// RootNode Academy — Gateway (Phase 2B revision). // // The only student-facing path to disposable Lab environments. Phase 2B changes from Phase 2A: // - No longer calls the Provisioner directly. All environment orchestration (student launch/ // reset, admin prepare/destroy/reset) is consolidated in the Platform API // (platform/api) — this Gateway proxies /api/environment/*, /api/admin/*, and everything // else to it, the same way it proxied to the old "identity" instance in Phase 2A. // - Lab ownership is no longer decided by a client-supplied ?sid= query param (Phase 2A // pilot mechanism, trivially forgeable). It's derived from the browser's rn_session cookie, // resolved server-side via the Platform API's session store // (GET /api/session/whoami) — nothing client-controlled (URL, token, query param, header) // can change who the Gateway thinks is asking. See // docs/PHASE2B_ADMIN_AND_PLATFORM_DATA.md §3. // - Ownership is verified on EVERY proxied Lab request (HTTP and WebSocket), not just the // initial navigation — a short-TTL cache keeps this cheap. const http = require('http'); const express = require('express'); const httpProxy = require('http-proxy'); const { provisioner } = require('./provisioner-client'); // still used ONLY for the token->environment lookup a Lab request needs; all mutating calls (provision/reset/destroy) now live in the Platform API. const envCache = require('./env-cache'); // Phase 4.1 — the ONE place that knows how to reach a Lab (direct Docker-internal address, or // via that lab_host's Lab Host Router). Every proxy.web/proxy.ws/labRequest call below uses // labTarget's output instead of building `http://${env.internal_ip}:${env.internal_port}` // itself. See lab-endpoint.js's header comment and docs/PHASE4_1_PREPRODUCTION_HARDENING.md §1. const { labTarget, labRequest } = require('./lab-endpoint'); const PORT = Number(process.env.GATEWAY_PORT) || 4200; const PLATFORM_API_URL = process.env.PLATFORM_API_URL || 'http://localhost:4300'; const STUDENT_COOKIE = 'rn_session'; const STUDENT_CSRF_COOKIE = 'rn_csrf_s'; // Phase 3A — see relayScenarioOutcome for why the Gateway needs this too // Routing hint only — NOT a security check by itself (ownership is always re-verified via the // session, see checkOwnership/resolveOwnedRunningEnvironment). Its only job is disambiguating // "this unprefixed request came from inside the Lab UI" from "this came from the dashboard," // which "does the student currently have a running env" cannot do — that's true on the // dashboard too, and naively using it hijacked EVERY unmatched dashboard request (including // GET /api/session/whoami) to the student's Lab container once one existed. Found by testing. const LAB_ACTIVE_COOKIE = 'rn_lab_active'; // Phase 3A: shortened from 5000ms so a revoked/expired session stops working here within a // tighter, testable bound (see platform/test/e2e-phase3a.js "revoked mid-use" case). const SESSION_CACHE_TTL_MS = 2000; const app = express(); // Deliberately no express.json() globally — see the Phase 2A note this carries forward: // body-parsing middleware breaks http-proxy's re-streaming of proxied POST bodies. const proxy = httpProxy.createProxyServer({ ws: true }); proxy.on('error', (err, req, res) => { console.error('[proxy error]', err.message); if (res && !res.headersSent && res.writeHead) { res.writeHead(502, { 'Content-Type': 'application/json' }); res.end(JSON.stringify({ error: 'unreachable' })); } else if (res && res.destroy) { res.destroy(); } }); // GET /healthz — the Gateway's own liveness check, answered locally without proxying anywhere // (Phase 4, docs/OBSERVABILITY.md §"uptime checks"). Deliberately doesn't depend on Platform // API/Provisioner being reachable, so it can distinguish "Gateway itself is down" from // "something behind it is down" for alerting. app.get('/healthz', (req, res) => res.json({ ok: true })); function parseCookies(req) { const header = req.headers.cookie; if (!header) return {}; return Object.fromEntries( header.split(';').map((p) => p.trim().split('=')).filter((p) => p.length === 2).map(([k, v]) => [k, decodeURIComponent(v)]) ); } // ── Session resolution — replaces Phase 2A's ?sid= entirely ──────────────────────────────── const sessionCache = new Map(); // cookie value -> { studentId, expiresAt } async function resolveSessionStudentId(req) { const token = parseCookies(req)[STUDENT_COOKIE]; if (!token) return null; const cached = sessionCache.get(token); if (cached && cached.expiresAt > Date.now()) return cached.studentId; try { const res = await fetch(`${PLATFORM_API_URL}/api/session/whoami`, { headers: { Cookie: `${STUDENT_COOKIE}=${token}` } }); if (res.status !