image/webp,image/apng,*/*;q=0.8",
+				"Accept": preferMarkdown ? MARKDOWN_FIRST_ACCEPT : BROWSER_ACCEPT,
 				"Accept-Language": "en-US,en;q=0.9",
 				"Cache-Control": "no-cache",
 				"Sec-Fetch-Dest": "document",
@@ -1275,6 +1292,10 @@ async function extractViaHttp(
 				activityMonitor.logComplete(activityId, response.status);
 				return { url, title: "", content: "", error: disabled, mimeType, status: response.status };
 			}
+			if (options?.rejectDirectImages) {
+				activityMonitor.logComplete(activityId, response.status);
+				return { url, title: "", content: "", error: options.rejectDirectImages, mimeType, status: response.status };
+			}
 			try {
 				const buffer = await readResponseBufferWithLimit(response, maxResponseSize, () => responseSizeLimitError(maxResponseSize));
 				const { resizeImage } = await import("@earendil-works/pi-coding-agent");
@@ -1350,7 +1371,24 @@ async function extractViaHttp(
 		if (!isHTML) {
 			activityMonitor.logComplete(activityId, response.status);
 			const title = extractTextTitle(text, url);
-			return { url, title, content: text, error: null };
+			if (mimeType !== "text/markdown" && mimeType !== "text/x-markdown") return { url, title, content: text, error: null };
+			const declaredLinks = discoverDeclaredWebLinks(null, response.headers.get("link"), response.url || url);
+			const markdown = { url, title, content: appendDeclaredWebLinks(text, declaredLinks), error: null, declaredLinks };
+			if (!preferMarkdown || text.trim().length >= MIN_USEFUL_CONTENT) return markdown;
+			// Short negotiated markdown is re-requested with the browser Accept header.
+			// That response stays authoritative for policy/config rejection, not-found
+			// status, cancellation, and any content it has. This markdown only fills a
+			// recoverable empty response, marked incomplete with its status and error.
+			const normal = await extractViaHttp(url, Math.max(1, timeoutMs - (Date.now() - startedAt)), signal, options, false);
+			if (signal?.aborted || isNonRecoverableHttpError(normal.error) ||
+				normal.status === 404 || normal.status === 410 || isAbortError(normal.error)) return normal;
+			const normalHasBody = normal.content.trim() !== appendDeclaredWebLinks("", normal.declaredLinks ?? []).trim();
+			if (normalHasBody || !markdown.content.trim()) return normal;
+			return {
+				...normal,
+				...markdown,
+				error: "Extracted content appears incomplete" + (normal.error ? `\nBrowser retry failed: ${normal.error}` : ""),
+			};
 		}
 
 		const { parseHTML } = await import("linkedom");

package/node_modules/pi-web-access/firecrawl.ts

@@ -37,12 +37,18 @@ interface FirecrawlConfig {
 	firecrawlApiKey?: unknown;
 	firecrawlApiVersion?: unknown;
 	firecrawlFreshScrape?: unknown;
+	firecrawlPdfMaxPages?: unknown;
+}
+
+interface FirecrawlPageMetadata {
+	numPages?: unknown;
+	totalPages?: unknown;
 }
 
 interface FirecrawlScrapeData {
 	title?: unknown;
 	markdown?: unknown;
-	metadata?: { title?: unknown };
+	metadata?: { title?: unknown } & FirecrawlPageMetadata;
 }
 
 interface FirecrawlSearchItem {
@@ -51,7 +57,7 @@ interface FirecrawlSearchItem {
 	snippet?: unknown;
 	url?: unknown;
 	markdown?: unknown;
-	metadata?: {
+	metadata?: FirecrawlPageMetadata & {
 		title?: unknown;
 		description?: unknown;
 		sourceURL?: unknown;
@@ -155,6 +161,28 @@ function allowFreshScrape(): boolean {
 	return configured;
 }
 
