Files
OmniRoute/open-sse/services/browserPool.ts
Diego Rodrigues de Sa e Souza 7db430a352 Release v3.8.14 (#3340)
* chore(release): open v3.8.14 development cycle

Version bump 3.8.13 -> 3.8.14 (root + electron + open-sse + openapi + lockfiles).
Seed the v3.8.14 changelog with the four post-tag hotfixes that shipped to
Docker/Electron in v3.8.13 but missed the immutable npm 3.8.13 (#3336 SSRF /
CodeQL #323, #3334/#3335/#3339 Electron packaging). i18n CHANGELOG mirrors get
the in-progress placeholder section.

* feat: add per-provider custom headers support for OpenAI/Anthropic-compatible nodes (#3338)

Integrated into release/v3.8.14

* fix: Kiro Builder ID token import fails with Bad credentials (#3333)

Integrated into release/v3.8.14 — adds Builder ID cached-creds + OIDC refresh path for Kiro token import, with regression tests (#3333).

* Improve code quality: auto-pr/docstrings-1780792063 (#3337)

Integrated into release/v3.8.14 — docstring for context analytics route re-export.

* fix(catalog): remove minimaxai/minimax-m3 from NVIDIA NIM tier (404 upstream) (#3329) (#3341)

NVIDIA NIM does not host minimaxai/minimax-m3 — every request returns
404 page not found, while sibling minimaxai/minimax-m2.7 on the same provider
works. Advertising a model that 404s is a catalog bug; remove it from the nvidia
tier (it remains on the tiers that actually serve MiniMax M3). Re-add only once
NVIDIA serves it.

Co-authored-by: mikmaneggahommie <mikmaneggahommie@users.noreply.github.com>

* fix(cli): write OpenCode config to ~/.config on all platforms incl. Windows (#3330) (#3343)

resolveOpencodeConfigDir used %APPDATA% on Windows, but OpenCode reads its
config from XDG ~/.config/opencode/ on every platform (on Windows:
%USERPROFILE%\.config\opencode\, NOT %APPDATA%). So a Windows user who
configured OpenCode via the dashboard had the file written where OpenCode never
looks — it silently had no effect.

Use the XDG path (XDG_CONFIG_HOME || ~/.config) unconditionally. Update the UI
note + route JSDoc, and flip the three tests that encoded the old %APPDATA%
behavior (t40 per-platform + card-note, cli-runtime-extended getCliConfigPaths).

Co-authored-by: abdulkadirozyurt <abdulkadirozyurt@users.noreply.github.com>

* fix(proxy): make auto-selection fallback opt-in (#3332) (#3344)

selectWorkingProxyFallback (Step 11 of resolveProxyForConnection) listed ALL
registry proxies, ignoring assignments and per-connection proxy_enabled, and
returned the first working one with level:'autoSelect'. So a single proxy added
to the registry silently became a global fallback for every connection's traffic.

Gate it behind a new PROXY_AUTO_SELECT_ENABLED feature flag (default off): the
fallback now no-ops unless the operator opts in. No registry proxy becomes a
silent global default anymore.

Co-authored-by: hertznsk <hertznsk@users.noreply.github.com>

* fix(sse): treat MiniMax M3 as multimodal so vision isn't stripped (#3328) (#3342)

MiniMax M3 via the opencode provider (oc/minimax-m3-free) appeared blind:
image inputs didn't reach the model, while the same model in Cline could
see them. Verified empirically that MiniMax M3 on the opencode upstream IS
multimodal -- a base64 image is described correctly (it returns 403 only
for remote image URLs, which it doesn't accept).

Root cause: OmniRoute treated MiniMax M3 as a non-vision model in two
places, so when compression was active the image was replaced with a text
placeholder before dispatch:
- compression's modelSupportsVision() heuristic (lite.ts) only matched
  gpt-4/4o/claude-3/gemini/vision -- minimax was absent -> replaceImageUrls
  stripped the image.
- the opencode minimax-m3-free catalog entry lacked supportsVision, so the
  combo vision-capability gate could also exclude/mishandle it.

