File size: 9,046 Bytes
22087d0 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 | import os
import json
import asyncio
import time
from datetime import datetime, timezone
from typing import Optional, Dict, Any
import structlog
from core.utils.logger import logger
from core.services import redis
from core.services.langfuse import langfuse
from core.utils.tool_output_streaming import (
set_tool_output_streaming_context,
clear_tool_output_streaming_context,
)
from core.agents.runner.config import AgentConfig
from core.agents.runner.services import (
REDIS_STREAM_TTL_SECONDS,
STOP_CHECK_INTERVAL,
stream_status_message,
check_terminating_tool_call,
update_agent_run_status,
send_completion_notification,
)
async def execute_agent_run(
agent_run_id: str,
thread_id: str,
project_id: str,
model_name: str,
agent_config: dict,
account_id: str,
cancellation_event: asyncio.Event,
is_new_thread: bool = False,
user_message: Optional[str] = None
) -> None:
execution_start = time.time()
structlog.contextvars.clear_contextvars()
structlog.contextvars.bind_contextvars(agent_run_id=agent_run_id, thread_id=thread_id)
logger.info(f"π Executing agent run: {agent_run_id}")
stop_checker = None
final_status = "failed"
stream_key = f"agent_run:{agent_run_id}:stream"
try:
from datetime import datetime, timezone
start_time = datetime.now(timezone.utc)
await stream_status_message("initializing", "Starting execution...", stream_key=stream_key)
await redis.verify_stream_writable(stream_key)
try:
await redis.expire(stream_key, REDIS_STREAM_TTL_SECONDS)
except Exception:
pass
from core.ai_models import model_manager
effective_model = model_manager.resolve_model_id(model_name)
stop_state = {'received': False, 'reason': None}
async def check_stop():
while not stop_state['received']:
try:
if cancellation_event.is_set():
stop_state['received'] = True
stop_state['reason'] = 'cancellation_event'
logger.info(f"π Stop detected via cancellation_event for {agent_run_id}")
break
if await redis.check_stop_signal(agent_run_id):
stop_state['received'] = True
stop_state['reason'] = 'stop_signal'
cancellation_event.set()
logger.info(f"π Stop detected via Redis for {agent_run_id}")
break
await asyncio.sleep(STOP_CHECK_INTERVAL)
except asyncio.CancelledError:
break
except Exception:
await asyncio.sleep(5.0)
stop_checker = asyncio.create_task(check_stop())
set_tool_output_streaming_context(agent_run_id=agent_run_id, stream_key=stream_key)
from core.agents.pipeline.context import PipelineContext
from core.agents.pipeline.stateless import StatelessCoordinator
ctx = PipelineContext(
agent_run_id=agent_run_id,
thread_id=thread_id,
project_id=project_id,
account_id=account_id,
model_name=effective_model,
agent_config=agent_config,
is_new_thread=is_new_thread,
cancellation_event=cancellation_event,
stream_key=stream_key,
user_message=user_message
)
coordinator = StatelessCoordinator()
first_response = False
complete_tool_called = False
total_responses = 0
stream_ttl_set = False
error_message = None
async for response in coordinator.execute(ctx):
if cancellation_event.is_set() or stop_state['received']:
logger.warning(f"π Agent run stopped: {stop_state.get('reason', 'cancellation_event')}")
final_status = "stopped"
error_message = f"Stopped by {stop_state.get('reason', 'cancellation_event')}"
break
if not first_response:
first_response_time_ms = (time.time() - execution_start) * 1000
logger.info(f"β±οΈ FIRST RESPONSE: {first_response_time_ms:.1f}ms")
first_response = True
try:
timing_msg = {
"type": "timing",
"first_response_ms": round(first_response_time_ms, 1),
"timestamp": datetime.now(timezone.utc).isoformat(),
"pipeline": "stateless"
}
await redis.stream_add(stream_key, {"data": json.dumps(timing_msg)}, maxlen=200, approximate=True)
except Exception:
pass
from core.services.db import serialize_row
if isinstance(response, dict):
response = serialize_row(response)
if response.get('type') == 'assistant':
metadata = response.get('metadata', '')
if isinstance(metadata, str) and 'complete' in metadata:
logger.debug(f"[STREAM] Sending assistant complete: message_id={response.get('message_id')}")
try:
await redis.stream_add(stream_key, {"data": json.dumps(response)}, maxlen=200, approximate=True)
if not stream_ttl_set:
try:
await asyncio.wait_for(redis.expire(stream_key, REDIS_STREAM_TTL_SECONDS), timeout=2.0)
stream_ttl_set = True
except:
pass
except Exception as e:
logger.warning(f"Failed to write to stream: {e}")
total_responses += 1
terminating = check_terminating_tool_call(response)
if terminating == 'complete':
complete_tool_called = True
if response.get('type') == 'status':
status = response.get('status')
if status in ['completed', 'failed', 'stopped', 'error']:
final_status = status if status != 'error' else 'failed'
if status in ['failed', 'error']:
error_message = response.get('message') or response.get('error')
logger.error(f"Agent run error: {error_message}")
break
if response.get('type') == 'error':
final_status = 'failed'
error_message = response.get('error')
logger.error(f"Agent run error: {error_message}")
break
if final_status == "failed" and not error_message:
final_status = "completed"
duration = (datetime.now(timezone.utc) - start_time).total_seconds()
logger.info(f"Agent run completed (duration: {duration:.2f}s, responses: {total_responses})")
completion_msg = {"type": "status", "status": "completed", "message": "Completed successfully"}
try:
await redis.stream_add(stream_key, {'data': json.dumps(completion_msg)}, maxlen=200, approximate=True)
except:
pass
await send_completion_notification(thread_id, agent_config, complete_tool_called)
if stop_state['reason']:
final_status = "stopped"
await update_agent_run_status(agent_run_id, final_status, error=error_message, account_id=account_id)
logger.info(f"β
Agent run completed: {agent_run_id} | status={final_status}")
except Exception as e:
logger.error(f"Error in agent run {agent_run_id}: {e}", exc_info=True)
await update_agent_run_status(agent_run_id, "failed", error=str(e), account_id=account_id)
finally:
from core.utils.lifecycle_tracker import log_cleanup_error
cleanup_errors = []
try:
clear_tool_output_streaming_context()
except Exception as e:
log_cleanup_error(agent_run_id, "streaming_context", e)
cleanup_errors.append(f"streaming_context: {e}")
if stop_checker and not stop_checker.done():
try:
stop_checker.cancel()
await stop_checker
except asyncio.CancelledError:
pass
except Exception as e:
log_cleanup_error(agent_run_id, "stop_checker", e)
cleanup_errors.append(f"stop_checker: {e}")
try:
await redis.expire(stream_key, REDIS_STREAM_TTL_SECONDS)
except Exception as e:
log_cleanup_error(agent_run_id, "redis_expire", e)
cleanup_errors.append(f"redis_expire: {e}")
if cleanup_errors:
logger.error(
f"[LIFECYCLE] CLEANUP_ERRORS agent_run={agent_run_id} "
f"count={len(cleanup_errors)} errors={cleanup_errors}"
)
|