== 200) return null; const { student_id } = await res.json(); sessionCache.set(token, { studentId: student_id, expiresAt: Date.now() + SESSION_CACHE_TTL_MS }); return student_id; } catch { return null; } } async function resolveByToken(token) { const cached = envCache.get(token); if (cached) return cached; const { status, body } = await provisioner.getEnvironmentByToken(token); if (status !== 200) return null; envCache.set(token, body); return body; } // internal_ip is set by the Provisioner as soon as a container exists, regardless of routing // mode (it's a Docker-internal fact, independent of how the Gateway will reach it) — so this // check stays valid unchanged for both 'direct' and 'router' lab_hosts. function isReachable(env) { return env && env.status === 'running' && env.internal_ip; } function sendLabError(res, status, error, message) { res.status(status).type('html').send(` RootNode Academy

${message}

(${error})

`); } // Shared ownership check: valid session AND session's student owns this environment. // Used identically by the explicit /lab/:token route, the cookie-fallback route, and the // WebSocket upgrade handler — one rule, everywhere a Lab request could be routed. async function checkOwnership(req, env) { if (!env) return { ok: false, reason: 'invalid_token' }; const studentId = await resolveSessionStudentId(req); if (!studentId) return { ok: false, reason: 'no_session' }; if (studentId !== env.student_id) return { ok: false, reason: 'forbidden' }; return { ok: true, studentId }; } // ── Progress relay (unchanged mechanism from Phase 2A, now passes environment_id + // validation_count through so Platform API can log a real scenario_attempts row) ────────── async function relayScenarioOutcome({ env, token, req, res, scenarioPath, extractOutcome }) { if (!isReachable(env)) return sendLabError(res, 503, 'lab_unavailable', 'Your lab is not currently running.'); const labRes = await labRequest(env, token, scenarioPath, { method: 'POST', headers: { 'Content-Type': 'application/json' }, body: JSON.stringify(req.body || {}), }); const body = await labRes.json().catch(() => ({})); res.status(labRes.status).json(body); const passed = extractOutcome(body); if (passed !== null) { const validationCount = Array.isArray(body.checks) ? body.checks.length : undefined; const cookies = parseCookies(req); const token = cookies[STUDENT_COOKIE]; const csrf = cookies[STUDENT_CSRF_COOKIE] || ''; // Phase 3A: /api/progress/record is CSRF-protected like every other student mutation. The // double-submit check needs BOTH the rn_csrf_s cookie AND a matching X-CSRF-Token header on // the request Platform API actually sees — so both are forwarded here, not just the header // (an earlier version of this fix sent only the header, which still 403'd). fetch(`${PLATFORM_API_URL}/api/progress/record`, { method: 'POST', headers: { 'Content-Type': 'application/json', Cookie: `${STUDENT_COOKIE}=${token}; ${STUDENT_CSRF_COOKIE}=${csrf}`, 'X-CSRF-Token': csrf, }, body: JSON.stringify({ scenario_id: req.params.scenarioId, status: passed ? 'completed' : 'in_progress', environment_id: env.environment_id, validation_count: validationCount, }), }).then((r) => { if (!r.ok) console.error('[progress relay] Platform API rejected the update:', r.status); }) .catch((e) => console.error('[progress relay] failed:', e.message)); } } app.post('/lab/:token/api/scenarios/:scenarioId/validate', express.json(), async (req, res) => { const env = await resolveByToken(req.params.token); const own = await checkOwnership(req, env); if (!own.ok) return sendLabError(res, own.reason === 'invalid_token' ? 404 : 403, own.reason, 'This lab is not available to you.'); await relayScenarioOutcome({ env, token: req.params.token, req, res, scenarioPath: `/api/scenarios/${req.params.scenarioId}/validate`, extractOutcome: (body) => (typeof body.passed === 'boolean' ? body.passed : null), }); }); app.post('/lab/:token/api/scenarios/:scenarioId/answer', express.json(), async (req, res) => { const env = await resolveByToken(req.params.token); const own = await checkOwnership(req, env); if (!own.ok) return sendLabError(res, own.reason === 'invalid_token' ? 