+function pdfParserOptions(): { parsers?: Array<{ type: "pdf"; maxPages: number }> } {
+	const configured = loadConfig().firecrawlPdfMaxPages;
+	if (configured === undefined || configured === null) return {};
+	if (typeof configured !== "number" || !Number.isInteger(configured) || configured < 1) {
+		throw new Error(`firecrawlPdfMaxPages in ${CONFIG_PATH} must be a positive integer`);
+	}
+	if (getApiVersion() === "v1") {
+		throw new Error(
+			`firecrawlPdfMaxPages in ${CONFIG_PATH} requires Firecrawl API v2; v1 has no PDF page cap. ` +
+			"Use firecrawlApiVersion \"v2\" or remove firecrawlPdfMaxPages",
+		);
+	}
+	return { parsers: [{ type: "pdf", maxPages: configured }] };
+}
+
+function withPdfTruncationNotice(content: string, metadata: FirecrawlPageMetadata | undefined): string {
+	const parsed = metadata?.numPages;
+	const total = metadata?.totalPages;
+	if (typeof parsed !== "number" || typeof total !== "number" || parsed >= total) return content;
+	return `${content}\n\n---\n\n*[Truncated: Only first ${parsed} of ${total} PDF pages parsed by Firecrawl]*`;
+}
+
 async function getApiKey(signal?: AbortSignal): Promise<string | null> {
 	const configKey = loadConfig().firecrawlApiKey;
 	return resolveCredential({
@@ -234,6 +262,7 @@ function scrapeBody(url: string): Record<string, unknown> {
 		formats: ["markdown"],
 		onlyMainContent: true,
 		...(allowFreshScrape() ? {} : { lockdown: true }),
+		...pdfParserOptions(),
 	};
 }
 
@@ -272,6 +301,7 @@ function mapRecencyFilter(value: SearchOptions["recencyFilter"]): string | undef
 }
 
 function searchBody(query: string, options: FirecrawlSearchOptions, numResults: number, filters: DomainFilters): Record<string, unknown> {
+	const parserOptions = pdfParserOptions();
 	return {
 		query,
 		limit: numResults,
@@ -284,6 +314,7 @@ function searchBody(query: string, options: FirecrawlSearchOptions, numResults:
 				formats: ["markdown"],
 				onlyMainContent: true,
 				...(allowFreshScrape() ? {} : { lockdown: true }),
+				...parserOptions,
 			},
 		} : {}),
 	};
@@ -317,7 +348,7 @@ function mapSearchResults(data: unknown, numResults: number, filters: DomainFilt
 		const snippet = firstString(item.description, item.snippet, item.metadata?.description) ?? "";
 		results.push({ title, url, snippet });
 		const markdown = firstString(item.markdown);
-		if (markdown) inlineContent.push({ url, title, content: markdown, error: null });
+		if (markdown) inlineContent.push({ url, title, content: withPdfTruncationNotice(markdown, item.metadata), error: null });
 		if (results.length >= numResults) break;
 	}
 	return { results, inlineContent };
@@ -419,5 +450,5 @@ export async function extractWithFirecrawl(
 	const title = typeof metadataTitle === "string" && metadataTitle.trim()
 		? metadataTitle.trim()
 		: typeof scrape.title === "string" ? scrape.title.trim() : "";
-	return { url, title, content, error: null };
+	return { url, title, content: withPdfTruncationNotice(content, scrape.metadata), error: null };
 }

package/node_modules/pi-web-access/gemini-search.ts

@@ -31,6 +31,7 @@ import { isKagiAvailable, searchWithKagi } from "./kagi.ts";
 import { isBochaAvailable, searchWithBocha } from "./bocha.ts";
 import { isOllamaAvailable, searchWithOllama } from "./ollama.ts";
 import { isSearXNGAvailable, searchWithSearXNG } from "./searxng.ts";
+import { isDegoogAvailable, searchWithDegoog } from "./degoog.ts";
 import { isDuckDuckGoAvailable, searchWithDuckDuckGo } from "./duckduckgo.ts";
 import { isAnySearchAvailable, searchWithAnySearch } from "./anysearch.ts";
 import { isXcrawlAvailable, searchWithXCrawl } from "./xcrawl.ts";
@@ -43,11 +44,12 @@ import { isSerplyAvailable, searchWithSerply } from "./serply.ts";
 import { isBaizhiAvailable, searchWithBaizhi } from "./baizhi.ts";
 import { isZaiAvailable, searchWithZai } from "./zai.ts";
 import { isValyuAvailable, searchWithValyu } from "./valyu.ts";
+import { isKeenableAvailable, searchWithKeenable } from "./keenable.ts";
 import { isKimiSearchAvailable, searchWithKimi } from "./kimi-search.ts";
 import { isMistralAvailable, searchWithMistral } from "./mistral-search.ts";
 import { getWebSearchConfigPath } from "./utils.ts";
 
