[FEATURE] Parallelize S3 message retrieval in S3SessionManager

[CrysisDeu]

Problem Statement

Summary

Optimize S3SessionManager.list_messages() to retrieve message files from S3 in parallel instead of sequentially, significantly improving performance for sessions with many messages.

Problem

Currently, list_messages() loads message files one at a time:

for key in message_keys:
    message_data = self._read_s3_object(key)  # Sequential - slow!

Performance Impact:

• 100 messages × 50ms latency = ~5 seconds
• Network I/O bottleneck: each request waits for the previous to complete

Proposed Solution

Solution

Use ThreadPoolExecutor to fetch multiple S3 objects concurrently:

with ThreadPoolExecutor(max_workers=None) as executor:
    future_to_key = {executor.submit(self._read_s3_object, key): key 
                     for key in message_keys}
    # Process results as they complete, maintaining order

Performance Improvement:

• 100 messages in parallel = ~50-200ms
• 10-100x speedup for large message histories

Implementation

• ✅ Added concurrent.futures imports
• ✅ Parallelized list_messages() using ThreadPoolExecutor
• ✅ Maintains message order (sorts by original index)
• ✅ Graceful error handling (individual failures don't stop operation)
• ✅ Backward compatible (no API changes)

Testing Checklist

• Test with various message counts (1, 10, 100, 1000)
• Verify message order is preserved
• Test error handling (missing files, network errors)
• Performance benchmarks (before/after comparison)

Use Case

any session with 5+ turns can see increased latency

Alternatives Solutions

No response

Additional Context

No response