Fimeg/letta-server
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

fix: stream return sending full message after yielding chunks (#6295)

Browse Source 
base

Co-authored-by: Letta Bot <noreply@letta.com>
This commit is contained in:
jnjpng
2025-11-20 10:24:42 -08:00
committed by Caren Thomas
parent 6fe0971bfa
commit 32e4caf0d2
1 changed files with 11 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

21
letta/agents/letta_agent_v3.py
View File

@@ -638,16 +638,17 @@ class LettaAgentV3(LettaAgentV2):
self.response_messages.extend(aggregated_persisted[new_message_idx:])
self.response_messages_for_metadata.extend(aggregated_persisted[new_message_idx:]) # Track for job metadata
if llm_adapter.supports_token_streaming() and tool_calls:
# Stream each tool return if tools were executed
response_tool_returns = [msg for msg in aggregated_persisted if msg.role == "tool"]
for tr in response_tool_returns:
# Skip streaming for aggregated parallel tool returns (no per-call tool_call_id)
if tr.tool_call_id is None and tr.tool_returns:
continue
tool_return_letta = tr.to_letta_messages()[0]
if include_return_message_types is None or tool_return_letta.message_type in include_return_message_types:
yield tool_return_letta
if llm_adapter.supports_token_streaming():
if tool_calls:
# Stream each tool return if tools were executed
response_tool_returns = [msg for msg in aggregated_persisted if msg.role == "tool"]
for tr in response_tool_returns:
# Skip streaming for aggregated parallel tool returns (no per-call tool_call_id)
if tr.tool_call_id is None and tr.tool_returns:
continue
tool_return_letta = tr.to_letta_messages()[0]
if include_return_message_types is None or tool_return_letta.message_type in include_return_message_types:
yield tool_return_letta
else:
filter_user_messages = [m for m in aggregated_persisted[new_message_idx:] if m.role != "user"]
letta_messages = Message.to_letta_messages_from_list(