-export const RESOLVED_SEARCH_PROVIDERS = ["openai", "brave", "parallel", "parallel-mcp", "tinyfish", "search1api", "searchinfinity", "querit", "tavily", "you", "firecrawl", "jina", "searxng", "duckduckgo", "perplexity", "gemini", "kimi", "exa", "serpdive", "kagi", "ollama", "anysearch", "xai", "mistral", "brightdata", "serpbase", "serpapi", "serper", "serply", "valyu", "bocha", "xcrawl", "baizhi", "zai"] as const;
+export const RESOLVED_SEARCH_PROVIDERS = ["openai", "brave", "parallel", "parallel-mcp", "tinyfish", "search1api", "searchinfinity", "querit", "tavily", "you", "firecrawl", "jina", "searxng", "degoog", "duckduckgo", "perplexity", "gemini", "kimi", "exa", "serpdive", "kagi", "ollama", "anysearch", "xai", "mistral", "brightdata", "serpbase", "serpapi", "serper", "serply", "valyu", "bocha", "xcrawl", "baizhi", "zai", "keenable"] as const;
 export const SEARCH_PROVIDERS = ["auto", "all", ...RESOLVED_SEARCH_PROVIDERS] as const;
 
 export type ResolvedSearchProvider = typeof RESOLVED_SEARCH_PROVIDERS[number];
@@ -112,7 +114,7 @@ export interface AttributedSearchResponse extends SearchResponse {
 
 const CONFIG_PATH = getWebSearchConfigPath();
 const DEFAULT_SEARCH_MODEL = "gemini-3.6-flash";
-// Explicit-only providers (Parallel MCP, DuckDuckGo, Kimi, AnySearch, XCrawl, xAI, Mistral, Bright Data, SerpBase, SerpApi, Serper, Serply, You.com, Valyu, Baizhi, Z.ai) are deliberately absent:
+// Explicit-only providers (Parallel MCP, degoog, DuckDuckGo, Kimi, AnySearch, XCrawl, xAI, Mistral, Bright Data, SerpBase, SerpApi, Serper, Serply, You.com, Valyu, Baizhi, Z.ai, Keenable) are deliberately absent:
 // `all` must never fan out to an opt-in or paid provider without the user asking for it.
 export const ALL_SEARCH_PROVIDERS: ResolvedSearchProvider[] = ["searxng", "openai", "exa", "brave", "parallel", "tinyfish", "search1api", "searchinfinity", "querit", "tavily", "firecrawl", "jina", "serpdive", "kagi", "ollama", "perplexity", "gemini", "bocha"];
 const VALID_ROUTING_KINDS = ["transient", "quota", "network", "invalid-response", "unsupported"] as const;
@@ -264,6 +266,8 @@ export interface FullSearchOptions extends SearchOptions {
 	provider?: SearchProviderSelection;
 	includeContent?: boolean;
 	extensionContext?: ExtensionContext;
+	/** Forwarded to providers that understand it (currently Exa); ignored elsewhere. */
+	category?: string;
 }
 