404 : 403, own.reason, 'This lab is not available to you.'); await relayScenarioOutcome({ env, token: req.params.token, req, res, scenarioPath: `/api/scenarios/${req.params.scenarioId}/answer`, extractOutcome: (body) => (typeof body.correct === 'boolean' ? body.correct : null), }); }); // ── Phase 3B learning relay — /start and /hint are pure Platform API operations, unlike // validate/answer above (those call the Lab first, then relay the outcome). The Lab has no // concept of attempts, hints-as-tracked-state, or assessments; these forward straight through // to Platform API with the browser's own session + CSRF token (same forwarding requirement // documented on relayScenarioOutcome above — both cookie AND header, not just the header). ──── async function relayToPlatformApi(req, res, platformPath, extraBody) { const cookies = parseCookies(req); const token = cookies[STUDENT_COOKIE]; const csrf = cookies[STUDENT_CSRF_COOKIE] || ''; try { const platformRes = await fetch(`${PLATFORM_API_URL}${platformPath}`, { method: 'POST', headers: { 'Content-Type': 'application/json', Cookie: `${STUDENT_COOKIE}=${token}; ${STUDENT_CSRF_COOKIE}=${csrf}`, 'X-CSRF-Token': csrf, }, body: JSON.stringify({ ...(req.body || {}), ...extraBody }), }); const body = await platformRes.json().catch(() => ({})); res.status(platformRes.status).json(body); } catch (e) { res.status(502).json({ error: 'platform_unreachable' }); } } app.post('/lab/:token/api/scenarios/:scenarioId/start', express.json(), async (req, res) => { const env = await resolveByToken(req.params.token); const own = await checkOwnership(req, env); if (!own.ok) return sendLabError(res, own.reason === 'invalid_token' ? 404 : 403, own.reason, 'This lab is not available to you.'); await relayToPlatformApi(req, res, `/api/scenarios/${req.params.scenarioId}/start`, { environment_id: env.environment_id }); }); app.post('/lab/:token/api/scenarios/:scenarioId/hint', express.json(), async (req, res) => { const env = await resolveByToken(req.params.token); const own = await checkOwnership(req, env); if (!own.ok) return sendLabError(res, own.reason === 'invalid_token' ? 404 : 403, own.reason, 'This lab is not available to you.'); await relayToPlatformApi(req, res, `/api/scenarios/${req.params.scenarioId}/hint`, { environment_id: env.environment_id }); }); // Strips `hints` from the Lab's own scenario-detail response before it reaches the browser — // the Lab backend still returns the full array (it has no student/attempt concept and was // never modified), but hints must only be revealed progressively through the endpoint above. // The Lab image itself stays byte-for-byte unmodified; this is a response rewrite at the // Gateway, not a Lab change. app.get('/lab/:token/api/scenarios/:scenarioId', async (req, res) => { const env = await resolveByToken(req.params.token); const own = await checkOwnership(req, env); if (!own.ok) return sendLabError(res, own.reason === 'invalid_token' ? 404 : 403, own.reason, 'This lab is not available to you.'); if (!isReachable(env)) return sendLabError(res, 503, 'lab_unavailable', 'Your lab is not currently running.'); try { const labRes = await labRequest(env, req.params.token, `/api/scenarios/${req.params.scenarioId}`); const body = await labRes.json().catch(() => ({})); const { hints, ...withoutHints } = body; res.status(labRes.status).json({ ...withoutHints, hint_count: Array.isArray(hints) ? hints.length : 0 }); } catch (e) { res.status(502).json({ error: 'lab_unreachable' }); } }); // ── /lab/:token/* — HTTP proxy into the student's own Lab, prefix-stripped ───────────────── app.use('/lab/:token', async (req, res, next) => { const { token } = req.params; const env = await resolveByToken(token); if (!env) return sendLabError(res, 404, 'invalid_token', 'This lab link is invalid or has expired.'); const own = await checkOwnership(req, env); if (!own.ok) { if (own.reason === 'no_session') return sendLabError(res, 401, 'no_session', 'Please log in.'); return sendLabError(res, 403, 'forbidden', 'This lab does not belong to your session.'); } if (env.status === 'provisioning' || env.status === 'requested' || env.status === 'resetting') { return sendLabError(res, 202, 'provisioning', 'Your lab is still starting. Try again in a moment.'); } if (!isReachable(env)) { return sendLabError(res, 503, 'lab_unavailable', 'Your lab is not currently running. Try recreating it.'); } // Marks this browser as "currently inside a lab page" — the only thing the unprefixed // fallback route below uses to decide whether to even attempt Lab routing. Path-scoped // isn't useful here (the whole point is covering root-relative /api/* and /shell-ws calls // the Lab UI makes), so Path=/ with a short lifetime