Skip to content

C
coderai
Commit bcd150e2 authored by Stefy Lanza (nextime / spora )'s avatar Stefy Lanza (nextime / spora )
Browse files
Options

Fix thinking display with timer thread and parse tool_call tags from content

parent 0c31d3fd
Hide whitespace changes
Inline Side-by-side
Showing with 144 additions and 103 deletions
coder
View file @ bcd150e2
......@@ -453,6 +453,7 @@ class CoderClient:
"""Handle streaming response from API."""
import time
import re
import threading
full_content = ""
tool_calls = []
......@@ -461,12 +462,27 @@ class CoderClient:
thinking_start_time = 0
last_elapsed = -1
thinking_line_printed = False
timer_thread = None
stop_timer = threading.Event()
def update_timer():
"""Background thread to update timer every second."""
nonlocal last_elapsed
while not stop_timer.is_set():
if in_thinking:
current_time = time.time()
elapsed = int(current_time - thinking_start_time)
if elapsed != last_elapsed:
last_elapsed = elapsed
print_thinking_line(elapsed, thinking_content)
time.sleep(0.5)
def print_thinking_line(elapsed, content, final=False):
"""Print or update the thinking line."""
nonlocal thinking_line_printed
# Filter out <tool> tags and newlines for display
# Filter out <tool> and <tool_call> tags and newlines for display
display_content = re.sub(r'<tool[^>]*>.*?</tool>', '', content, flags=re.DOTALL)
display_content = re.sub(r'<tool_call[^>]*>.*?</tool_call>', '', display_content, flags=re.DOTALL)
display_content = display_content.replace('\n', ' ').replace('\r', '')
display_content = display_content.strip()
......@@ -476,120 +492,145 @@ class CoderClient:
line = f"[{elapsed}s] Thinking: [{display_content}]"
if not thinking_line_printed:
# First time - just print
print(f"\r{Colors.DIM}{line}{Colors.RESET}", end='', flush=True)
thinking_line_printed = True
else:
# Overwrite existing line
# Clear to end of line and print new content
print(f"\r\033[K{Colors.DIM}{line}{Colors.RESET}", end='', flush=True)
# Clear line and print
print(f"\r\033[K{Colors.DIM}{line}{Colors.RESET}", end='', flush=True)
thinking_line_printed = True
if final:
print() # Move to new line
def parse_tool_calls_from_content(text):
"""Parse <tool_call> tags from content."""
pattern = r'<tool_call>\s*(\{[^}]+\})\s*</tool_call>'
matches = re.findall(pattern, text, re.DOTALL)
parsed = []
for match in matches:
try:
tool_data = json.loads(match)
parsed.append({
'id': f'call_{len(parsed)}',
'type': 'function',
'function': {
'name': tool_data.get('name', ''),
'arguments': json.dumps(tool_data.get('arguments', {}))
}
})
except json.JSONDecodeError:
continue
return parsed
# Start timer thread
timer_thread = threading.Thread(target=update_timer, daemon=True)
timer_thread.start()
# Use iter_content with smaller chunk size for better real-time handling
buffer = ""
last_timer_update = time.time()
for chunk in response.iter_content(chunk_size=256, decode_unicode=True):
current_time = time.time()
# Update thinking timer every second
if in_thinking:
elapsed = int(current_time - thinking_start_time)
if elapsed != last_elapsed and current_time - last_timer_update >= 1.0:
last_elapsed = elapsed
last_timer_update = current_time
print_thinking_line(elapsed, thinking_content)
if chunk:
buffer += chunk
# Process complete lines from buffer
while '\n' in buffer:
line, buffer = buffer.split('\n', 1)
# Handle SSE format
if line.startswith('data: '):
line = line[6:]
if line == '[DONE]':
break
if not line:
continue
try:
for chunk in response.iter_content(chunk_size=256, decode_unicode=True):
if chunk:
buffer += chunk
try:
data = json.loads(line)
delta = data.get('choices', [{}])[0].get('delta', {})
# Process complete lines from buffer
while '\n' in buffer:
line, buffer = buffer.split('\n', 1)
# Handle SSE format
if line.startswith('data: '):
line = line[6:]
if line == '[DONE]':
break
if not line:
continue
# Handle content
content = delta.get('content')