 function errorMessage(err: unknown): string {
@@ -330,6 +334,16 @@ function classifyProviderError(provider: ResolvedSearchProvider, err: unknown):
 	const message = errorMessage(err);
 	const lower = message.toLowerCase();
 	const status = providerErrorStatus(message);
+	// An HTTP 200 OpenAI stream reports its failure as an error type/code such as
+	// rate_limit_exceeded or server_error. Reading only that message with spaces
+	// lets the shared keywords below classify it like prose, without widening them
+	// for other providers, which echo foreign text (Bright Data zone names,
+	// envelope codes) that must keep classifying as before. OpenAI error types
+	// the shared keywords don't name (usage limit, authentication_error,
+	// permission_error, invalid_api_key, api_error, internal_error, overloaded)
+	// are added per branch for these messages only.
+	const openAIStreamError = provider === "openai" && lower.includes("openai api stream error");
+	const text = openAIStreamError ? lower.replace(/_/g, " ") : lower;
 	let kind: SearchProviderErrorKind = "unknown";
 	const mentionsUnsupportedWebSearch = /(?:web[_ -]?search|web[_ -]?search_preview|(?:the )?tool)\b.*\b(?:unsupported|not supported|does not support|doesn't support|unknown|unrecognized|unavailable|not found)|\b(?:unsupported|not supported|does not support|doesn't support|unknown|unrecognized|unavailable|not found)\b.*\b(?:web[_ -]?search|web[_ -]?search_preview|(?:the )?tool)/i.test(lower);
 	if (err instanceof CredentialResolutionError || /(?:api )?key (?:not found|missing)|credential resolution/.test(lower)) {
@@ -350,20 +364,23 @@ function classifyProviderError(provider: ResolvedSearchProvider, err: unknown):
 		kind = "quota";
 	} else if (status !== undefined && (status === 408 || status === 425 || status >= 500)) {
 		kind = "transient";
-	} else if (/rate limit|quota|too many requests/.test(lower)) {
+	} else if (/rate limit|quota|too many requests/.test(text) || (openAIStreamError && /usage limit/.test(text))) {
 		kind = "quota";
-	} else if (/unauthorized|forbidden|permission denied/.test(lower)) {
+	} else if (/unauthorized|forbidden|permission denied/.test(text) || (openAIStreamError && /authentication error|permission error|invalid api key/.test(text))) {
 		kind = "auth";
-	} else if (/bad request|invalid request/.test(lower)) {
+	} else if (/bad request|invalid request/.test(text)) {
 		kind = "invalid-request";
-	} else if (/invalid json|no parseable response|no parseable results|invalid response|returned empty response|no web_search_call/.test(lower)) {
+	} else if (/invalid json|no parseable response|no parseable results|invalid response|returned empty response|no web_search_call/.test(text)) {
 		kind = "invalid-response";
-	} else if (/temporar|service unavailable|server error/.test(lower)) {
+	} else if (/temporar|service unavailable|server error/.test(text) || (openAIStreamError && /internal error|\bapi error|overloaded/.test(text))) {
 		kind = "transient";
-	} else if (err instanceof TypeError || /fetch failed|network|econnreset|econnrefused|enotfound|etimedout|timed out|socket/.test(lower)) {
+	} else if (err instanceof TypeError || /fetch failed|network|econnreset|econnrefused|enotfound|etimedout|timed out|socket/.test(text)) {
