What causes "File {path} does not exist." in run-llama/llama_index?

Check paths before constructing: [p for p in files if Path(p).is_file()] or pre-resolve with Path(...).resolve()

run-llama/llama_index · error · ValueError

File {path} does not exist.

Error message

File {path} does not exist.

What it means

During SimpleDirectoryReader construction, each path in input_files is checked with fs.isfile(path); a path that is not an existing file raises ValueError('File {path} does not exist.'). This is eager validation so failures surface before any loading starts.

Source

Thrown at llama-index-core/llama_index/core/readers/file/base.py:290

        self.fs = fs or get_default_fs()
        self.errors = errors
        self.encoding = encoding

        self.exclude = exclude
        self.recursive = recursive
        self.exclude_hidden = exclude_hidden
        self.exclude_empty = exclude_empty
        self.required_exts = required_exts
        self.num_files_limit = num_files_limit
        self.raise_on_error = raise_on_error
        _Path = Path if is_default_fs(self.fs) else PurePosixPath

        if input_files:
            self.input_files = []
            for path in input_files:
                if not self.fs.isfile(path):
                    raise ValueError(f"File {path} does not exist.")
                input_file = _Path(path)
                self.input_files.append(input_file)
        elif input_dir:
            if not self.fs.isdir(input_dir):
                raise ValueError(f"Directory {input_dir} does not exist.")
            self.input_dir = _Path(input_dir)
            self.exclude = exclude
            self.input_files = self._add_files(self.input_dir)

        self.file_extractor = file_extractor or {}
        self.file_metadata = file_metadata or _DefaultFileMetadataFunc(self.fs)
        self.filename_as_id = filename_as_id

    def is_hidden(self, path: Path | PurePosixPath) -> bool:
        return any(
            part.startswith(".") and part not in [".", ".."] for part in path.parts
        )

View on GitHub (pinned to afd0fef371)

Solutions

Check paths before constructing: [p for p in files if Path(p).is_file()] or pre-resolve with Path(...).resolve()
Use absolute paths (e.g. build them from a project-root anchor) instead of cwd-relative ones
When using a custom fsspec fs, pass paths as that filesystem expects (scheme/keys), since fs.isfile is what validates

Example fix

// before
reader = SimpleDirectoryReader(input_files=["data/report.pdf"])  # cwd mismatch -> raises

// after
from pathlib import Path
root = Path(__file__).parent
files = [str(p) for p in (root / "data").glob("*.pdf") if p.is_file()]
reader = SimpleDirectoryReader(input_files=files)

Defensive patterns

Strategy: validation

Validate before calling

from pathlib import Path

files = [Path(p) for p in input_files]
missing = [p for p in files if not p.is_file()]
if missing:
    raise FileNotFoundError(f"missing input files: {missing}")
reader = SimpleDirectoryReader(input_files=[str(p) for p in files])

Try / catch

try:
    reader = SimpleDirectoryReader(input_files=files)
except ValueError as e:
    if "does not exist" in str(e):
        files = [f for f in files if Path(f).is_file()]
        if not files:
            raise
        reader = SimpleDirectoryReader(input_files=files)
    else:
        raise

Prevention

Resolve paths to absolute form at process start (anchor to project root or __file__)
Filter/validate file lists before constructing the reader
Remember the check runs in the constructor — failures are eager, so catch them at setup, not per-query

When it happens

Trigger: Passing input_files=["data/report.pdf"] where the file was moved/deleted, the relative path is resolved from a different working directory, or a remote fsspec filesystem (fs=...) is given local-style paths.

Common situations: Relative paths breaking when the process cwd differs (cron jobs, notebooks run from another directory), stale file lists, or using a custom fs while passing paths valid only on local disk.

Related errors

AI-assisted analysis of run-llama/llama_index@afd0fef371 (2026-08-15). Data as JSON: /api/errors/eaf8805622d1f94d. Report an issue: GitHub.