Skip to content

~95% on SimpleQA (e.g. Qwen3.6-27B on a 3090). Supports all local and cloud LLMs (llama.cpp, Ollama, Google, ...). 10+ search engines - arXiv, PubMed, your private documents. Everything Local & Encrypted.

docs(ui-tests): note local rate-limit gotcha in README (#4119)

LearningCircuit7fcc336eb4f4
diff --git a/tests/ui_tests/README.md b/tests/ui_tests/README.md
index c722f478c7..9ba1e89897 100644
--- a/tests/ui_tests/README.md
+++ b/tests/ui_tests/README.md
@@ -11,6 +11,26 @@ npm install puppeteer
## Running Tests
+### Local environment notes
+
+The auth endpoints rate-limit at **5 logins per 15 min** and **3 registrations per hour** per IP. The full suite re-authenticates dozens of times, so a local run against a default-configured server gets blocked after the first few tests and shows a wave of `Login failed` failures that are not real test failures.
+
+Two options for local runs:
+
+1. **Disable rate limiting** (simplest):
+ ```bash
+ LDR_DISABLE_RATE_LIMITING=true python -m local_deep_research.web.app
+ ```
+
+2. **Use the CI test-user pattern**: pre-create `test_admin` once, then run the suite with `CI=true` so `auth_helper.js` logs in with that user instead of registering fresh users each time.
+ ```bash
+ python scripts/ci/init_test_database.py # one-time setup
+ CI=true node tests/ui_tests/run_all_tests.js --shard=<shard>
+ ```
+ Note: `CI=true` requires `--shard=<name>` (the runner fails fast otherwise to catch matrix misconfiguration). Iterate shards manually for a full local run.
+
+Either approach is fine; option 1 is closer to how CI runs the suite (CI launches the server with rate limiting off).
+
### Run All UI Tests (Recommended)
```bash
cd /path/to/local-deep-research