Open-source, secure environment with real-world tools for enterprise-grade agents.
Python agentaiai-agentai-agentscode-interpretercopilotdevelopmentdevtoolsgptgpt-4javascriptllmnextjsopenaipythonreactsoftwaretypescript
Cleanup log handling
Tomas Valenta5bfa0ab371a7
diff --git a/api-service/codegen/callbacks/log_parser.py b/api-service/codegen/callbacks/log_parser.pynew file mode 100644index 000000000..bdee4b606--- /dev/null+++ b/api-service/codegen/callbacks/log_parser.py@@ -0,0 +1,126 @@+import datetime+import uuid++from typing import Dict, List++from codegen.agent.base import parse_action_string, separate_thought_and_action+++class LogStreamParser:+ """+ To allow continuous streaming of partially finished logs we parse the logs every time we receive new token or tool output.+ We reparse only the tokens that corresponds to the buffered logs and not all the tokens that we have ever received.++ Buffered logs are partially finished logs that are either unfinished (still being streamed)+ or they are tools' logs for which we haven't received outputs yet.+ We can keep track of the buffered logs because there will always be more tool logs than tool outputs - we must parse the actions from tokens before they executed.+ Because of that all buffered logs can be flushed to finished logs when we receive output for the last tool in the buffered logs.++ We are matching the tools' logs and tools' outputs by the order in which they are parsed and ingested - if we need to support asynchronous execution of tools we need to change this matching.+ """++ def __init__(self) -> None:+ # All "finished" logs that the parser saved.+ self._logs: List[Dict[str, str]] = []++ self._token_buffer: str = ""++ # Logs have both `thought` logs and `tool` logs.+ # These logs can be partially parsed or they can have missing outputs.+ # They are flushed to self._logs after outputs for all tools in self._logs.buffer are ingested.+ self._logs_buffer: List[Dict[str, str]] = []++ # This is a list of tools' outputs that corresponds to the tool logs in the self._logs.buffer.+ self._tools_output_buffer: List[Dict[str, str]] = []+ # List of ids corresponding to the buffered logs.+ self._id_buffer: List[str] = []+ # List of timestamps corresponding to the buffered logs.+ self._start_timestamp_buffer: List[str] = []++ def _parse(self):+ """This function should be called only after ingesting new token or tool output to update the buffered logs."""+ thought, action_string = separate_thought_and_action(self._token_buffer)+ tools_logs = [+ {"type": "tool", "name": action.attrib["tool"], "input": action.text or ""}+ for action in parse_action_string(action_string)+ ]++ # Update tools' logs with the information from ingested tools' outputs.+ for i in range(len(self._tools_output_buffer)):+ tools_logs[i] = {+ **tools_logs[i],+ **self._tools_output_buffer[i],+ }++ # We overwrite the current buffered logs with their newer version.+ # If you ingested token or tool output this will save the change so you can call self.get_logs and get the new logs.+ self._logs_buffer = [+ {+ "type": "thought",+ "content": thought.removeprefix("Thought:")+ .replace("Action:", "")+ .strip(),+ },+ *tools_logs,+ ]++ # Add a new uuids to id buffer if we have less ids that there are logs in the logs buffer.+ self._id_buffer.extend(+ (+ str(uuid.uuid4())+ for _ in range(len(self._logs_buffer) - len(self._id_buffer))+ )+ )++ # Assign the stable ids to logs+ for log, id in zip(self._logs_buffer, self._id_buffer):+ log["id"] = id++ # Add a new timestamps to timestamps buffer if we have less timestamps that there are logs in the logs buffer.+ self._start_timestamp_buffer.extend(+ (+ str(datetime.datetime.now())+ for _ in range(+ len(self._logs_buffer) - len(self._start_timestamp_buffer)+ )+ )+ )++ # Assign the corresponding timestamps to logs+ for log, timestamp in zip(self._logs_buffer, self._start_timestamp_buffer):+ log["created_at"] = timestamp++ def ingest_token(self, token: str):+ """Ingest token and update the logs."""+ self._token_buffer += token+ self._parse()+ return self++ def ingest_tool_output(self, output: str):+ """Ingest output from tool and update the logs."""+ self._tools_output_buffer.append(+ {+ "finish_at": str(datetime.datetime.now()),+ "output": output,+ }+ )+ self._parse()++ # If we received output for the last tool in the buffered logs we save the buffered logs to logs and reset all buffers.+ # We know this is is the output for the last tool if the number of tool logs in buffered logs is equal to the number of buffered outputs.+ if len(self._tools_output_buffer) == len(+ [log for log in self._logs_buffer if log["type"] == "tool"]+ ):+ self._logs.extend(self._logs_buffer)+ self._token_buffer = ""+ self._logs_buffer = []+ self._tools_output_buffer = []+ self._id_buffer = []++ return self++ def get_logs(self):+ return [+ *self._logs,+ *self._logs_buffer,+ ]diff --git a/api-service/codegen/callbacks/log_queue.py b/api-service/codegen/callbacks/log_queue.pynew file mode 100644index 000000000..ce2bf6116--- /dev/null+++ b/api-service/codegen/callbacks/log_queue.py@@ -0,0 +1,27 @@+from asyncio import Queue, ensure_future+from typing import Coroutine+++class LogQueue:+ def __init__(self) -> None:+ self.queue: Queue[Coroutine] = Queue()+ self.work = ensure_future(self.worker())++ async def worker(self):+ while True:+ for _ in range(self.queue.qsize() - 1):+ old_coro = self.queue.get_nowait()+ try:+ old_coro.close()+ except Exception as e:+ print(e)++ task = await self.queue.get()+ try:+ await ensure_future(task)+ self.queue.task_done()+ except Exception as e:+ print(e)++ def close(self):+ self.work.cancel()diff --git a/api-service/codegen/callbacks/logs.py b/api-service/codegen/callbacks/logs.pyindex e432a8157..84117776c 100644--- a/api-service/codegen/callbacks/logs.py+++ b/api-service/codegen/callbacks/logs.py@@ -1,162 +1,15 @@-import datetime--import uuidimport asyncio-from asyncio import Queue, ensure_future-from typing import Coroutine, Dict, Any, List, Union++from typing import Dict, Any, List, Union+from codegen.callbacks.log_queue import LogQueuefrom langchain.callbacks.base import AsyncCallbackHandlerfrom langchain.schema import AgentAction, AgentFinish, LLMResultfrom pydantic import PrivateAttr-from codegen.agent.base import parse_action_string, separate_thought_and_action+from codegen.callbacks.log_parser import LogStreamParserfrom database import Database-class LogStreamParser:- """- To allow continuous streaming of partially finished logs we parse the logs every time we receive new token or tool output.- We reparse only the tokens that corresponds to the buffered logs and not all the tokens that we have ever received.-- Buffered logs are partially finished logs that are either unfinished (still being streamed)- or they are tools' logs for which we haven't received outputs yet.- We can keep track of the buffered logs because there will always be more tool logs than tool outputs - we must parse the actions from tokens before they executed.- Because of that all buffered logs can be flushed to finished logs when we receive output for the last tool in the buffered logs.-- We are matching the tools' logs and tools' outputs by the order in which they are parsed and ingested - if we need to support asynchronous execution of tools we need to change this matching.- """-- def __init__(self) -> None:- # All "finished" logs that the parser saved.- self._logs: List[Dict[str, str]] = []-- self._token_buffer: str = ""-- # Logs have both `thought` logs and `tool` logs.- # These logs can be partially parsed or they can have missing outputs.- # They are flushed to self._logs after outputs for all tools in self._logs.buffer are ingested.- self._logs_buffer: List[Dict[str, str]] = []-- # This is a list of tools' outputs that corresponds to the tool logs in the self._logs.buffer.- self._tools_output_buffer: List[Dict[str, str]] = []- # List of ids corresponding to the buffered logs.- self._id_buffer: List[str] = []- # List of timestamps corresponding to the buffered logs.- self._start_timestamp_buffer: List[str] = []-- def _parse(self):- """This function should be called only after ingesting new token or tool output to update the buffered logs."""- thought, action_string = separate_thought_and_action(self._token_buffer)- tools_logs = [- {"type": "tool", "name": action.attrib["tool"], "input": action.text or ""}- for action in parse_action_string(action_string)- ]-- # Update tools' logs with the information from ingested tools' outputs.- for i in range(len(self._tools_output_buffer)):- tools_logs[i] = {- **tools_logs[i],- **self._tools_output_buffer[i],- }-- # We overwrite the current buffered logs with their newer version.- # If you ingested token or tool output this will save the change so you can call self.get_logs and get the new logs.- self._logs_buffer = [- {- "type": "thought",- "content": thought.removeprefix("Thought:")- .replace("Action:", "")- .strip(),- },- *tools_logs,- ]-- # Add a new uuids to id buffer if we have less ids that there are logs in the logs buffer.- self._id_buffer.extend(- (- str(uuid.uuid4())- for _ in range(len(self._logs_buffer) - len(self._id_buffer))- )- )-- # Assign the stable ids to logs- for log, id in zip(self._logs_buffer, self._id_buffer):- log["id"] = id-- # Add a new timestamps to timestamps buffer if we have less timestamps that there are logs in the logs buffer.- self._start_timestamp_buffer.extend(- (- str(datetime.datetime.now())- for _ in range(- len(self._logs_buffer) - len(self._start_timestamp_buffer)- )- )- )-- # Assign the corresponding timestamps to logs- for log, timestamp in zip(self._logs_buffer, self._start_timestamp_buffer):- log["created_at"] = timestamp-- def ingest_token(self, token: str):- """Ingest token and update the logs."""- self._token_buffer += token- self._parse()- return self-- def ingest_tool_output(self, output: str):- """Ingest output from tool and update the logs."""- self._tools_output_buffer.append(- {- "finish_at": str(datetime.datetime.now()),- "output": output,- }- )- self._parse()-- # If we received output for the last tool in the buffered logs we save the buffered logs to logs and reset all buffers.- # We know this is is the output for the last tool if the number of tool logs in buffered logs is equal to the number of buffered outputs.- if len(self._tools_output_buffer) == len(- [log for log in self._logs_buffer if log["type"] == "tool"]- ):- self._logs.extend(self._logs_buffer)- self._token_buffer = ""- self._logs_buffer = []- self._tools_output_buffer = []- self._id_buffer = []-- return self-- def get_logs(self):- return [- *self._logs,- *self._logs_buffer,- ]---class LogQueue:- def __init__(self) -> None:- self.queue: Queue[Coroutine] = asyncio.Queue()- self.work = ensure_future(self.worker())-- async def worker(self):- while True:- for _ in range(self.queue.qsize() - 1):- old_coro = self.queue.get_nowait()- try:- old_coro.close()- except Exception as e:- print(e)-- task = await self.queue.get()- try:- await ensure_future(task)- self.queue.task_done()- except Exception as e:- print(e)-- def close(self):- self.work.cancel()--class LogsCallbackHandler(AsyncCallbackHandler):_database: Database = PrivateAttr()_run_id: str = PrivateAttr()