 		kind = "network";
-	} else if (/invalid or missing|invalid config|failed to parse|must be an? |configuration/.test(lower)) {
+	} else if (/invalid or missing|invalid config|failed to parse|must be an? |configuration/.test(text)) {
 		kind = "config";
+	} else if (openAIStreamError) {
+		// The same stream threw "no parseable response output" before its error was surfaced.
+		kind = "invalid-response";
 	}
 	return new SearchProviderError(provider, kind, message, status, err);
 }
@@ -406,9 +423,11 @@ async function searchWithResolvedProvider(
 	if (provider === "baizhi") return { ...(await searchWithBaizhi(query, options)), provider };
 	if (provider === "zai") return { ...(await searchWithZai(query, options)), provider };
 	if (provider === "valyu") return { ...(await searchWithValyu(query, options)), provider };
+	if (provider === "keenable") return { ...(await searchWithKeenable(query, options)), provider };
 	if (provider === "xcrawl") return { ...(await searchWithXCrawl(query, options)), provider };
 	if (provider === "perplexity") return { ...(await searchWithPerplexity(query, options)), provider };
 	if (provider === "searxng") return { ...(await searchWithSearXNG(query, options)), provider };
+	if (provider === "degoog") return { ...(await searchWithDegoog(query, options)), provider };
 	if (provider === "duckduckgo") return { ...(await searchWithDuckDuckGo(query, options)), provider };
 	if (provider === "kimi") return { ...(await searchWithKimi(query, options, options.extensionContext)), provider };
 	if (provider === "gemini") {
@@ -459,9 +478,11 @@ async function isResolvedProviderAvailable(provider: ResolvedSearchProvider, opt
 	if (provider === "baizhi") return isBaizhiAvailable();
 	if (provider === "zai") return isZaiAvailable();
 	if (provider === "valyu") return isValyuAvailable();
+	if (provider === "keenable") return isKeenableAvailable();
 	if (provider === "xcrawl") return isXcrawlAvailable();
 	if (provider === "perplexity") return isPerplexityAvailable();
 	if (provider === "searxng") return isSearXNGAvailable();
+	if (provider === "degoog") return isDegoogAvailable();
 	if (provider === "duckduckgo") return isDuckDuckGoAvailable();
 	if (provider === "gemini") return isGeminiApiAvailable() || await isGeminiWebOptionallyAvailable();
 	if (provider === "kimi") return isKimiSearchAvailable(options.extensionContext);
@@ -486,6 +507,7 @@ export function providerLabel(provider: ResolvedSearchProvider): string {
 	if (provider === "firecrawl") return "Firecrawl";
 	if (provider === "serpdive") return "SERPdive";
 	if (provider === "searxng") return "SearXNG";
+	if (provider === "degoog") return "degoog";
 	if (provider === "duckduckgo") return "DuckDuckGo";
 	if (provider === "kagi") return "Kagi";
 	if (provider === "bocha") return "Bocha";
@@ -503,6 +525,7 @@ export function providerLabel(provider: ResolvedSearchProvider): string {
 	if (provider === "baizhi") return "Baizhi";
 	if (provider === "zai") return "Z.ai";
 	if (provider === "valyu") return "Valyu";
+	if (provider === "keenable") return "Keenable";
 	return provider.charAt(0).toUpperCase() + provider.slice(1);
 }
 
