bunware.org
forked from
cameron.stream/void
this repo has no description
Implement streaming for Letta API calls to prevent 524 timeouts
- Replace synchronous agents.messages.create with create_stream
- Add step streaming with max_steps=100 for better responsiveness
- Increase client timeout from 5 to 10 minutes
- Add condensed info-level logging for streaming chunks with emojis
- Add debug-level logging for full chunk details
- Maintain backward compatibility with existing response handling
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
cameron.stream
+36
-4
bsky.py
+36
-4
bsky.py
···
58
58
# Create a client with extended timeout for LLM operations
59
59
CLIENT= Letta(
60
60
token=os.environ["LETTA_API_KEY"],
61
-
timeout=300 # 5 minutes timeout for API calls
61
+
timeout=600 # 10 minutes timeout for API calls - higher than Cloudflare's 524 timeout
62
62
)
63
63
64
64
# Use the "Bluesky" project
···
313
313
logger.info(f"💬 Sending to LLM: @{author_handle} mention | msg: \"{mention_text[:50]}...\" | context: {len(thread_context)} chars, {thread_handles_count} users")
314
314
315
315
try:
316
-
message_response = CLIENT.agents.messages.create(
317
-
agent_id = void_agent.id,
318
-
messages = [{"role":"user", "content": prompt}]
316
+
# Use streaming to avoid 524 timeout errors
317
+
message_stream = CLIENT.agents.messages.create_stream(
318
+
agent_id=void_agent.id,
319
+
messages=[{"role": "user", "content": prompt}],
320
+
stream_tokens=False, # Step streaming only (faster than token streaming)
321
+
max_steps=100
319
322
)
323
+
324
+
# Collect the streaming response
325
+
all_messages = []
326
+
for chunk in message_stream:
327
+
# Log condensed chunk info
328
+
if hasattr(chunk, 'message_type'):
329
+
if chunk.message_type == 'reasoning_message':
330
+
logger.info(f"🧠 Reasoning: {chunk.reasoning[:100]}...")
331
+
elif chunk.message_type == 'tool_call_message':
332
+
logger.info(f"🔧 Tool call: {chunk.tool_call.name}({chunk.tool_call.arguments[:50]}...)")
333
+
elif chunk.message_type == 'tool_return_message':
334
+
logger.info(f"📋 Tool result: {chunk.name} - {chunk.status}")
335
+
elif chunk.message_type == 'assistant_message':
336
+
logger.info(f"💬 Assistant: {chunk.content[:100]}...")
337
+
else:
338
+
logger.info(f"📨 {chunk.message_type}: {str(chunk)[:100]}...")
339
+
else:
340
+
logger.info(f"📦 Stream status: {chunk}")
341
+
342
+
# Log full chunk for debugging
343
+
logger.debug(f"Full streaming chunk: {chunk}")
344
+
all_messages.append(chunk)
345
+
if str(chunk) == 'done':
346
+
break
347
+
348
+
# Convert streaming response to standard format for compatibility
349
+
message_response = type('StreamingResponse', (), {
350
+
'messages': [msg for msg in all_messages if hasattr(msg, 'message_type')]
351
+
})()
320
352
except Exception as api_error:
321
353
import traceback
322
354
error_str = str(api_error)