commit-0/commit0

no such file: 'commit0/repos/pyjwt/spec.pdf'

chenqianhe opened this issue · 3 comments

chenqianhe commented

I see spec.pdf.bz2, seems to lack unzip

RemoteTraceback: 
"""
Traceback (most recent call last):
  File "/Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/multiprocessing/pool.py", line 125, in worker
    result = (True, func(*args, **kwds))
                    ^^^^^^^^^^^^^^^^^^^
  File "/Users/chenqianhe/PycharmProjects/venv/commit0/lib/python3.12/site-packages/agent/run_agent.py", line 132, in run_agent_for_repo
    message = get_message(agent_config, repo_path, test_file=test_file)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/chenqianhe/PycharmProjects/venv/commit0/lib/python3.12/site-packages/agent/agent_utils.py", line 255, in get_message
    + get_specification(specification_pdf_path=Path(repo_path, "spec.pdf"))[
      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/chenqianhe/PycharmProjects/venv/commit0/lib/python3.12/site-packages/agent/agent_utils.py", line 271, in get_specification
    document = fitz.open(specification_pdf_path)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/chenqianhe/PycharmProjects/venv/commit0/lib/python3.12/site-packages/pymupdf/__init__.py", line 2777, in __init__
    raise FileNotFoundError(msg)
pymupdf.FileNotFoundError: no such file: '/Users/chenqianhe/PycharmProjects/commit0/repos/pyjwt/spec.pdf'
wenting-zhao commented

Hi! This is addressed in 72552d4e6b45974715d004ae72e4eb81ae12af83.

Please pull the latest main or run pip install -U commit0.

I tested with

agent config aider --use-spec-info
agent run fillin-spec

It seems to work fine for me. When I checked vim logs/agent/simpy/fillin-spec/2024-09-28_15-01-36/src__simpy__resources__resource/.aider.chat.history.md The spec information is included in the chat.

Let me know if you run into more errors!

chenqianhe commented

I can run it normally.

btw, spec.pdf is generated by you, why not save it in markdown or other formats that are easier for programs to read?
At the same time, can the content of spec be optimized, such as removing useless information(e.g. version update records)?

wenting-zhao commented

The point of saving specs in PDFs is because they has richer information such as figures. We believe that reading specs is a multi-modal problem. For example, an overview figure would help human software engineers to understand the overall structure of a repo. As another example, figures in seaborn also help with understanding how the end deliverable code should look like. We understand that this is very out of scope at the moment, but we are planning for the future.