feat: add progress logging for v2 and v3

vision_agent/agent/agent.py

@@ -1,6 +1,6 @@
 from abc import ABC, abstractmethod
 from pathlib import Path
-from typing import Dict, List, Optional, Union
+from typing import Dict, List, Optional, Union, Any
 
 
 class Agent(ABC):
@@ -13,7 +13,7 @@ def __call__(
  pass
 
  @abstractmethod
- def log_progress(self, description: str) -> None:
+ def log_progress(self, data: Dict[str, Any]) -> None:
  """Log the progress of the agent.
  This is a hook that is intended for reporting the progress of the agent.
  """

vision_agent/agent/agent_coder.py

@@ -3,7 +3,7 @@
 import os
 import sys
 from pathlib import Path
-from typing import Dict, List, Optional, Union
+from typing import Dict, List, Optional, Union, Any
 
 from rich.console import Console
 from rich.syntax import Syntax
@@ -206,5 +206,5 @@ def chat(
 
  return f"{IMPORT_HELPER}\n{code}"
 
- def log_progress(self, description: str) -> None:
- _LOGGER.info(description)
+ def log_progress(self, data: Dict[str, Any]) -> None:
+ _LOGGER.info(data)

vision_agent/agent/vision_agent.py

@@ -451,7 +451,7 @@ def __init__(
  reflect_model: Optional[Union[LLM, LMM]] = None,
  max_retries: int = 2,
  verbose: bool = False,
- report_progress_callback: Optional[Callable[[str], None]] = None,
+ report_progress_callback: Optional[Callable[[Dict[str, Any]], None]] = None,
  ):
  """VisionAgent constructor.
 
@@ -518,23 +518,23 @@ def __call__(
  self_reflection=self_reflection,
  )
 
- def log_progress(self, description: str) -> None:
- _LOGGER.info(description)
+ def log_progress(self, data: Dict[str, Any]) -> None:
+ _LOGGER.info(data)
  if self.report_progress_callback:
- self.report_progress_callback(description)
+ self.report_progress_callback(data)
 
  def _report_visualization_via_callback(
  self, images: Sequence[Union[str, Path]]
  ) -> None:
  """This is intended for streaming the visualization images via the callback to the client side."""
  if self.report_progress_callback:
- self.report_progress_callback("<VIZ>")
+ self.report_progress_callback({"log": "<VIZ>"})
  if images:
  for img in images:
  self.report_progress_callback(
- f"<IMG>base:64{convert_to_b64(img)}</IMG>"
+ {"log": f"<IMG>base:64{convert_to_b64(img)}</IMG>"}
  )
- self.report_progress_callback("</VIZ>")
+ self.report_progress_callback({"log": "</VIZ>"})
 
  def chat_with_workflow(
  self,
@@ -618,8 +618,8 @@ def chat_with_workflow(
  tool_results["answer"] = answer
  all_tool_results.append(tool_results)
 
- self.log_progress(f"\tCall Result: {call_results}")
- self.log_progress(f"\tAnswer: {answer}")
+ self.log_progress({"log": f"\tCall Result: {call_results}"})
+ self.log_progress({"log": f"\tAnswer: {answer}"})
  answers.append({"task": task_str, "answer": answer})
  task_depend[task["id"]]["answer"] = answer # type: ignore
  task_depend[task["id"]]["call_result"] = call_results # type: ignore
@@ -644,18 +644,22 @@ def chat_with_workflow(
  final_answer,
  reflection_images,
  )
- self.log_progress(f"Reflection: {reflection}")
+ self.log_progress({"log": f"Reflection: {reflection}"})
  parsed_reflection = parse_reflect(reflection)
  if parsed_reflection["Finish"]:
  break
  else:
  reflections += "\n" + parsed_reflection["Reflection"]
  else:
- self.log_progress("Self Reflection skipped based on user request.")
+ self.log_progress(
+ {"log": "Self Reflection skipped based on user request."}
+ )
  break
  # '<ANSWER>' is a symbol to indicate the end of the chat, which is useful for streaming logs.
  self.log_progress(
- f"The Vision Agent has concluded this chat. <ANSWER>{final_answer}</ANSWER>"
+ {
+ "log": f"The Vision Agent has concluded this chat. <ANSWER>{final_answer}</ANSWER>"
+ }
  )
 
  if visualize_output:
@@ -718,8 +722,10 @@ def retrieval(
  }
 
  self.log_progress(
- f"""Going to run the following tool(s) in sequence:
+ {
+ "log": f"""Going to run the following tool(s) in sequence:
 {tabulate(tabular_data=[tool_results], headers="keys", tablefmt="mixed_grid", maxcolwidths=_MAX_TABULATE_COL_WIDTH)}"""
+ }
  )
 
  def parse_tool_results(result: Dict[str, Union[Dict, List]]) -> Any:
@@ -764,7 +770,9 @@ def create_tasks(
  else:
  task_list = []
  self.log_progress(
- f"""Planned tasks:
-{tabulate(task_list, headers="keys", tablefmt="mixed_grid", maxcolwidths=_MAX_TABULATE_COL_WIDTH)}"""
+ {
+ "log": "Planned tasks:",
+ "plan": task_list,
+ }
  )
  return task_list

