diff --git a/pyproject.toml b/pyproject.toml
index 2a3a9c7..962f7e1 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,6 +1,6 @@
 [project]
 name = "ollama-queue-proxy"
-version = "0.3.1"
+version = "0.3.2"
 description = "Drop-in HTTP proxy for Ollama with priority queuing, per-client auth, and model-aware failover"
 readme = "README.md"
 license = { text = "MIT" }
diff --git a/src/ollama_queue_proxy/routes/status.py b/src/ollama_queue_proxy/routes/status.py
index 9c158fc..c62c963 100644
--- a/src/ollama_queue_proxy/routes/status.py
+++ b/src/ollama_queue_proxy/routes/status.py
@@ -202,6 +202,14 @@ async def metrics(request: Request):
                 f'oqp_embedding_cache_hits_total{{client="{_pm_label(client_id)}",model="{_pm_label(model)}",'
                 f'endpoint="{_pm_label(endpoint)}"}} {count}'
             )
+        # Emit zero for miss labels absent from hits so Prometheus always has the time series
+        for label in cache_misses:
+            if label not in cache_hits:
+                client_id, model, endpoint = label.split(",", 2)
+                lines.append(
+                    f'oqp_embedding_cache_hits_total{{client="{_pm_label(client_id)}",model="{_pm_label(model)}",'
+                    f'endpoint="{_pm_label(endpoint)}"}} 0'
+                )
 
         lines += [
             "# HELP oqp_embedding_cache_misses_total Embedding cache misses",