Sabo/webui-develop
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Phase 8: TypeScript migration, i18n rewrite, Activity Tree, Projects API, Heartbeats

Browse Source
This commit is contained in:
Rose
2026-04-29 11:50:00 +02:00
parent c705fad626
commit 255914c9f1
43 changed files with 17948 additions and 6899 deletions
Download Patch File Download Diff File Expand all files Collapse all files

47
api/streaming.py
View File

@@ -777,9 +777,13 @@ def _sse(handler, event, data):
def _run_agent_streaming(session_id, msg_text, model, workspace, stream_id, attachments=None, agent=None):
"""Run agent in background thread, writing SSE events to STREAMS[stream_id]."""
print(f'[DEBUG streaming] started stream_id={stream_id}', flush=True)
q = STREAMS.get(stream_id)
print(f'[DEBUG streaming] STREAMS keys={list(STREAMS.keys())}', flush=True)
if q is None:
print(f'[DEBUG streaming] queue is None for stream_id={stream_id}', flush=True)
return
print(f'[DEBUG streaming] queue found, agent={agent}', flush=True)
s = None
_rt = {}
old_cwd = None
@@ -937,12 +941,41 @@ def _run_agent_streaming(session_id, msg_text, model, workspace, stream_id, atta
_reasoning_text = '' # accumulates reasoning/thinking trace for persistence
_live_tool_calls = [] # tool progress fallback when final messages omit tool IDs
_token_buf = [] # token text buffer for batching
_token_buf_timer = None # threading.Timer reference
_token_buf_closed = False # True after sentinel seen
def _flush_token_buf():
nonlocal _token_buf_timer
if _token_buf_closed or not _token_buf:
return
# Grab and clear the buffer atomically
batch = ''.join(_token_buf)
_token_buf.clear()
# Cancel any pending timer
if _token_buf_timer is not None:
_token_buf_timer.cancel()
_token_buf_timer = None
# _buf_closed guard ensures we never put after sentinel
if not _token_buf_closed:
put('token', {'text': batch})
def on_token(text):
nonlocal _token_sent
nonlocal _token_sent, _token_buf_timer, _token_buf_closed
if text is None:
# Flush any remaining buffered tokens, then mark closed
_flush_token_buf()
_token_buf_closed = True
return # end-of-stream sentinel
_token_sent = True
put('token', {'text': text})
_token_buf.append(text)
if len(_token_buf) >= 20:
# Flush immediately on 20-token threshold
_flush_token_buf()
elif _token_buf_timer is None:
# Start 100ms debounce timer (only if not already pending)
_token_buf_timer = threading.Timer(0.1, _flush_token_buf)
_token_buf_timer.start()
def on_reasoning(text):
nonlocal _reasoning_text
@@ -1318,6 +1351,13 @@ def _run_agent_streaming(session_id, msg_text, model, workspace, stream_id, atta
if isinstance(_rm, dict) and _rm.get('role') == 'assistant':
_rm['reasoning'] = _reasoning_text
break
# Tag the last assistant message with per-turn token usage so the UI
# can display it on that specific message instead of the cumulative total.
if s.messages:
for _rm in reversed(s.messages):
if isinstance(_rm, dict) and _rm.get('role') == 'assistant':
_rm['_usage'] = {'in': input_tokens, 'out': output_tokens}
break
s.save()
# Sync to state.db for /insights (opt-in setting)
try:
@@ -1342,6 +1382,9 @@ def _run_agent_streaming(session_id, msg_text, model, workspace, stream_id, atta
usage['context_length'] = getattr(_cc, 'context_length', 0) or 0
usage['threshold_tokens'] = getattr(_cc, 'threshold_tokens', 0) or 0
usage['last_prompt_tokens'] = getattr(_cc, 'last_prompt_tokens', 0) or 0
# Send cumulative session totals separately so UI can label them as "session total"
usage['_session_input_tokens'] = s.input_tokens or 0
usage['_session_output_tokens'] = s.output_tokens or 0
# (reasoning trace already attached + saved above, before s.save())
raw_session = s.compact() | {'messages': s.messages, 'tool_calls': tool_calls}
put('done', {'session': redact_session_data(raw_session), 'usage': usage})