@faable/faable 2.1.0 → 2.2.0

package/dist/api/FaableApi.js

@@ -252,9 +252,9 @@ class FaableApi {
     async addAppWafRule(app_id, params) {
         return data(this.client.post(`/app/${app_id}/waf/rules`, params));
     }
-    async removeAppWafRule(app_id, pattern) {
+    async removeAppWafRule(app_id, params) {
         return data(this.client.delete(`/app/${app_id}/waf/rules`, {
-            data: { pattern }
+            data: params
         }));
     }
 }

package/dist/commands/deploy/waf/add_rule.js

@@ -1,24 +1,83 @@
 import { log } from '../../../log.js';
 
+/** How a rule reads in a sentence, for the confirmation the user gets back. */
+const describe_selector = (s) => {
+    if (s.pattern && s.user_agent) {
+        return `requests for ${s.pattern} from user-agents matching ${s.user_agent}`;
+    }
+    if (s.user_agent)
+        return `requests from user-agents matching ${s.user_agent}`;
+    return `requests for ${s.pattern}`;
+};
+/** Does this returned rule carry everything we asked for? */
+const rule_matches_request = (rule, want, action) => rule.action === action &&
+    (rule.pattern ?? '') === (want.pattern ?? '') &&
+    (rule.user_agent ?? '') === (want.user_agent ?? '');
 /**
  * Shared handler for `block` and `sink` — the two differ only in the action
  * they store and in what the edge answers, so the flow (resolve app, write,
- * explain, tell the user how to check) lives here once.
+ * verify, explain, tell the user how to check) lives here once.
  */
 const add_rule = async (opts) => {
-    const { api, app_id, app_name, app_url, pattern, action, description } = opts;
-    await api.addAppWafRule(app_id, { pattern, action, description });
+    const { api, app_id, app_name, app_url, pattern, user_agent, action, description, force } = opts;
+    const want = { pattern, user_agent };
+    const waf = await api.addAppWafRule(app_id, {
+        pattern,
+        user_agent,
+        action,
+        description,
+        force
+    });
+    // Read the rule back before telling the user it worked.
+    //
+    // An older Faable server has no `user_agent` on this endpoint and drops it
+    // silently, storing the path half on its own — which turns "block /login for
+    // this bot" into "block /login for everyone". No server-side schema can
+    // prevent that (the old server is the old code), so the client checks that
+    // what came back is what it asked for, and undoes the write if it is not.
+    if (user_agent &&
+        !waf.rules.some(r => rule_matches_request(r, want, action))) {
+        await api
+            .removeAppWafRule(app_id, { pattern, user_agent, action })
+            // If the rollback itself is unsupported, fall back to the path-only
+            // shape the old server actually stored.
+            .catch(() => api.removeAppWafRule(app_id, { pattern, action }))
+            .catch(() => undefined);
+        throw new Error(`This Faable server does not support user-agent rules yet, so the rule was not created.\n` +
+            `  Update the CLI and try again, or ask support to upgrade the platform.`);
+    }
     const answer = action === 'deny' ? '403' : '404';
-    log.info(`🛡️  ${pattern} → ${answer} at the edge for ${app_name} (${app_id}).`);
+    log.info(`🛡️  ${describe_selector(want)} → ${answer} at the edge`);
+    log.info(`   for ${app_name} (${app_id}).`);
     log.info(``);
     log.info(action === 'deny'
         ? `Matching requests are blocked before they reach your app, so they no longer wake it.`
         : `Faable answers matching requests with a ${answer} itself, so they no longer wake your app.`);
+    if (user_agent) {
+        log.info(``);
+        log.info(`⚠️  A user-agent rule has no path restriction: it applies to EVERY path` +
+            (pattern ? ` that also matches ${pattern}.` : ` on this app.`));
+        if (action === 'sink' && !pattern) {
+            // Worth saying out loud: the sink router sits above the sleep-through
+            // probe, so a monitor caught by this rule gets the 404 instead of the
+            // health answer it expects.
+            log.info(`   If that user-agent is an uptime monitor, it will now see a 404 instead of a health check.`);
+        }
+    }
     log.info(``);
     log.info(`It takes about 20s to reach the edge. Then check it with:`);
-    log.info(`  curl -s -o /dev/null -w '%{http_code}\\n' https://${app_url}<path>`);
+    log.info(user_agent
+        ? `  curl -s -o /dev/null -w '%{http_code}\\n' -A '${user_agent}' https://${app_url}/`
+        : `  curl -s -o /dev/null -w '%{http_code}\\n' https://${app_url}<path>`);
     log.info(``);
-    log.info(`Undo: faable deploy waf rm '${pattern}' -a ${app_id}`);
+    log.info(`Undo: faable deploy waf rm ${undo_args(want)} -a ${app_id}`);
 };
+/** The exact arguments that reverse this rule. */
+const undo_args = (s) => [
+    s.pattern ? `'${s.pattern}'` : '',
+    s.user_agent ? `--user-agent '${s.user_agent}'` : ''
+]
+    .filter(Boolean)
+    .join(' ');
 
