Add Promethus histograms for some parts of the code (#114)

* Add support for exporting code block times as histograms

* Adjust perf metrics

* Aff perf metrics for message processing and title gen

* Add metric for opening app home views

* Performance metrics are now defined individually

* Remove performance metric from slack.py

* New metrics for ticket title gen and app home rendering

* Prevent .labels({}) from happening

* Make import consistent

nephthys/events/app_home_opened.py

@@ -2,6 +2,7 @@ import logging
 import traceback
 from typing import Any
 
+from prometheus_client import Histogram
 from slack_sdk.web.async_client import AsyncWebClient
 
 from nephthys.utils.env import env
@@ -21,6 +22,13 @@ async def on_app_home_opened(event: dict[str, Any], client: AsyncWebClient):
     await open_app_home("default", client, user_id)
 
 
+APP_HOME_RENDER_DURATION = Histogram(
+    "nephthys_app_home_render_duration_seconds",
+    "How long it takes to load the app home screen",
+    ["home_type"],
+)
+
+
 async def open_app_home(home_type: str, client: AsyncWebClient, user_id: str):
     try:
         await client.views_publish(view=get_loading_view(), user_id=user_id)
@@ -37,7 +45,11 @@ async def open_app_home(home_type: str, client: AsyncWebClient, user_id: str):
             view = get_unknown_user_view(name)
         else:
             logging.info(f"Opening {home_type} for {user_id}")
-            async with perf_timer(f"Rendering app home (type={home_type})"):
+            async with perf_timer(
+                f"Rendering app home (type={home_type})",
+                APP_HOME_RENDER_DURATION,
+                home_type=home_type,
+            ):
                 match home_type:
                     case "default" | "dashboard":
                         view = await get_helper_view(user)

nephthys/events/message.py

@@ -4,6 +4,7 @@ from typing import Any
 from typing import Dict
 
 from openai import OpenAIError
+from prometheus_client import Histogram
 from slack_sdk.errors import SlackApiError
 from slack_sdk.web.async_client import AsyncWebClient
 
@@ -19,6 +20,11 @@ from prisma.models import User
 # Message subtypes that should be handled by on_message (messages with no subtype are always handled)
 ALLOWED_SUBTYPES = ["file_share", "me_message", "thread_broadcast"]
 
+TICKET_TITLE_GENERATION_DURATION = Histogram(
+    "nephthys_ticket_title_generation_duration_seconds",
+    "How long it takes to generate a ticket title using AI",
+)
+
 
 async def handle_message_sent_to_channel(event: Dict[str, Any], client: AsyncWebClient):
     """Tell a non-helper off because they sent a thread message with the 'send to channel' box checked."""
@@ -185,7 +191,9 @@ async def handle_new_question(
             event, client, text=user_facing_message_text, ticket_url=ticket_url
         )
 
-    async with perf_timer("AI ticket title generation"):
+    async with perf_timer(
+        "AI ticket title generation", TICKET_TITLE_GENERATION_DURATION
+    ):
         title = await generate_ticket_title(text)
 
     user_facing_message_ts = user_facing_message["ts"]

nephthys/utils/performance.py

@@ -3,12 +3,21 @@ import logging
 from contextlib import asynccontextmanager
 from time import perf_counter
 
+from prometheus_client import Histogram
+
 
 @asynccontextmanager
-async def perf_timer(name: str):
+async def perf_timer(
+    name: str, histogram_metric: Histogram | None = None, **metric_labels
+):
     start_time = perf_counter()
     yield
     duration = perf_counter() - start_time
     # Get the name of the function that called the `perf_timer()`
     function_name = inspect.stack()[2].function
     logging.debug(f"{function_name}: {name} took {duration:.3f}s")
+    if histogram_metric:
+        if metric_labels:
+            histogram_metric.labels(**metric_labels).observe(duration)
+        else:
+            histogram_metric.observe(duration)

nephthys/utils/slack.py

@@ -33,8 +33,8 @@ async def handle_message(event: Dict[str, Any], client: AsyncWebClient):
         and event["message"].get("subtype") == "tombstone"
     ) or event.get("subtype") == "message_deleted"
 
-    async with perf_timer("Processing message event (total time)"):
-        if event["channel"] == env.slack_help_channel:
+    if event["channel"] == env.slack_help_channel:
+        async with perf_timer("Processing message event (total time)"):
             if is_message_deletion:
                 await on_message_deletion(event, client)
             else: