diff --git a/.coverage b/.coverage
new file mode 100644
index 0000000..3e63c09
Binary files /dev/null and b/.coverage differ
diff --git a/main.py b/main.py
index c9a7715..7f77b18 100644
--- a/main.py
+++ b/main.py
@@ -1,24 +1,34 @@
-import argparse
 import os
+import time
+import argparse
 
 import uvicorn
 from fastapi import FastAPI, UploadFile
 from fastapi.encoders import jsonable_encoder
+from fastapi.responses import PlainTextResponse
+from starlette.middleware.base import BaseHTTPMiddleware
+os.environ['PROMETHEUS_DISABLE_CREATED_SERIES'] = 'true'
+# pylint: disable=wrong-import-position
+from prometheus_client import generate_latest, Counter, Histogram, Summary, REGISTRY
 
 from config import TEMP_DIR
 
 
 os.makedirs(TEMP_DIR, exist_ok=True)
 
+
 # Specify mode
 parser = argparse.ArgumentParser(description='Start service with different modes.')
 parser.add_argument('--langchain', action='store_true')
 parser.add_argument('--towhee', action='store_true')
+parser.add_argument('--max_observation', default=1000)
 
 args = parser.parse_args()
 
+
 USE_LANGCHAIN = args.langchain
 USE_TOWHEE = args.towhee
+MAX_OBSERVATION = args.max_observation
 
 assert (USE_LANGCHAIN and not USE_TOWHEE ) or (USE_TOWHEE and not USE_LANGCHAIN), \
     'The service should start with either "--langchain" or "--towhee".'
@@ -28,12 +38,90 @@
 if USE_TOWHEE:
     from src_towhee.operations import chat, insert, drop  # pylint: disable=C0413
 
+
 app = FastAPI()
 origins = ['*']
 
+
+# Define metrcis
+requests_total = Counter('requests_total', 'Cumulative requests')
+requests_success_total = Counter('requests_success_total', 'Cumulative successful requests')
+requests_failed_total = Counter('requests_failed_total', 'Cumulative failed requests')
+
+endpoint_requests_total = Counter('endpoint_requests_total', 'Cumulative requests of each endpoint', ['endpoint'])
+endpoint_requests_success_total = Counter('endpoint_requests_success_total', 'Cumulative successful requests of each endpoint', ['endpoint'])
+endpoint_requests_failed_total = Counter('endpoint_requests_failed_total', 'Cumulative failed requests of each endpoint', ['endpoint'])
+
+latency_seconds_histogram = Histogram('latency_seconds_histogram', 'Request process latency histogram')
+endpoint_latency_seconds_histogram = Histogram(
+    'endpoint_latency_seconds_histogram', 'Request process latency histogram of each endpoint', ['endpoint']
+)
+
+latency_seconds_summary = Summary('latency_seconds_summary', 'Request process latency summary')
+endpoint_latency_seconds_summary = Summary('endpoint_latency_seconds_summary', 'Request process latency summary of each endpoint', ['endpoint'])
+
+
+# Define middleware to collect metrics
+class RequestMetricsMiddleware(BaseHTTPMiddleware):
+    """
+    Middleware to process requests.
+    """
+    async def dispatch(self, request, call_next):
+        path = request.scope.get('path')
+        is_req = path != '/metrics'
+
+        if not is_req:
+            try:
+                response = await call_next(request)
+                return response
+            except Exception as e:
+                raise e
+
+        begin = time.time()
+        requests_total.inc()
+        endpoint_requests_total.labels(path).inc()
+        try:
+            response = await call_next(request)
+            if response.status_code / 100 < 4:
+                requests_success_total.inc()
+                endpoint_requests_success_total.labels(path).inc()
+                end = time.time()
+                if path in latencies:
+                    latencies[path].append(end - begin)
+                    latencies[path] = latencies[path][-MAX_OBSERVATION:]
+                else:
+                    latencies[path] = [end - begin]
+                latencies['all'].append(end - begin)
+                latencies['all'] = latencies['all'][-MAX_OBSERVATION:]
+                latency_seconds_histogram.observe(end - begin)
+                endpoint_latency_seconds_histogram.labels(path).observe(end - begin)
+                latency_seconds_summary.observe(end - begin)
+                endpoint_latency_seconds_summary.labels(path).observe(end - begin)
+                return response
+            else:
+                requests_failed_total.inc()
+                endpoint_requests_failed_total.labels(path).inc()
+        except Exception as e:
+            requests_failed_total.inc()
+            endpoint_requests_failed_total.labels(path).inc()
+            raise e
+
+
+app.add_middleware(RequestMetricsMiddleware)
+
+
 @app.get('/')
 def check_api():
-    return jsonable_encoder({'status': True, 'msg': 'ok'}), 200
+    res = jsonable_encoder({'status': True, 'msg': 'ok'}), 200
+    return res
+
+
+@app.get('/metrics')
+async def metrics():
+    registry = REGISTRY
+    data = generate_latest(registry)
+    return PlainTextResponse(content=data, media_type='text/plain')
+
 
 @app.get('/answer')
 def do_answer_api(session_id: str, project: str, question: str):
diff --git a/requirements.txt b/requirements.txt
index c1516ab..35c55cf 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -11,3 +11,4 @@ uvicorn
 towhee>=1.1.0
 pymilvus
 elasticsearch>=8.0.0
+prometheus-client