Publisher agent "directory not found" when using multi agent approach with local docs

[emilmirzayev]

Hi. I have previously faced several issues with multi_agent approach. #684 has solved one of them. But now there is another issue when using local docs as knowledge source and not following guidelines. It creates the folder based on the task but publisher_agent throws this error when trying to save the report:

PUBLISHER: Publishing final research report based on retrieved data...
Traceback (most recent call last):
  File "C:\Users\Emil\Desktop\gpt-researcher\multi_agents\main.py", line 62, in <module>
    asyncio.run(main())
  File "C:\Users\Emil\AppData\Local\Programs\Python\Python311\Lib\asyncio\runners.py", line 190, in run
    return runner.run(main)
           ^^^^^^^^^^^^^^^^
  File "C:\Users\Emil\AppData\Local\Programs\Python\Python311\Lib\asyncio\runners.py", line 118, in run
    return self._loop.run_until_complete(task)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\Emil\AppData\Local\Programs\Python\Python311\Lib\asyncio\base_events.py", line 653, in run_until_complete
    return future.result()
           ^^^^^^^^^^^^^^^
  File "C:\Users\Emil\Desktop\gpt-researcher\multi_agents\main.py", line 57, in main
    research_report = await chief_editor.run_research_task()
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\Emil\Desktop\gpt-researcher\multi_agents\agents\master.py", line 65, in run_research_task
    result = await chain.ainvoke({"task": self.task})
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\Emil\Desktop\gpt-researcher\RESEARCHER_ENV\Lib\site-packages\langgraph\pregel\__init__.py", line 1316, in ainvoke
    async for chunk in self.astream(
  File "C:\Users\Emil\AppData\Local\Programs\Python\Python311\Lib\contextlib.py", line 222, in __aexit__
    await self.gen.athrow(typ, value, traceback)
  File "C:\Users\Emil\AppData\Local\Programs\Python\Python311\Lib\contextlib.py", line 222, in __aexit__
    await self.gen.athrow(typ, value, traceback)
  File "C:\Users\Emil\Desktop\gpt-researcher\RESEARCHER_ENV\Lib\site-packages\langgraph\channels\base.py", line 57, in afrom_checkpoint
    yield value
  File "C:\Users\Emil\Desktop\gpt-researcher\RESEARCHER_ENV\Lib\site-packages\langgraph\channels\manager.py", line 37, in AsyncChannelsManager
    yield {
  File "C:\Users\Emil\Desktop\gpt-researcher\RESEARCHER_ENV\Lib\site-packages\langgraph\pregel\__init__.py", line 1195, in astream
    _panic_or_proceed(done, inflight, loop.step, asyncio.TimeoutError)
  File "C:\Users\Emil\Desktop\gpt-researcher\RESEARCHER_ENV\Lib\site-packages\langgraph\pregel\__init__.py", line 1349, in _panic_or_proceed
    raise exc
  File "C:\Users\Emil\Desktop\gpt-researcher\RESEARCHER_ENV\Lib\site-packages\langgraph\pregel\executor.py", line 123, in done
    task.result()
  File "C:\Users\Emil\Desktop\gpt-researcher\RESEARCHER_ENV\Lib\site-packages\langgraph\pregel\retry.py", line 74, in arun_with_retry
    await task.proc.ainvoke(task.input, task.config)
  File "C:\Users\Emil\Desktop\gpt-researcher\RESEARCHER_ENV\Lib\site-packages\langchain_core\runnables\base.py", line 2911, in ainvoke
    input = await step.ainvoke(input, config, **kwargs)
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\Emil\Desktop\gpt-researcher\RESEARCHER_ENV\Lib\site-packages\langgraph\utils.py", line 124, in ainvoke
    ret = await asyncio.create_task(
          ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\Emil\Desktop\gpt-researcher\multi_agents\agents\publisher.py", line 133, in run
    final_research_report = await self.publish_research_report(research_state, publish_formats)
                            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\Emil\Desktop\gpt-researcher\multi_agents\agents\publisher.py", line 90, in publish_research_report
    await self.write_report_by_formats(layout, publish_formats)
  File "C:\Users\Emil\Desktop\gpt-researcher\multi_agents\agents\publisher.py", line 124, in write_report_by_formats
    await write_text_to_md(layout, self.output_dir)
  File "C:\Users\Emil\Desktop\gpt-researcher\multi_agents\agents\publisher.py", line 156, in write_text_to_md      
    await write_to_file(md_path, md_content)
  File "C:\Users\Emil\Desktop\gpt-researcher\multi_agents\agents\publisher.py", line 141, in write_to_file
    async with aiofiles.open(filename, "w", encoding='utf-8') as file:
  File "C:\Users\Emil\Desktop\gpt-researcher\RESEARCHER_ENV\Lib\site-packages\aiofiles\base.py", line 63, in __aenter__
    return await self
           ^^^^^^^^^^
  File "C:\Users\Emil\Desktop\gpt-researcher\RESEARCHER_ENV\Lib\site-packages\aiofiles\base.py", line 59, in __await__
    self._obj = yield from self._coro.__await__()
                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\Emil\Desktop\gpt-researcher\RESEARCHER_ENV\Lib\site-packages\aiofiles\threadpool\__init__.py", line 92, in _open
    f = await loop.run_in_executor(executor, cb)
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\Emil\AppData\Local\Programs\Python\Python311\Lib\concurrent\futures\thread.py", line 58, in run   
    result = self.fn(*self.args, **self.kwargs)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
FileNotFoundError: [Errno 2] No such file or directory: './outputs/run_1722601499_Your task is to write a small report on \\report.md'

