From db2f47ddd1f1893eb0ef3b9da41b108db50198c2 Mon Sep 17 00:00:00 2001
From: Pratima Kumari <104563658+Pratimakumari8@users.noreply.github.com>
Date: Fri, 3 Jul 2026 11:41:54 +0530
Subject: [PATCH] fix: add Tinker warmup step before integration tests

The Tinker proxy cold-starts on first inference because the
tokenizer downloads on demand (>120s). The existing health
check only confirms the API server is up, not that it is
ready to serve inference requests.

This adds a warmup step that sends a real inference request
and retries until it succeeds (timeout 300s) before the
OpenClaw gateway check and test run begin.
---
 tests/integration/run_tinker_stack_integration.sh | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/tests/integration/run_tinker_stack_integration.sh b/tests/integration/run_tinker_stack_integration.sh
index 3e36791..322f665 100755
--- a/tests/integration/run_tinker_stack_integration.sh
+++ b/tests/integration/run_tinker_stack_integration.sh
@@ -47,6 +47,16 @@ compose up -d --build
 echo "==> Waiting for CLaaS API ..."
 timeout 120 bash -c 'until curl -sf http://127.0.0.1:8080/ >/dev/null 2>&1; do sleep 5; done'
 
+# ── 2b. Warm up Tinker (tokenizer downloads on first inference) ────
+echo "==> Warming up Tinker inference (tokenizer download may take >120s) ..."
+timeout 300 bash -c '
+  until curl -sf \
+    -H "Content-Type: application/json" \
+    -X POST http://127.0.0.1:8080/v1/chat/completions \
+    -d "{\"model\":\"'"$MODEL"'\",\"messages\":[{\"role\":\"user\",\"content\":\"hi\"}],\"max_tokens\":1}" \
+    -o /dev/null 2>&1; do sleep 10; done'
+echo "==> Tinker warmed up."
+
 echo "==> Waiting for OpenClaw gateway ..."
 timeout 180 bash -c '
   until curl -sf \