vision_agent/agent/vision_agent_v2.py

@@ -165,7 +165,7 @@ def write_and_exec_code(
  tool_info: str,
  exec: Execute,
  retrieved_ltm: str,
- log_progress: Callable[..., str],
+ log_progress: Callable[[Dict[str, Any]], None],
  max_retry: int = 3,
  verbosity: int = 0,
 ) -> Tuple[bool, str, str, Dict[str, List[str]]]:
@@ -179,7 +179,23 @@ def write_and_exec_code(
  success, result = exec.run_isolation(code)
  if verbosity == 2:
  _CONSOLE.print(Syntax(code, "python", theme="gruvbox-dark", line_numbers=True))
- log_progress(f"\tCode success: {success}\n\tResult: {str(result)}", code)
+ log_progress(
+ {
+ "log": f"Code success: {success}",
+ }
+ )
+ log_progress(
+ {
+ "log": "Code:",
+ "code": code,
+ }
+ )
+ log_progress(
+ {
+ "log": "Result:",
+ "result": str(result),
+ }
+ )
  _LOGGER.info(f"\tCode success: {success}, result: {str(result)}")
  working_memory: Dict[str, List[str]] = {}
  while not success and counter < max_retry:
@@ -206,7 +222,18 @@ def write_and_exec_code(
  _CONSOLE.print(
  Syntax(code, "python", theme="gruvbox-dark", line_numbers=True)
  )
- log_progress(f"\tDebugging reflection: {reflection}\n\tResult: {result}")
+ log_progress(
+ {
+ "log": "Debugging reflection:",
+ "reflection": reflection,
+ }
+ )
+ log_progress(
+ {
+ "log": "Result:",
+ "result": result,
+ }
+ )
  _LOGGER.info(f"\tDebugging reflection: {reflection}, result: {result}")
 
  if success:
@@ -227,7 +254,7 @@ def run_plan(
  exec: Execute,
  code: str,
  tool_recommender: Sim,
- log_progress: Callable[..., str],
+ log_progress: Callable[[Dict[str, Any]], None],
  long_term_memory: Optional[Sim] = None,
  verbosity: int = 0,
 ) -> Tuple[str, str, List[Dict[str, Any]], Dict[str, List[str]]]:
@@ -239,8 +266,7 @@ def run_plan(
 
  for task in active_plan:
  log_progress(
- f"""Going to run the following task(s) in sequence:
-{tabulate(tabular_data=[task], headers="keys", tablefmt="mixed_grid", maxcolwidths=_MAX_TABULATE_COL_WIDTH)}"""
+ {"log": "Going to run the following task(s) in sequence:", "task": task}
  )
  _LOGGER.info(
  f"""
@@ -250,7 +276,7 @@ def run_plan(
  tool_info = "\n".join([e["doc"] for e in tools])
 
  if verbosity == 2:
- log_progress(f"Tools retrieved: {[e['desc'] for e in tools]}")
+ log_progress({"log": f"Tools retrieved: {[e['desc'] for e in tools]}"})
  _LOGGER.info(f"Tools retrieved: {[e['desc'] for e in tools]}")
 
  if long_term_memory is not None:
@@ -282,7 +308,17 @@ def run_plan(
  Syntax(code, "python", theme="gruvbox-dark", line_numbers=True)
  )
 
- log_progress(f"\tCode success: {success}\n\tResult: {str(result)}")
+ log_progress(
+ {
+ "log": f"Code success: {success}",
+ }
+ )
+ log_progress(
+ {
+ "log": "Result:",
+ "result": str(result),
+ }
+ )
  _LOGGER.info(f"\tCode success: {success} result: {str(result)}")
 
  task["success"] = success
@@ -320,7 +356,7 @@ def __init__(
  tool_recommender: Optional[Sim] = None,
  long_term_memory: Optional[Sim] = None,
  verbosity: int = 0,
- report_progress_callback: Optional[Callable[..., Any]] = None,
+ report_progress_callback: Optional[Callable[[Dict[str, Any]], None]] = None,
  ) -> None:
  self.planner = OpenAILLM(temperature=0.0, json_mode=True)
  self.coder = OpenAILLM(temperature=0.0)
@@ -376,8 +412,10 @@ def chat_with_workflow(
 
  user_req, plan = write_plan(chat, plan, TOOL_DESCRIPTIONS, self.planner)
  self.log_progress(
- f"""Plan:
-{tabulate(tabular_data=plan, headers="keys", tablefmt="mixed_grid", maxcolwidths=_MAX_TABULATE_COL_WIDTH)}"""
+ {
+ "log": "Plans:",
+ "plan": plan,
+ }
  )
  _LOGGER.info(
  f"""Plan:
@@ -412,8 +450,12 @@ def chat_with_workflow(
 
  retries += 1
 
- self.log_progress("The Vision Agent V2 has concluded this chat.")
- self.log_progress(f"<ANSWER>Plan success: {success}</ANSWER>")
+ self.log_progress(
+ {
+ "log": f"The Vision Agent V2 has concluded this chat.\nSuccess: {success}",
+ "finished": True,
+ }
+ )
 
  return {
  "code": working_code,
@@ -423,7 +465,7 @@ def chat_with_workflow(
  "plan": plan,
  }
 
- def log_progress(self, description: str, code: Optional[str] = "") -> None:
+ def log_progress(self, data: Dict[str, Any]) -> None:
  if self.report_progress_callback is not None:
- self.report_progress_callback(description, code)
+ self.report_progress_callback(data)
  pass