Add 'minimax-m3' to the vision heuristic and supportsVision: true to the
opencode minimax-m3-free entry. TDD: a failing-then-passing test in
compression/lite.test.ts proves replaceImageUrls now keeps images for
minimax-m3 ids, plus a registry assertion mirroring the #2822 qwen test.

Reported-by: @mikmaneggahommie

* docs(i18n): translate 25 core documentation files to Indonesian (#3348)

Integrated into release/v3.8.14 — Indonesian i18n docs.

* fix(review): resolve /review-reviews battery findings (LEDGER-1..11) on v3.8.14 (#3350)

Integrated into release/v3.8.14 — /review-reviews battery hardening (LEDGER-1..11) for #3338 custom-headers + #3333 kiro, plus cycle-test drift fixes (#3329/#3330/#3332).

* fix(provider-proxy): honor per-account proxy toggles (#3349)

Integrated into release/v3.8.14 — honor per-account proxy toggles + auto-fallback opt-in via PROXY_AUTO_SELECT_ENABLED.

* fix(dashboard): remove duplicate Distribute Proxies button on provider page (#3352)

* fix(providers): reduce proxy label noise (#3346)

Integrated into release/v3.8.14 — reduce proxy label noise + a11y (aria-label/sr-only).

* fix(duckduckgo): restore bare Response contract and rebase onto release/v3.8.14 (#3323)

Integrated into release/v3.8.14 — browser-backed cookie providers (duckduckgo/claude-web) with restored executor contract + unit tests.

* fix(noauth): expose only usable model aliases (#3345)

Integrated into release/v3.8.14 — noauth usable-alias filtering + registry alias plumbing (veo-free).

* fix(dashboard): stop infinite config-load loop on Hermes Agent detail page (#3353)

* fix(electron): tree-kill the server on exit/update to release the omniroute.exe lock (#3347) (#3354)

* chore(release): finalize v3.8.14 changelog + clear release-gate drift

- CHANGELOG: finalize the v3.8.14 section (date, full New Features/Bug Fixes/
  Maintenance coverage of all 16 cycle commits, Contributors hall of 12).
- docs: document OMNIROUTE_BROWSER_POOL + WEB_COOKIE_USE_BROWSER (#3323) in
  .env.example + ENVIRONMENT.md; regenerate the id/llm.txt strict mirror (#3348
  had translated it; llm.txt mirrors must match root).
- test(proxy-fetch): #3323 made tlsClient.available a computed getter — stub it
  via Object.defineProperty instead of assignment (5 tests were red on the base).

* fix(translator): coerce Gemini functionDeclaration parameters to an OBJECT schema (#3357) (#3360)

* fix(gemini): resolve truncation/suppression of false positive textual tool call markers in backticks (#3358)

Integrated into release/v3.8.14 — Gemini/Antigravity textual tool-call marker normalization (no false-positive suppression + split-chunk buffering).

* docs(changelog): add #3358 Gemini textual tool-call normalization to v3.8.14

* fix(dashboard): surface real analytics error instead of generic placeholder (#3356) (#3361)

The Analytics page discarded the server's error body on a non-OK response and
rendered a generic "An error occurred", so users (and maintainers) could not see
why /api/usage/analytics 500'd after an upgrade. Now the route returns the real
reason via buildErrorBody (sanitized, Hard Rule #12) and the page surfaces it via
a new readFetchErrorMessage helper that handles both the OpenAI-style and legacy
error shapes.

