debug-agent-py 0.5.1

AI-powered runtime debugging agent for Python web applications

Python Debug Agent

An AI-powered runtime debugging agent that embeds directly into your Python web application. Add one dependency, configure an LLM key, and chat with your live app at /agent to inspect memory, threads, GC, modules, database connections, Redis, Django models/URLs, Celery tasks, Flask extensions, Jinja2 templates, signals, routes, HTTP requests, and more — 68 diagnostic tools across 24 inspectors.

Version Support

Python Version	Status
3.8	Not supported
3.9	Minimum supported
3.10	Supported
3.11	Supported
3.12	Supported
3.13	Supported
3.14	Tested

Uses from __future__ import annotations so X | Y union types work on Python 3.9+. Optional dependencies (Flask, SQLAlchemy, Redis, Celery) are not required at install time.

Quick Start

1. Install

pip install debug-agent-py

2. Integrate (Flask)

from flask import Flask
from debug_agent import setup_debug_agent

app = Flask(__name__)

# One line to integrate
setup_debug_agent(app)

3. Configure LLM

export LLM_API_KEY=your-key
export LLM_BASE_URL=https://open.bigmodel.cn/api/coding/paas/v4  # default
export LLM_MODEL=glm-5.2                                          # default

Supports any OpenAI-compatible endpoint.

4. Run and open

http://localhost:8000/agent

Features

• Streaming AI responses with real-time tool call badges (pending / success / error)
• Context compression — automatically summarizes old conversation when token limit is approached
• Dark-themed chat UI with full markdown rendering (tables, code blocks, lists)
• Max tool rounds (25) with forced final summary when limit is reached
• 68 diagnostic tools across 24 inspectors
• Works with Flask, FastAPI, and Django
• Zero external dependencies (no Datadog, no Grafana, no APM)

Inspectors & Tools (68)

Memory Inspector

Tool	Description
get_tracemalloc_stats	Python tracemalloc top allocations by file/line
get_object_counts	Count objects by type via gc
get_gc_stats	gc.get_stats() generation details
get_ref_cycles	Count reference cycles detected by gc
trigger_gc	Force garbage collection

Threads Inspector

Tool	Description
get_thread_info	List all threads with name, daemon, alive status
get_thread_count	Active thread count
get_thread_summary	Thread state distribution
get_thread_stacks	Current frame/stack for all threads

Database Inspector

Tool	Description
get_sqlalchemy_engines	Find SQLAlchemy engines and pool status
get_db_connections	Inspect database connection pools
get_db_pool_config	Pool configuration: size, timeout, recycle settings

Modules Inspector

Tool	Description
get_loaded_modules	List loaded Python modules (sys.modules) with versions
get_module_count	Total loaded module count
get_installed_packages	List installed packages from pip

Async Tasks Inspector

Tool	Description
get_async_tasks	List pending asyncio tasks
get_event_loop_info	Event loop details: type, running state
get_pending_callbacks	List scheduled callbacks on the event loop

Runtime Inspector

Tool	Description
get_memory_info	Process memory info (RSS, VMS, shared)
get_cpu_usage	CPU usage percentage
get_python_info	Python version, implementation, executable path
get_open_fds	Open file descriptor count and limits

System Inspector

Tool	Description
get_system_info	Hostname, platform, CPU cores, disk
get_environment_variables	Environment variables (masked secrets)
get_disk_usage	Disk usage for the working directory

Framework Inspector

Tool	Description
get_routes	List all registered web routes/endpoints
get_middleware	List registered middleware

HTTP Tracker Inspector

Tool	Description
get_recent_requests	Recent HTTP requests ring buffer
get_slow_requests	Slowest requests by duration
get_error_requests	Error requests (4xx/5xx)
get_request_stats	P50/P95/P99 latency, error rate

Redis Inspector

Tool	Description
get_redis_info	Redis server info: memory, clients, persistence
get_redis_keys	Scan Redis keyspace with pattern matching
get_redis_config	Redis runtime configuration (CONFIG GET)
get_redis_slowlog	Redis slow query log entries

Django Inspector

Tool	Description
get_django_models	List Django models with app label, table name, field count
get_django_urls	List all URL patterns with view names and namespaces
get_django_settings	Key Django settings (DBs, INSTALLED_APPS, MIDDLEWARE)
get_django_migrations	Migration status per app: applied vs pending

Celery Inspector

Tool	Description
get_celery_tasks	List registered Celery tasks with routing info
get_celery_workers	Active Celery workers with pool and concurrency
get_celery_queues	Queue depth and message stats per queue

Flask Extensions Inspector

Tool	Description
get_flask_extensions	List registered Flask extensions and their bindings
get_flask_blueprints	List Flask blueprints with URL prefixes and routes
get_flask_config	Flask configuration object values (secrets masked)

Jinja2 Inspector

Tool	Description
get_jinja_templates	List loaded Jinja2 templates with loader paths
get_jinja_filters	List registered Jinja2 filters, tests, and globals

Signals Inspector

Tool	Description
get_signal_handlers	List Python signal handlers registered via signal module
get_django_signals	List Django signal receivers connected to senders

WSGI/ASGI Inspector

Tool	Description
get_wsgi_info	WSGI server details (Gunicorn/uWSGI workers, config)
get_asgi_apps	List ASGI application scope and middleware chain

Logging Inspector

Tool	Description
get_logging_tree	Python logging module logger hierarchy and levels
get_recent_logs	Recent log entries from the built-in ring buffer
get_logging_config	Current logging configuration (handlers, formatters, levels)
set_log_level	Dynamically set the log level for a named logger

Cache Inspector

Tool	Description
get_cache_info	Stats for registered caches (hit rate, miss count, key count)
get_cache_keys	List keys from a registered cache with optional prefix filter
clear_cache	Clear all entries from a registered cache

Outbound HTTP Inspector

Tool	Description
get_http_pool_stats	HTTP client connection pool stats (connections, keepalive, timeouts)
get_outbound_summary	Aggregated outbound HTTP call stats (total, avg latency, error rate)

File Descriptor Inspector

Tool	Description
get_fd_count	Current number of open file descriptors
get_fd_limit	File descriptor soft and hard limits (RLIMIT_NOFILE)
get_fd_list	List open file descriptors with type and details

Metrics Inspector

Tool	Description
get_registered_metrics	List all registered metrics from prometheus_client
get_metric_value	Get current value of a specific metric by name

FastAPI Inspector

Tool	Description
get_fastapi_openapi	List FastAPI routes, schemas, and OpenAPI spec details

Warnings Inspector

Tool	Description
get_warnings	List captured Python warnings with category, message, and location

Custom Tools

from debug_agent import debug_tool

@debug_tool('check_redis', 'Check Redis connection')
def check_redis():
    return {'connected': True}

Configuration

Env Var	Default	Description
LLM_BASE_URL	https://open.bigmodel.cn/api/coding/paas/v4	LLM endpoint
LLM_API_KEY	(required)	API key
LLM_MODEL	glm-5.2	Model name
LLM_MAX_TOOL_ROUNDS	25	Max tool-calling rounds
LLM_CONTEXT_WINDOW_TOKENS	100000	Context window size

Run the Demo

The demo uses Flask + redis-py + SQLAlchemy + Celery. Start Redis with Docker Compose first:

Docker Compose

# docker-compose.yml
services:
  redis:
    image: redis:7-alpine
    ports:
      - "6379:6379"
    command: redis-server --save 60 1 --loglevel warning

docker compose up -d

Start the app

export LLM_API_KEY=your-key
cd demo && python app.py
# Open http://localhost:8000/agent

PyPI

License

MIT