📄 819% (8.19x) speedup for Transcript.get_by_id_async in assemblyai/transcriber.py

⏱️ Runtime : 3.03 milliseconds → 330 microseconds (best of 8 runs)

📝 Explanation and details

The key optimization is moving the ThreadPoolExecutor from instance-level to module-level.

In the original code, every Transcript object created its own ThreadPoolExecutor in __init__(), which is extremely expensive. Each executor spawns multiple threads and allocates significant resources. When creating many Transcript instances (as suggested by the test patterns), this leads to:

• Thread pool creation overhead for each instance
• Memory bloat from multiple thread pools
• OS-level thread context switching penalties

The optimized version uses a single shared _executor at module level. All Transcript instances now share the same thread pool, eliminating the per-instance creation cost while maintaining identical async behavior.

Performance impact: The line profiler shows wait_for_completion_async() improved from 568μs per hit to 63μs per hit (~9x faster), and the overall runtime improved from 3.03ms to 330μs (818% speedup).

This optimization is particularly effective for workloads that create multiple Transcript objects or call get_by_id_async() frequently, as each call creates a new instance. The shared executor maintains thread reuse efficiency while dramatically reducing initialization overhead.

✅ Correctness verification report:

import concurrent.futures

# imports
import pytest
from assemblyai.transcriber import Transcript

# --- Unit tests ---

# Basic Test Cases





def test_edge_none_id_raises_typeerror():
    """Test that passing None as transcript_id raises TypeError."""
    with pytest.raises(TypeError):
        Transcript.get_by_id_async(None)










#------------------------------------------------
import concurrent.futures
import threading
import time

# imports
import pytest
from assemblyai.transcriber import Transcript

# ================================
# Unit tests for get_by_id_async()
# ================================

# ----------- Basic Tests ------------



















#------------------------------------------------
from assemblyai.transcriber import Transcript
import pytest

def test_Transcript_get_by_id_async():
    with pytest.raises(TypeError, match="__dict__\\ must\\ be\\ set\\ to\\ a\\ dictionary,\\ not\\ a\\ 'ShellMutableMap'"):
        Transcript.get_by_id_async(Transcript, '')

To edit these changes git checkout codeflash/optimize-Transcript.get_by_id_async-mgv02utj and push.