Documentation Index
Fetch the complete documentation index at: https://docs.praison.ai/llms.txt
Use this file to discover all available pages before exploring further.
Profiler Module
The Profiler module provides standardized profiling for performance monitoring across praisonai and praisonai-agents.
Features
- Import timing
- Function execution timing
- Flow tracking
- File/module usage tracking
- Memory usage (tracemalloc)
- API call profiling (wall-clock time)
- Streaming profiling (TTFT, total time)
- Statistics (p50, p95, p99)
- cProfile integration
- Flamegraph generation
- Line-level profiling
- JSON/HTML export
Import
from praisonai.profiler import Profiler, profile, profile_imports
Quick Examples
Profile a Function
from praisonai.profiler import profile
@profile
def my_function():
# Your code here
pass
Profile a Block
from praisonai.profiler import Profiler
with Profiler.block("my_operation"):
do_something()
Profile API Calls
from praisonai.profiler import Profiler
import requests
with Profiler.api_call("https://api.example.com") as call:
response = requests.get("https://api.example.com/data")
Profile Streaming
from praisonai.profiler import Profiler
with Profiler.streaming("chat") as tracker:
tracker.first_token()
for chunk in stream:
tracker.chunk()
Profile Imports
from praisonai.profiler import profile_imports
with profile_imports():
import heavy_module
Profiler Class
Static Methods
Profiler.block(name)
Context manager for profiling a code block.
with Profiler.block("database_query"):
results = db.query(...)
Profiler.api_call(url)
Context manager for profiling API calls.
with Profiler.api_call("https://api.openai.com/v1/chat") as call:
response = openai.chat.completions.create(...)
Profiler.streaming(name)
Context manager for profiling streaming operations.
with Profiler.streaming("llm_stream") as tracker:
tracker.first_token() # Call when first token arrives
for chunk in stream:
tracker.chunk() # Call for each chunk
Profiler.report()
Print a profiling report to console.
Profiler.get_statistics()
Get profiling statistics.
Returns: dict with p50, p95, p99 percentiles
stats = Profiler.get_statistics()
print(f"p50: {stats['p50']}ms")
print(f"p95: {stats['p95']}ms")
print(f"p99: {stats['p99']}ms")
Profiler.export_json(path)
Export profiling data to JSON.
Profiler.export_json("profile_data.json")
Profiler.export_html(path)
Export profiling data to HTML report.
Profiler.export_html("profile_report.html")
Profiler.reset()
Reset all profiling data.
Decorators
@profile
Decorator to profile a function.
from praisonai.profiler import profile
@profile
def expensive_operation():
# This function will be profiled
pass
@profile
async def async_operation():
# Async functions are also supported
pass
Context Managers
profile_imports()
Profile import times for modules.
from praisonai.profiler import profile_imports
with profile_imports():
import pandas
import numpy
import tensorflow
# Check which imports were slow
Profiler.report()
Example: Full Profiling Session
from praisonai.profiler import Profiler, profile, profile_imports
# Profile imports
with profile_imports():
from praisonaiagents import Agent
# Profile function
@profile
def run_agent(prompt):
agent = Agent(name="Test")
return agent.start(prompt)
# Profile blocks
with Profiler.block("full_workflow"):
with Profiler.block("setup"):
# Setup code
pass
with Profiler.block("execution"):
result = run_agent("Hello")
with Profiler.block("cleanup"):
# Cleanup code
pass
# Get report
Profiler.report()
# Export data
Profiler.export_json("profile.json")
Profiler.export_html("profile.html")
# Get statistics
stats = Profiler.get_statistics()
print(f"Median execution time: {stats['p50']}ms")