instead — long enough to cover a work // session, short enough that leaving the lab page eventually stops privileging Lab-routing // for this browser. res.append('Set-Cookie', `${LAB_ACTIVE_COOKIE}=1; Path=/; HttpOnly; SameSite=Lax; Max-Age=28800`); req.url = req.url.replace(/^\/lab\/[^/]+/, '') || '/'; const target = labTarget(env, token); proxy.web(req, res, { target: target.url, agent: target.agent, headers: target.headers, changeOrigin: target.changeOrigin }); }); // Resolves "the environment this browser's session currently owns and can reach," for the // unprefixed fallback case (no /lab/:token in the URL — the existing Lab UI's own root-relative // calls once a browser is already inside a lab session). Shared by both the HTTP fallback and // the WS upgrade handler so there's exactly one place this logic lives. Gated on // LAB_ACTIVE_COOKIE first — see its definition above for why "has a running env" alone isn't // a safe signal (it's true on the dashboard too). // Returns { env, token } (not just env) — Phase 4.1's labTarget needs the routing token // alongside the environment row for router-mode lab_hosts, and this is the only place that // derives the token from launch_path rather than the URL, so it has to hand both back. async function resolveOwnedRunningEnvironment(req) { const cookies = parseCookies(req); if (!cookies[LAB_ACTIVE_COOKIE]) return null; const sessionToken = cookies[STUDENT_COOKIE]; if (!sessionToken) return null; const studentId = await resolveSessionStudentId(req); if (!studentId) return null; try { const envRes = await fetch(`${PLATFORM_API_URL}/api/environment/status`, { headers: { Cookie: `${STUDENT_COOKIE}=${sessionToken}` } }); const envInfo = await envRes.json().catch(() => null); if (envInfo?.status !== 'running' || !envInfo.launch_path) return null; const lookupToken = envInfo.launch_path.split('/')[2]; const env = await resolveByToken(lookupToken); return isReachable(env) && env.student_id === studentId ? { env, token: lookupToken } : null; } catch { return null; } } // Paths the unmodified Lab UI actually calls unprefixed (see backend/server.js's route list // and frontend/src/App.jsx et al.). LAB_ACTIVE_COOKIE alone isn't a safe enough signal — it // persists for the rest of the browser's session once a student has EVER visited their lab, // which would otherwise keep misrouting dashboard calls (GET /api/session/whoami, // /api/environment/*, /api/admin/*, /api/roster*, ...) to the Lab long after the student left // it. Found by testing. Everything NOT matching this allowlist always goes to the Platform // API, regardless of LAB_ACTIVE_COOKIE. const LAB_OWNED_PATHS = [/^\/api\/scenarios\b/, /^\/api\/bundles\b/, /^\/api\/sessions\b/, /^\/api\/cache\b/, /^\/api\/health\b/, /^\/shell-ws$/, /^\/assets\//]; function looksLikeLabPath(pathname) { return LAB_OWNED_PATHS.some((re) => re.test(pathname)); } // ── Default fallback — unprefixed Lab traffic for a browser already "inside" a lab session, // or the Platform API (student dashboard, admin portal, roster/curriculum/progress) ──────── app.use(async (req, res) => { if (looksLikeLabPath(req.path)) { const owned = await resolveOwnedRunningEnvironment(req); if (owned) { const target = labTarget(owned.env, owned.token); return proxy.web(req, res, { target: target.url, agent: target.agent, headers: target.headers, changeOrigin: target.changeOrigin }); } } proxy.web(req, res, { target: PLATFORM_API_URL }); }); // ── WebSocket upgrade — same ownership rule as HTTP, applied to the /shell-ws handshake ──── const server = http.createServer(app); server.on('upgrade', async (req, socket, head) => { const labMatch = req.url.match(/^\/lab\/([^/]+)(\/.*)?$/); if (labMatch) { const [, token, rest] = labMatch; const env = await resolveByToken(token); const own = await checkOwnership(req, env); if (!own.ok || !isReachable(env)) return socket.destroy(); req.url = rest || '/'; const target = labTarget(env, token); return proxy.ws(req, socket, head, { target: target.url, agent: target.agent, headers: target.headers, changeOrigin: target.changeOrigin }); } // Unprefixed WS (the existing Lab UI's /shell-ws, hit while "inside" a lab session). const owned = await resolveOwnedRunningEnvironment(req); if (owned) { const target = labTarget(owned.env, owned.token); return proxy.ws(req, socket, head, { target: target.url, agent: target.agent, headers: target.headers, changeOrigin: target.changeOrigin }); } socket.destroy(); }); server.listen(PORT, () => console.log(`Gateway listening on :${PORT}`));