Reported-by: @superti4r

---------

Co-authored-by: PizzaV <103120356+pizzav-xyz@users.noreply.github.com>
Co-authored-by: Someres <168349709+quanturbo@users.noreply.github.com>
Co-authored-by: Dong Mengzhe <154944819+Lang-Qiu@users.noreply.github.com>
Co-authored-by: mikmaneggahommie <mikmaneggahommie@users.noreply.github.com>
Co-authored-by: abdulkadirozyurt <abdulkadirozyurt@users.noreply.github.com>
Co-authored-by: hertznsk <hertznsk@users.noreply.github.com>
Co-authored-by: Krisna Santosa <54174372+KrisnaSantosa15@users.noreply.github.com>
Co-authored-by: Randi <55005611+rdself@users.noreply.github.com>
Co-authored-by: Wilson <pedbookmed@gmail.com>
Co-authored-by: Paijo <14921983+oyi77@users.noreply.github.com>
Co-authored-by: Ardem2025 <ardemb22@gmail.com>
2026-06-07 07:20:02 -03:00

381 lines
11 KiB
TypeScript

/**
* browserPool.ts — Shared stealth browser pool for web-cookie providers.
*
* The DuckDuckGo VQD challenge and Claude web's Cloudflare Turnstile both
* validate values that only a real browser can produce (DOM layout
* measurements like offsetWidth/Height, getBoundingClientRect,
* getComputedStyle, iframe contentWindow probes). Plain Node fetch + a
* VM-stubs solver structurally runs the JS but cannot match those values,
* so the server rejects the request.
*
* This pool keeps one Chromium instance warm and serves "browser contexts"
* (one per provider) on demand. Each context owns one or more pages; the
* caller is expected to be polite (one page per request, close on done).
*
* The pool prefers `cloakbrowser` (npm) when available — its binary-level
* fingerprint patches (--fingerprint-timezone, --fingerprint-locale, and
* dozens more) are the only thing that gets past DuckDuckGo's anti-bot
* in this environment. Falls back to plain `playwright` if cloakbrowser
* is not installed; the fallback works for Claude web (which only needs
* valid cookies) but not for DDG's VQD challenge.
*
* Opt-in: pool only launches Chromium when an executor explicitly asks
* for a context, so users who never use the browser-backed path pay zero
* startup cost. Set OMNIROUTE_BROWSER_POOL=off to fully disable.
*/
import { Buffer } from "node:buffer";
type Browser = import("playwright").Browser;
type BrowserContext = import("playwright").BrowserContext;
type Page = import("playwright").Page;
export interface BrowserPoolContextOptions {
cookieDomain: string;
cookieString?: string | null;
warmupUrl?: string | null;
userAgent?: string;
locale?: string;
timezone?: string;
preferCloakbrowser?: boolean;
}
export interface PooledContext {
id: string;
context: BrowserContext;
warmupPage: Page | null;
lastUsed: number;
isStealth: boolean;
}
interface PoolState {
browser: Browser | null;
contexts: Map<string, PooledContext>;
pendingContexts: Map<string, Promise<PooledContext>>;
launching: Promise<Browser> | null;
lastActivity: number;
idleTimer: NodeJS.Timeout | null;
evictTimer: NodeJS.Timeout | null;
cloakLaunch: ((opts: unknown) => Promise<Browser>) | null;
cloakLaunchResolved: boolean;
}
const POOL_IDLE_TIMEOUT_MS = 5 * 60 * 1000;
const CONTEXT_TTL_MS = 10 * 60 * 1000; // 10 min — evict stale contexts
const EVICT_INTERVAL_MS = 60 * 1000; // check every 60s
const DEFAULT_USER_AGENT =
"Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/146.0.0.0 Safari/537.36";
const state: PoolState = {
browser: null,
contexts: new Map(),
pendingContexts: new Map(),
launching: null,
lastActivity: 0,
idleTimer: null,
evictTimer: null,
cloakLaunch: null,
cloakLaunchResolved: false,
};
async function resolveCloakLaunch(): Promise<((opts: unknown) => Promise<Browser>) | null> {