if content:
# Check for thinking tags
if '<think>' in content:
in_thinking = True
thinking_start_time = time.time()
last_timer_update = thinking_start_time
last_elapsed = 0
thinking_content = ""
thinking_line_printed = False
# Print initial thinking line
print_thinking_line(0, "")
continue
try:
data = json.loads(line)
delta = data.get('choices', [{}])[0].get('delta', {})
# Handle content
content = delta.get('content')
if content:
# Check for thinking tags
if '<think>' in content:
in_thinking = True
thinking_start_time = time.time()
last_elapsed = 0
thinking_content = ""
thinking_line_printed = False
# Print initial thinking line immediately
print_thinking_line(0, "")
continue
if in_thinking:
if '</think>' in content:
# End of thinking
in_thinking = False
parts = content.split('</think>', 1)
think_part = parts[0]
if think_part:
thinking_content += think_part
# Show final thinking line
elapsed = int(time.time() - thinking_start_time)
print_thinking_line(elapsed, thinking_content, final=True)
# Get content after </think>
actual_content = parts[1] if len(parts) > 1 else ""
if actual_content:
# Check for tool calls in the content
parsed_tools = parse_tool_calls_from_content(actual_content)
if parsed_tools:
tool_calls.extend(parsed_tools)
else:
print(actual_content, end='', flush=True)
full_content += actual_content
else:
# Still thinking - accumulate
thinking_content += content
# Update display every few chars
if len(thinking_content) % 10 == 0:
elapsed = int(time.time() - thinking_start_time)
print_thinking_line(elapsed, thinking_content)
else:
# Check for tool calls in normal content too
parsed_tools = parse_tool_calls_from_content(content)
if parsed_tools:
tool_calls.extend(parsed_tools)
else:
print(content, end='', flush=True)
full_content += content
if in_thinking:
if '</think>' in content:
# End of thinking
in_thinking = False
parts = content.split('</think>', 1)
think_part = parts[0]
if think_part:
thinking_content += think_part
# Handle tool calls (OpenAI format)
delta_tool_calls = delta.get('tool_calls')
if delta_tool_calls:
for tc in delta_tool_calls:
index = tc.get('index', 0)
# Show final thinking line
elapsed = int(current_time - thinking_start_time)
print_thinking_line(elapsed, thinking_content, final=True)
# Extend tool_calls list if needed
while len(tool_calls) <= index:
tool_calls.append({
'id': '',
'type': 'function',
'function': {'name': '', 'arguments': ''}
})
# Get content after </think>
actual_content = parts[1] if len(parts) > 1 else ""
if actual_content:
print(actual_content, end='', flush=True)
full_content += actual_content
else:
# Still thinking - accumulate
thinking_content += content
else:
# Normal content
print(content, end='', flush=True)
full_content += content
# Handle tool calls
delta_tool_calls = delta.get('tool_calls')
if delta_tool_calls:
for tc in delta_tool_calls:
index = tc.get('index', 0)
# Extend tool_calls list if needed
while len(tool_calls) <= index:
tool_calls.append({
'id': '',
'type': 'function',
'function': {'name': '', 'arguments': ''}
})
# Update tool call
if 'id' in tc:
tool_calls[index]['id'] = tc['id']
if 'function' in tc:
if 'name' in tc['function']:
tool_calls[index]['function']['name'] = tc['function']['name']
if 'arguments' in tc['function']:
tool_calls[index]['function']['arguments'] += tc['function']['arguments']
except json.JSONDecodeError:
continue
# Update tool call
if 'id' in tc:
tool_calls[index]['id'] = tc['id']
if 'function' in tc:
if 'name' in tc['function']:
tool_calls[index]['function']['name'] = tc['function']['name']
if 'arguments' in tc['function']:
tool_calls[index]['function']['arguments'] += tc['function']['arguments']
except json.JSONDecodeError:
continue
finally:
stop_timer.set()
if timer_thread:
timer_thread.join(timeout=1)
if in_thinking:
print() # End thinking line if still in thinking
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment