zylon-ai/private-gpt · critical · FileNotFoundError

Settings file not found for profile '{profile}'. Searched in

Error message

Settings file not found for profile '{profile}'. Searched in folders: {_settings_folders} with file name '{profile_file_name}'

What it means

Raised by load_settings_from_profile when none of the configured settings folders contain the profile's YAML file. PrivateGPT resolves profiles to file names ('default' -> settings.yaml, 'override' -> settings.override.yaml, other -> settings-<profile>.yaml) and searches every folder in PGPT_SETTINGS_FOLDER (comma-separated, defaulting to the project root). If no candidate file exists anywhere, the profile cannot be loaded and the app aborts during settings bootstrap.

Source

Thrown at private_gpt/settings/settings_loader.py:72

    else:
        profile_file_name = f"settings-{profile}.yaml"

    config: dict[str, Any] = {}
    found = False
    for settings_folder in _settings_folders:
        path = Path(settings_folder) / profile_file_name
        if not path.is_file():
            continue
        with Path(path).open("r") as f:
            raw = load_yaml_with_envvars(f)
        if not isinstance(raw, dict):
            raise TypeError(f"Config file has no top-level mapping: {path}")
        config = raw
        found = True
        break

    if not found:
        raise FileNotFoundError(
            f"Settings file not found for profile '{profile}'. "
            f"Searched in folders: {_settings_folders} with file name '{profile_file_name}'"
        )

    return config


@typing.no_type_check
def discover_models_from_environment(
    environ: dict[str, Any] = os.environ,
) -> list[dict[str, Any]]:
    """Discover model configurations from environment variables.

    This function parses environment variables with the pattern:
    PGPT_MODELS_<MODEL_ID>_<PARAMETER>[__<NESTED_PARAM>]

    Examples:
        PGPT_MODELS_GPT4_API_KEY=sk-123 ->

View on GitHub (pinned to 4a030776a3)

Solutions

1. Create settings.yaml (copy settings.sample.yaml if present) in the project root or in a folder listed in the error message
2. Fix PGPT_SETTINGS_FOLDER so every comma-separated entry is an existing absolute path containing the profile file
3. Check the profile naming rule: PGPT_PROFILES=myprofile requires settings-myprofile.yaml exactly (hyphen, not underscore)
4. Verify the running working directory / container WORKDIR matches where the settings files were mounted
5. If running tests, ensure a settings-test.yaml exists or remove the fixture that injects the 'test' profile

Example fix

# before
export PGPT_PROFILES=prod
# FileNotFoundError: Settings file not found for profile 'prod' ...

# after: create the correctly named file in a searched folder
cp settings.yaml settings-prod.yaml
# or point the loader at the right folder
export PGPT_SETTINGS_FOLDER=/app/config

import os
from pathlib import Path

def resolve_profile_file(profile: str) -> Path:
    name = ("settings.yaml" if profile == "default"
            else "settings.override.yaml" if profile == "override"
            else f"settings-{profile}.yaml")
    folders = [f.strip() for f in os.environ.get("PGPT_SETTINGS_FOLDER", ".").split(",") if f.strip()]
    for folder in folders:
        p = Path(folder) / name
        if p.is_file():
            return p
    raise SystemExit(f"missing {name}; searched {folders}")

null

try:
    cfg = load_settings_from_profile(profile)
except FileNotFoundError as e:
    # log searched folders from the message, create file or fix PGPT_SETTINGS_FOLDER
    raise SystemExit(f"settings bootstrap failed: {e}") from e

When it happens

Trigger: Starting private-gpt (or importing its Settings) when settings.yaml is missing; setting PGPT_PROFILES=prod without creating settings-prod.yaml in any searched folder; setting PGPT_SETTINGS_FOLDER to a wrong or misspelled path; running from a working directory where the default settings folder does not contain settings.yaml; tests importing tests.fixtures which adds the 'test' profile without a settings-test.yaml.

Common situations: Fresh clone where the user renamed or deleted settings.yaml; Docker/CI image that does not copy settings files in; a typo in PGPT_SETTINGS_FOLDER or in the profile name passed via PGPT_PROFILES; deploying with a profile file named settings_prod.yaml (underscore) instead of settings-prod.yaml (hyphen).

Related errors

• Default model '{self._default_model_id}' not found in regist
• Config file has no top-level mapping: {path}
• Environment variable {env_var} is not set and not default wa
• Default LLM model '{model_id}' could not be initialized: {e}
• No model specified and no models are configured

AI-assisted analysis of zylon-ai/private-gpt@4a030776a3 (2026-08-15). Data as JSON: /api/errors/339bdf2ca75c03eb. Report an issue: GitHub.