if (state.cloakLaunchResolved) return state.cloakLaunch;
state.cloakLaunchResolved = true;
try {
const mod = (await import("cloakbrowser")) as unknown as {
launch?: (opts: unknown) => Promise<Browser>;
};
state.cloakLaunch = mod.launch ?? null;
} catch {
state.cloakLaunch = null;
}
return state.cloakLaunch;
}
function isPoolEnabled(): boolean {
const flag = process.env.OMNIROUTE_BROWSER_POOL;
if (flag === undefined) return true;
return flag !== "off" && flag !== "0" && flag !== "false";
}
function resetIdleTimer(): void {
if (state.idleTimer) clearTimeout(state.idleTimer);
state.idleTimer = setTimeout(() => {
void shutdownPool("idle-timeout");
}, POOL_IDLE_TIMEOUT_MS);
state.idleTimer.unref?.();
}
function evictStaleContexts(): void {
const now = Date.now();
for (const [key, pooled] of state.contexts) {
if (now - pooled.lastUsed > CONTEXT_TTL_MS) {
console.log("[BrowserPool] Evicted stale context:", key, "(idle", ((now - pooled.lastUsed) / 1000).toFixed(0) + "s)");
state.contexts.delete(key);
pooled.context.close().catch(() => {});
}
}
if (state.contexts.size === 0 && !state.launching) {
void shutdownPool("all-contexts-evicted");
}
}
function startEvictTimer(): void {
if (state.evictTimer) clearInterval(state.evictTimer);
state.evictTimer = setInterval(() => evictStaleContexts(), EVICT_INTERVAL_MS);
state.evictTimer.unref?.();
}
async function launchBrowser(): Promise<Browser> {
if (state.browser) return state.browser;
if (state.launching) return state.launching;
state.launching = (async () => {
const cloakLaunch = await resolveCloakLaunch();
let browser: Browser;
if (cloakLaunch) {
browser = await cloakLaunch({
headless: true,
args: ["--no-sandbox", "--disable-dev-shm-usage"],
});
} else {
// Fallback: plain Playwright. Works for Claude web (cookie-only
// auth) but DDG's VQD challenge will detect this Chromium build.
const { chromium } = await import("playwright");
browser = await chromium.launch({
headless: true,
args: [
"--no-sandbox",
"--disable-dev-shm-usage",
"--disable-blink-features=AutomationControlled",
],
});
}
state.browser = browser;
state.launching = null;
return browser;
})();
try {
return await state.launching;
} catch (err) {
state.launching = null;
throw err;
}
}
function parseCookieString(
raw: string,
domain: string
): Array<{
name: string;
value: string;
domain: string;
path: string;
expires: number;
httpOnly: boolean;
secure: boolean;
sameSite: "Lax" | "Strict" | "None";
}> {
return raw
.split(";")
.map((p) => p.trim())
.filter(Boolean)
.map((pair) => {
const eq = pair.indexOf("=");
if (eq < 0) return null;
const name = pair.slice(0, eq).trim();
const value = pair.slice(eq + 1).trim();
if (!name || !value) return null;
return {
name,
value,
domain: domain.startsWith(".") ? domain : `.${domain}`,
path: "/",
expires: -1,
httpOnly: false,
secure: true,
sameSite: "Lax" as const,
};
})
.filter(Boolean) as Array<{
name: string;
value: string;
domain: string;
path: string;
expires: number;
httpOnly: boolean;
secure: boolean;
sameSite: "Lax" | "Strict" | "None";
}>;
}
export async function acquireBrowserContext(
key: string,
options: BrowserPoolContextOptions
): Promise<PooledContext> {
if (!isPoolEnabled()) {
throw new Error(
"browserPool: OMNIROUTE_BROWSER_POOL=off — context requested but pool is disabled"
);
}
const existing = state.contexts.get(key);
if (existing) {
existing.lastUsed = Date.now();
state.lastActivity = Date.now();
resetIdleTimer();
return existing;
}
// Dedup concurrent creations for the same key
const pending = state.pendingContexts.get(key);
if (pending) return pending;
const createPromise = (async (): Promise<PooledContext> => {
const browser = await launchBrowser();
const isStealth = state.cloakLaunch !== null;
const context = await browser.newContext({
userAgent: options.userAgent || DEFAULT_USER_AGENT,
locale: options.locale || "en-US",
timezoneId: options.timezone || "America/New_York",
viewport: { width: 1280, height: 800 },
});
if (options.cookieString) {
const cookies = parseCookieString(options.cookieString, options.cookieDomain);
if (cookies.length > 0) {
await context.addCookies(cookies);
}
}
let warmupPage: Page | null = null;
if (options.warmupUrl) {
try {
warmupPage = await context.newPage();
await warmupPage.goto(options.warmupUrl, {
waitUntil: "domcontentloaded",
timeout: 30000,
});
// Give the warmup a moment for the upstream's status/auth/country
// JSON endpoints to fire. Without this, the first chat request would
// pay the warmup cost on the hot path.
await new Promise((r) => setTimeout(r, 1500));
} catch (err) {
try {
await warmupPage?.close();
} catch {
/* ignore */
}
warmupPage = null;
void err;
}
}
// Guard: if shutdownPool() ran while we were creating this context,
// the browser we obtained is now closed. Close our temp context and
// throw so the caller knows to retry.
if (state.browser !== browser) {
await context.close().catch(() => {});
if (warmupPage) {
await warmupPage.close().catch(() => {});
}
throw new Error("Pool shut down during context creation");
}
const pooled: PooledContext = {
id: key,
context,
warmupPage,
lastUsed: Date.now(),
isStealth,
};
state.contexts.set(key, pooled);
state.lastActivity = Date.now();
resetIdleTimer();
startEvictTimer();
return pooled;
})();
state.pendingContexts.set(key, createPromise);
createPromise
.then(() => state.pendingContexts.delete(key))
.catch(() => state.pendingContexts.delete(key));
return createPromise;
}
export async function openPage(pooled: PooledContext): Promise<Page> {
return pooled.context.newPage();
}
export async function releaseBrowserContext(key: string): Promise<void> {
const pooled = state.contexts.get(key);
if (!pooled) return;
state.contexts.delete(key);
try {
await pooled.context.close();
} catch {
/* ignore */
}
if (state.contexts.size === 0) {
await shutdownPool("last-context-closed");
}
}
export async function shutdownPool(reason: string): Promise<void> {
if (state.idleTimer) {
clearTimeout(state.idleTimer);
state.idleTimer = null;
}
if (state.evictTimer) {
clearInterval(state.evictTimer);
state.evictTimer = null;
}
state.pendingContexts.clear();
for (const [key, pooled] of state.contexts) {
try {
await pooled.context.close();
} catch {
/* ignore */
}
state.contexts.delete(key);
}
if (state.browser) {
try {
await state.browser.close();
} catch {
/* ignore */
}
state.browser = null;
}
state.lastActivity = Date.now();
// Avoid unused-parameter lint: log reason via debug if anyone hooks
// process.on('exit') and prints state.
void reason;
}
export function getBrowserPoolStatus(): {
enabled: boolean;
contexts: number;
browserRunning: boolean;
stealthAvailable: boolean;
lastActivityAgoMs: number;
} {
return {
enabled: isPoolEnabled(),
contexts: state.contexts.size,
browserRunning: state.browser !== null,
stealthAvailable: state.cloakLaunch !== null,
lastActivityAgoMs: state.lastActivity === 0 ? -1 : Date.now() - state.lastActivity,
};
}
export async function readPageResponseBody(
response: import("playwright").Response
): Promise<{ status: number; headers: Record<string, string>; body: Buffer }> {
const headers: Record<string, string> = {};
for (const [name, value] of Object.entries(response.headers())) {
headers[name] = value;
}
const body = await response.body();
return { status: response.status(), headers, body: Buffer.from(body) };
}