-export { add_rule };
+export { add_rule, describe_selector, rule_matches_request, undo_args };

package/dist/commands/deploy/waf/block.js

@@ -3,13 +3,17 @@ import { resolve_app_id } from '../resolve_app_id.js';
 import { add_rule } from './add_rule.js';
 
 const waf_block = {
-    command: 'block <pattern>',
-    describe: 'Block a path at the edge with a 403 (never reaches your app)',
+    command: 'block [pattern]',
+    describe: 'Block requests at the edge with a 403 (they never reach your app)',
     builder: yargs => yargs
         .positional('pattern', {
         type: 'string',
-        demandOption: true,
         description: 'Anchored path regex, e.g. ^/\\.well-known/ (quote it in your shell)'
+    })
+        .option('user-agent', {
+        alias: 'u',
+        type: 'string',
+        description: 'RE2 fragment matched against User-Agent, e.g. YisouSpider. With a path, BOTH must match.'
     })
         .option('app', {
         alias: 'a',
@@ -20,8 +24,23 @@ const waf_block = {
         alias: 'd',
         type: 'string',
         description: 'Why this rule exists (shown in `waf list`)'
+    })
+        .option('force', {
+        type: 'boolean',
+        default: false,
+        description: 'Accept a user-agent that also matches a known search/AI crawler or uptime monitor'
+    })
+        .check(({ pattern, userAgent }) => {
+        if (!pattern && !userAgent) {
+            throw new Error('Give a path pattern, --user-agent, or both.\n' +
+                "  e.g. faable deploy waf block '^/wp-admin'\n" +
+                '       faable deploy waf block --user-agent YisouSpider');
+        }
+        return true;
     })
         .example("$0 deploy waf block '^/\\.well-known/'", 'Stop scanner probes under /.well-known from waking the app')
+        .example('$0 deploy waf block --user-agent YisouSpider', 'Refuse one crawler outright, on every path')
+        .example("$0 deploy waf block '^/api/' --user-agent curl", 'Refuse /api/ to that user-agent only, leaving browsers untouched')
         .showHelpOnFail(false),
     handler: async (args) => {
         const ctx = await requireApi();
@@ -33,8 +52,10 @@ const waf_block = {
             app_name: app.name,
             app_url: app.url,
             pattern: args.pattern,
+            user_agent: args.userAgent,
             action: 'deny',
-            description: args.description
+            description: args.description,
+            force: args.force
         });
     }
 };

package/dist/commands/deploy/waf/format.js

@@ -3,7 +3,9 @@ const action_label = (action) => action === 'sink'
     ? '404 (answered by Faable, app not woken)'
     : action === 'deny'
         ? '403 (blocked at the edge)'
-        : action;
+        : action === 'probe'
+            ? 'answered by Faable, app not woken (uptime monitors)'
+            : action;
 /**
  * Render the effective WAF of an app.
  *
@@ -22,9 +24,13 @@ const format_waf = (waf) => {
         out.push('  (none)');
     }
     for (const p of waf.platform_profiles) {
-        out.push(`  • ${p.name} — ${p.rule_count} rule(s), ${action_label(p.action)}`);
+        // Say what the ruleset matches ON, not just what it answers: without it a
+        // `probe` profile of user-agents reads as if it blocked paths.
+        const on = p.match === 'user_agent' ? ' by user-agent' : '';
+        out.push(`  • ${p.name} — ${p.rule_count} rule(s)${on}, ${action_label(p.action)}`);
         for (const r of p.rules ?? []) {
-            out.push(`      ${r.pattern}${r.description ? `  # ${r.description}` : ''}`);
+            const what = p.match === 'user_agent' ? `UA ~ ${r.pattern}` : r.pattern;
+            out.push(`      ${what}${r.description ? `  # ${r.description}` : ''}`);
         }
     }
     out.push('');
@@ -33,11 +39,26 @@ const format_waf = (waf) => {
         out.push('  (none)');
     }
     for (const r of waf.rules) {
-        out.push(`  • ${r.pattern} → ${action_label(r.action)}`);
+        out.push(`  • ${rule_subject(r)} → ${action_label(r.action)}`);
         if (r.description)
             out.push(`      ${r.description}`);
     }
     return out;
 };
+/**
+ * What a rule selects, in one line.
+ *
+ * The three shapes have to be visibly different here: a user-agent rule has no
+ * path restriction (it applies to the whole app) and a combined one is an AND,
+ * so printing either as a bare pattern would misrepresent its reach.
+ */
+const rule_subject = (r) => {
+    if (r.pattern && r.user_agent) {
+        return `${r.pattern} + UA ~ ${r.user_agent}`;
+    }
+    if (r.user_agent)
+        return `UA ~ ${r.user_agent} (any path)`;
+    return r.pattern ?? '(unknown)';
+};
 
