A Flask extension that provides multi-dimensional profiling capabilities for Flask applications.

• Code profiling: Statistical profiling using pyinstrument to identify performance bottlenecks
• SQL profiling: Database query profiling using sqltap to track query performance
• Search profiling: Custom profiler for OpenSearch/Elasticsearch operations via trace logging

pip install flask-multiprofiler

For search profiling support, install with the appropriate extra:

# For OpenSearch
pip install flask-multiprofiler[opensearch]

# For Elasticsearch
pip install flask-multiprofiler[elasticsearch]

from flask import Flask
from flask_multiprofiler import MultiProfiler

app = Flask(__name__)
profiler = MultiProfiler(app)

# Or using the factory pattern
profiler = MultiProfiler()
profiler.init_app(app)

1. Navigate to /profiler in your Flask application
2. Start a profiling session by selecting which profilers to enable
3. Use your application normally - profiling data is collected automatically
4. View profiling reports through the web interface

# Storage directory for profiling sessions (default: "{Flask.instance_path}/profiler")
MULTIPROFILER_STORAGE = "/path/to/storage"

# Logger name for search profiling (default: "opensearchpy.trace")
MULTIPROFILER_SEARCH_TRACE_LOGGER = "elasticsearchpy.trace"

# Session lifetime (default: 60 minutes)
MULTIPROFILER_ACTIVE_SESSION_LIFETIME = timedelta(minutes=30)

# Endpoints to ignore during profiling (regex patterns)
MULTIPROFILER_IGNORED_ENDPOINTS = ["static", r"profiler\..+", "api.health"]

# Permission function (default: lambda: True)
MULTIPROFILER_PERMISSION = lambda: current_user.is_admin

This project uses uv for dependency management and just for task automation. Tests use testcontainers to run PostgreSQL and OpenSearch in Docker containers.

# Run tests
just test

# Run all code-quality checks
just check

• Each profiling session creates its own SQLite database for storing profiling reports in a self-contained HTML format
• Profilers can be enabled/disabled individually per session
• Web interface provides session management and report viewing

The search profiler currently supports:

• OpenSearch client (opensearch-py)
• Elasticsearch client v7.x only (elasticsearch~=7.17.12)

Both clients use the same logging mechanism (plain Python logger) for trace logging, which our profiler intercepts. Elasticsearch v8.x uses a different tracing approach and is not currently supported.

Planned features and improvements:

• Redis profiling: Track Redis operations and performance
• Enhanced search query viewer: Improved visualization and analysis of search queries
• SQL profiler rewrite: Replace sqltap with custom implementation to add stack traces and better integration

This module intentionally keeps things simple and will not support:

• Customizing rendering: Reports are stored as fully renderable HTML pages to keep complexity low
• REST API: No plans to expose profiling data via API endpoint
• Async support: Focused on traditional Flask applications only
• Full type hints: Minimal type annotations to maintain simplicity
• Translations/I18N: This is an sysadmin/developer interface