So, I said maybe I can solve it by explicitly creating the folder (although I thought it was already there as i could see it in outputs folder). Came up with this in publisher.py:

import os
import aiofiles

from .utils.file_formats import (
    write_md_to_pdf,
    write_md_to_word,
    write_text_to_md,
)

from .utils.views import print_agent_output


class PublisherAgent:
    def __init__(self, output_dir: str, websocket=None, stream_output=None, headers=None):
        self.websocket = websocket
        self.stream_output = stream_output
        self.output_dir = output_dir
        self.headers = headers or {}
        
        # Ensure the output directory exists
        os.makedirs(self.output_dir, exist_ok=True)
        
    async def publish_research_report(self, research_state: dict, publish_formats: dict):
        layout = self.generate_layout(research_state)
        await self.write_report_by_formats(layout, publish_formats)
        return layout

    def generate_layout(self, research_state: dict):
        sections = '\n\n'.join(f"{value}"
                                 for subheader in research_state.get("research_data")
                                 for key, value in subheader.items())
        references = '\n'.join(f"{reference}" for reference in research_state.get("sources"))
        headers = research_state.get("headers")
        layout = f"""# {headers.get('title')}
#### {headers.get("date")}: {research_state.get('date')}

## {headers.get("introduction")}
{research_state.get('introduction')}

## {headers.get("table_of_contents")}
{research_state.get('table_of_contents')}

{sections}

## {headers.get("conclusion")}
{research_state.get('conclusion')}

## {headers.get("references")}
{references}
"""
        return layout

    async def write_report_by_formats(self, layout: str, publish_formats: dict):
        if publish_formats.get("pdf"):
            await write_md_to_pdf(layout, self.output_dir)
        if publish_formats.get("docx"):
            await write_md_to_word(layout, self.output_dir)
        if publish_formats.get("markdown"):
            await write_text_to_md(layout, self.output_dir)

    async def run(self, research_state: dict):
        task = research_state.get("task")
        publish_formats = task.get("publish_formats")
        if self.websocket and self.stream_output:
            await self.stream_output("logs", "publishing", f"Publishing final research report based on retrieved data...", self.websocket)
        else:
            print_agent_output(output="Publishing final research report based on retrieved data...", agent="PUBLISHER")
        final_research_report = await self.publish_research_report(research_state, publish_formats)
        return {"report": final_research_report}


# utils/file_formats.py
async def write_to_file(filename: str, text: str):
    # Ensure the directory exists
    os.makedirs(os.path.dirname(filename), exist_ok=True)
    async with aiofiles.open(filename, "w", encoding='utf-8') as file:
        await file.write(text)

async def write_md_to_pdf(md_content: str, output_dir: str):
    # Assume a function that converts markdown to PDF and writes to a file
    pdf_path = os.path.join(output_dir, 'report.pdf')
    await write_to_file(pdf_path, md_content)

async def write_md_to_word(md_content: str, output_dir: str):
    # Assume a function that converts markdown to Word and writes to a file
    word_path = os.path.join(output_dir, 'report.docx')
    await write_to_file(word_path, md_content)

async def write_text_to_md(md_content: str, output_dir: str):
    md_path = os.path.join(output_dir, 'report.md')
    await write_to_file(md_path, md_content)

However, it still gives the same error. it does output the text, but the error appears to be when publishing to a file.

My task.json:

{
  "query": "Your task is to write a small report on Working capital structure of Pernod Ricard. the years are from 2012 to 2023 but 2020 and 2015 are missing. covid and economic instability issues with the numbers. Write about main components and their interplay. Do not mention the table numbers separately. They will be given as one big table and you will reference the table itself.",
  "max_sections": 1,
  "publish_formats": {
    "markdown": true,
    "pdf": false,
    "docx": false
  },
  "source": "local",
  "follow_guidelines": false,
  "model": "gpt-4o-mini",
  "guidelines": [
"Something"
  ],
  "verbose": true
}

I also tried to solve it by editing file_formats.py in utils like this, but error persisted.

async def write_to_file(filename: str, text: str) -> None:
    """Asynchronously write text to a file in UTF-8 encoding.

    Args:
        filename (str): The filename to write to.
        text (str): The text to write.
    """
    # Ensure the directory exists
    os.makedirs(os.path.dirname(filename), exist_ok=True)

    # Convert text to UTF-8, replacing any problematic characters
    text_utf8 = text.encode('utf-8', errors='replace').decode('utf-8')

    async with aiofiles.open(filename, "w", encoding='utf-8') as file:
        await file.write(text_utf8)

I did my best based on my limited understanding. If someone has some ideas, i welcome them.

[emilmirzayev]

@ruzicka02 i think it happened with the recent couple changes, maybe you face this one too.

[assafelovic]

Hey @emilmirzayev is this still happening with latest updates? Has this occurred on several research queries?

[emilmirzayev]

Hi @assafelovic , I will try with updated library and report here.

[ElishaKay]

Welcome @emilmirzayev

I had a similar issue yesterday & found that it was fixed by leveraging absolute paths.

Have a look at the main.py edit in this commit that was merged into master yesterday.

Perhaps there's a key there for your issue as well.

P.S. The path of least resistance that we've found to work best for users is to run via Docker:

Flow here

Running via Docker with that flow will enable you to trigger multi_agents (with local docs as a knowledge source) via localhost:3000 - should be a bit smoother & more intuitive.

Hope it helps.

May the force be with you