@@ -530,7 +553,7 @@ async function searchWithProviders(
 			: await isResolvedProviderAvailable(provider, options),
 	})))).filter((entry) => entry.available).map((entry) => entry.provider);
 	if (providers.length === 0) {
-		throw new Error("No configured search provider available for provider \"all\". Parallel MCP, DuckDuckGo, Kimi, AnySearch, xAI, Mistral, Bright Data, SerpBase, SerpApi, Serper, Serply, You.com, Valyu, Baizhi, Z.ai, and XCrawl are excluded.");
+		throw new Error("No configured search provider available for provider \"all\". Parallel MCP, degoog, DuckDuckGo, Kimi, AnySearch, xAI, Mistral, Bright Data, SerpBase, SerpApi, Serper, Serply, You.com, Valyu, Baizhi, Z.ai, XCrawl, and Keenable are excluded.");
 	}
 
 	const settled = await Promise.allSettled(
@@ -825,7 +848,7 @@ export async function search(query: string, options: FullSearchOptions = {}): Pr
 		"  3. Set OPENAI_API_KEY, BRAVE_API_KEY, PARALLEL_API_KEY, TINYFISH_API_KEY, SEARCH1API_KEY, SEARCHINFINITY_API_KEY, QUERIT_API_KEY, TAVILY_API_KEY, FIRECRAWL_BASE_URL, JINA_API_KEY, SERPDIVE_API_KEY, KAGI_API_KEY, BOCHA_API_KEY, OLLAMA_API_KEY, SEARXNG_BASE_URL, EXA_API_KEY, PERPLEXITY_API_KEY, GEMINI_API_KEY, or CLOUDFLARE_API_KEY env vars\n" +
 		"  4. Set GOOGLE_GEMINI_BASE_URL with CLOUDFLARE_API_KEY for Cloudflare AI Gateway routing\n" +
 		"  5. Sign into gemini.google.com in a supported Chromium-based browser\n" +
-		"  6. Explicitly select provider: \"anysearch\" for anonymous AnySearch, \"xcrawl\" for XCrawl, \"xai\" for Grok, \"mistral\" for Mistral Conversations web search, \"brightdata\" with brightdataSerpZone for paid Bright Data SERP, \"serpbase\", \"serpapi\", \"serper\", or \"serply\" for Google SERP, \"you\" for You.com, \"zai\" for Z.ai GLM Coding Plan search, or \"valyu\" for research search"
+		"  6. Explicitly select provider: \"anysearch\" for anonymous AnySearch, \"xcrawl\" for XCrawl, \"xai\" for Grok, \"mistral\" for Mistral Conversations web search, \"brightdata\" with brightdataSerpZone for paid Bright Data SERP, \"serpbase\", \"serpapi\", \"serper\", or \"serply\" for Google SERP, \"you\" for You.com, \"zai\" for Z.ai GLM Coding Plan search, \"valyu\" for research search, or \"keenable\" for keyless Keenable search"
 	);
 }
 

package/node_modules/pi-web-access/gemini-web.ts

@@ -92,11 +92,16 @@ export async function resolveGeminiFetch(): Promise<typeof fetch> {
 export interface GeminiWebOptions {
 	youtubeUrl?: string;
 	model?: string;
-	files?: string[];
+	files?: GeminiWebFile[];
 	signal?: AbortSignal;
 	timeoutMs?: number;
 }
 
+export interface GeminiWebFile {
+	path: string;
+	mimeType: string;
+}
+
 export async function isGeminiWebAvailable(chromeProfile?: string): Promise<CookieMap | null> {
 	if (!isBrowserCookieAccessAllowed()) return null;
 
@@ -180,7 +185,7 @@ async function runGeminiWebOnce(
 	prompt: string,
 	cookieMap: CookieMap,
 	model: string,
-	files: string[] | undefined,
+	files: GeminiWebFile[] | undefined,
 	timeoutMs: number,
 	signal?: AbortSignal,
 ): Promise<GeminiWebResult> {
@@ -190,8 +195,8 @@ async function runGeminiWebOnce(
 
 	const uploaded: Array<{ id: string; name: string }> = [];
 	if (files) {
-		for (const filePath of files) {
-			uploaded.push(await uploadFile(filePath, cookieHeader, effectiveSignal));
+		for (const file of files) {
+			uploaded.push(await uploadFile(file, cookieHeader, effectiveSignal));
 		}
 	}
 
@@ -357,14 +362,14 @@ function decodeEmailEscapes(value: string): string {
 }
 
 async function uploadFile(
-	filePath: string,
+	file: GeminiWebFile,
 	cookieHeader: string,
 	signal: AbortSignal,
 ): Promise<{ id: string; name: string }> {
-	const data = readFileSync(filePath);
-	const fileName = basename(filePath);
+	const data = readFileSync(file.path);
+	const fileName = basename(file.path);
 	const boundary = "----FormBoundary" + Math.random().toString(36).slice(2);
-	const header = `--${boundary}\r\nContent-Disposition: form-data; name="file"; filename="${fileName}"\r\nContent-Type: application/octet-stream\r\n\r\n`;
+	const header = `--${boundary}\r\nContent-Disposition: form-data; name="file"; filename="${fileName}"\r\nContent-Type: ${file.mimeType}\r\n\r\n`;
 	const footer = `\r\n--${boundary}--\r\n`;
 
 	const body = Buffer.concat([
@@ -400,7 +405,7 @@ function buildFReqPayload(
 ): string {
 	const promptPayload =
 		uploaded.length > 0
-			? [prompt, 0, null, uploaded.map((file) => [[file.id, 1]])]
+			? [prompt, 0, null, uploaded.map((file) => [[file.id, 1], file.name])]
 			: [prompt];
 	const innerList = [promptPayload, null, null];
 	return JSON.stringify([null, JSON.stringify(innerList)]);