Official Crawl4AI Cloud: a hosted web scraping, extraction, and search API for AI agents. Free API key, 1,000 credits a month, no card. Fully open source: this repo is the Apache-2.0 crawler that powers it, and you can self-host it.
Python aiai-agentscrawl4ai-cloudcrawlerdata-extractionllmmarkdownmcpopen-sourceplaywrightpythonragscraperweb-crawlerweb-crawlingweb-scraping
fix(playground): forward the typed params, and clear status when config is empty
ntohidie138aa0f68e5
diff --git a/deploy/docker/static/playground/index.html b/deploy/docker/static/playground/index.htmlindex f905652f..1b0353b5 100644--- a/deploy/docker/static/playground/index.html+++ b/deploy/docker/static/playground/index.html@@ -568,11 +568,14 @@}// Validate the editor's params against /config/dump and return the- // normalized { type, params } the crawl endpoints expect. Throws with the- // server's message so the caller can surface the offending field.+ // { type, params } the crawl endpoints expect. Throws with the server's+ // message so the caller can surface the offending field.async function validateConfig() {const text = cm.getValue().trim();- if (!text) return {};+ if (!text) {+ setCfgStatus('', 'text-secondary'); // no config is not an error+ return {};+ }let params;try {@@ -585,7 +588,10 @@setCfgStatus('✖ expected a JSON object', 'text-red-400');throw new Error('Advanced Config must be a JSON object of parameters');}- if (!Object.keys(params).length) return {};+ if (!Object.keys(params).length) {+ setCfgStatus('', 'text-secondary'); // no config is not an error+ return {};+ }const cfgType = document.getElementById('cfg-type').value;const res = await authFetch('/config/dump', {@@ -601,7 +607,11 @@}setCfgStatus('✓ validated', 'text-green-400');- return await res.json();+ // Send back what the user typed, not /config/dump's echo. dump()+ // fills in server-derived fields -- BrowserConfig gains generated+ // `headers` -- that the untrusted gate then refuses on /crawl, so+ // the echo is a validation result, not a re-submittable payload.+ return { type: cfgType, params };}// ================ SERVER COMMUNICATION ================diff --git a/deploy/docker/tests/test_playground_config_protocol.py b/deploy/docker/tests/test_playground_config_protocol.pyindex dcb330e2..926b113d 100644--- a/deploy/docker/tests/test_playground_config_protocol.py+++ b/deploy/docker/tests/test_playground_config_protocol.py@@ -96,10 +96,29 @@ class TestTemplatesSurviveTheUntrustedGate(unittest.TestCase):def test_stream_reaches_the_shape_should_use_stream_reads(self):"""runCrawl picks the streaming endpoint off crawler_config.params.stream."""params = json.loads(read_templates()["CrawlerRunConfig"])- dumped = CrawlerRunConfig.load(- {"type": "CrawlerRunConfig", "params": params}, provenance=Provenance.UNTRUSTED+ self.assertIs(params.get("stream"), True)++ def test_the_validated_params_are_forwarded_not_the_dump_echo(self):+ """dump() adds server-derived fields the untrusted gate then refuses.++ BrowserConfig picks up generated `headers`, which is not allowlisted, so+ posting /config/dump's echo to /crawl 400s. The page must forward the+ params the user typed.+ """+ params = json.loads(read_templates()["BrowserConfig"])+ echo = BrowserConfig.load(+ {"type": "BrowserConfig", "params": params}, provenance=Provenance.UNTRUSTED).dump()- self.assertIs(dumped["params"].get("stream"), True)+ forbidden = set(echo["params"]) - UNTRUSTED_FIELD_ALLOWLIST["BrowserConfig"]+ self.assertTrue(+ forbidden,+ "if dump() ever became re-submittable this guard can go, but check the page first",+ )+ page = read_page()+ self.assertIn("return { type: cfgType, params };", page,+ "validateConfig must forward the typed params, not the echo")+ self.assertNotIn("return await res.json();", page,+ "the /config/dump echo must not be forwarded to /crawl")def test_the_old_code_protocol_is_still_refused(self):"""Guards the reason for this change: {type, code} cannot come back."""