-export { action_label, format_waf };
+export { action_label, format_waf, rule_subject };

package/dist/commands/deploy/waf/rm.js

@@ -1,30 +1,57 @@
 import { requireApi } from '../../../api/context.js';
 import { log } from '../../../log.js';
 import { resolve_app_id } from '../resolve_app_id.js';
+import { describe_selector } from './add_rule.js';
 
 const waf_rm = {
-    command: 'rm <pattern>',
+    command: 'rm [pattern]',
     describe: 'Remove one of your WAF rules',
     builder: yargs => yargs
         .positional('pattern', {
         type: 'string',
-        demandOption: true,
         description: 'The exact pattern to remove (see `faable deploy waf list`)'
+    })
+        // A rule is addressed by everything that defines it, not by its path
+        // alone: "/login" and "/login for YisouSpider" are two different rules
+        // that share a pattern, and removing one must not take out the other.
+        .option('user-agent', {
+        alias: 'u',
+        type: 'string',
+        description: 'The user-agent of the rule to remove'
+    })
+        .option('action', {
+        type: 'string',
+        choices: ['deny', 'sink'],
+        description: 'Only needed when the same rule exists as both'
     })
         .option('app', {
         alias: 'a',
         type: 'string',
         description: 'App Identifier (defaults to the linked app)'
+    })
+        .check(({ pattern, userAgent }) => {
+        if (!pattern && !userAgent) {
+            throw new Error('Give the path pattern, --user-agent, or both.');
+        }
+        return true;
     })
         .example("$0 deploy waf rm '^/robots\\.txt$'", 'Stop handling /robots.txt at the edge')
+        .example('$0 deploy waf rm --user-agent YisouSpider', 'Stop blocking that crawler')
         .showHelpOnFail(false),
     handler: async (args) => {
         const ctx = await requireApi();
         const app_id = await resolve_app_id(args.app, ctx.appId, ctx.api);
         const app = await ctx.api.getApp(app_id);
-        await ctx.api.removeAppWafRule(app_id, args.pattern);
-        log.info(`🗑️  Removed ${args.pattern} from ${app.name} (${app_id}).`);
-        log.info(`Requests for it reach your app again within ~20s, once the edge picks up the change.`);
+        await ctx.api.removeAppWafRule(app_id, {
+            pattern: args.pattern,
+            user_agent: args.userAgent,
+            action: args.action
+        });
+        log.info(`🗑️  Removed ${describe_selector({
+            pattern: args.pattern,
+            user_agent: args.userAgent
+        })} from ${app.name} (${app_id}).`);
+        log.info(`They reach your app again within ~20s, once the edge picks up the change.`);
     }
 };
 

package/dist/commands/deploy/waf/sink.js

@@ -3,13 +3,17 @@ import { resolve_app_id } from '../resolve_app_id.js';
 import { add_rule } from './add_rule.js';
 
 const waf_sink = {
-    command: 'sink <pattern>',
-    describe: 'Answer a path with a 404 from Faable, without waking your app',
+    command: 'sink [pattern]',
+    describe: 'Answer requests with a 404 from Faable, without waking your app',
     builder: yargs => yargs
         .positional('pattern', {
         type: 'string',
-        demandOption: true,
         description: 'Anchored path regex, e.g. ^/robots\\.txt$ (quote it in your shell)'
+    })
+        .option('user-agent', {
+        alias: 'u',
+        type: 'string',
+        description: 'RE2 fragment matched against User-Agent, e.g. YisouSpider. With a path, BOTH must match.'
     })
         .option('app', {
         alias: 'a',
@@ -20,11 +24,25 @@ const waf_sink = {
         alias: 'd',
         type: 'string',
         description: 'Why this rule exists (shown in `waf list`)'
+    })
+        .option('force', {
+        type: 'boolean',
+        default: false,
+        description: 'Accept a user-agent that also matches a known search/AI crawler or uptime monitor'
+    })
+        .check(({ pattern, userAgent }) => {
+        if (!pattern && !userAgent) {
+            throw new Error('Give a path pattern, --user-agent, or both.\n' +
+                "  e.g. faable deploy waf sink '^/wp-admin'\n" +
+                '       faable deploy waf sink --user-agent YisouSpider');
+        }
+        return true;
     })
         .example("$0 deploy waf sink '^/robots\\.txt$'", 'Let Faable answer /robots.txt with a 404 instead of starting your app')
-        .epilogue('Use `sink` for paths your app does not serve anyway: Faable replies ' +
-        'with the same 404 your app would have, without the cold start. ' +
-        'Use `block` instead when you want the request refused outright.')
+        .example('$0 deploy waf sink --user-agent SemrushBot', 'Answer one crawler with a 404 instead of waking the app for it')
+        .epilogue('Use `sink` for requests your app would 404 anyway: Faable replies ' +
+        'with the same 404, without the cold start. Use `block` instead ' +
+        'when you want the request refused outright.')
         .showHelpOnFail(false),
     handler: async (args) => {
         const ctx = await requireApi();
@@ -36,8 +54,10 @@ const waf_sink = {
             app_name: app.name,
             app_url: app.url,
             pattern: args.pattern,
+            user_agent: args.userAgent,
             action: 'sink',
-            description: args.description
+            description: args.description,
+            force: args.force
         });
     }
 };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@faable/faable",
-  "version": "2.1.0",
+  "version": "2.2.0",
   "main": "dist/index.js",
   "license": "MIT",
   "author": "Marc Pomar <marc@faable.com>",