zylon-ai/private-gpt · error · ImportError

OpenAI Responses LLM dependencies are not installed. Install

Error message

OpenAI Responses LLM dependencies are not installed. Install with `uv sync --inexact --extra llm-openai`.

What it means

ImportError raised by _load_openai_responses_base when llama_index.llms.openai cannot be imported or has no OpenAIResponses attribute. It fires at module import time (the base class is loaded immediately when TYPE_CHECKING is off), so merely importing the module with the dependency missing crashes. The message names the required extra: llm-openai.

Source

Thrown at private_gpt/components/llm/custom/openairesponses.py:39

        OpenAIResponses as OpenAIResponsesBase,
    )
    from openai.types.responses import (  # ty:ignore[unresolved-import]
        ResponseFunctionToolCall,
    )

logger = logging.getLogger(__name__)


def _load_openai_responses_base() -> type[Any]:
    try:
        return cast(
            type[Any],
            importlib.import_module("llama_index.llms.openai").OpenAIResponses,
        )
    except (ImportError, AttributeError) as e:
        from private_gpt.utils.dependencies import format_missing_dependency_message

        raise ImportError(
            format_missing_dependency_message(
                "OpenAI Responses LLM",
                extras="llm-openai",
            )
        ) from e


if not TYPE_CHECKING:
    OpenAIResponsesBase = _load_openai_responses_base()

_RESPONSES_REASONING_EFFORT_MAP = {
    ReasoningEffort.LOW: "low",
    ReasoningEffort.MEDIUM: "medium",
    ReasoningEffort.HIGH: "high",
    ReasoningEffort.MAX: "high",
}

View on GitHub (pinned to 4a030776a3)

Solutions

Run: uv sync --inexact --extra llm-openai.
Confirm the class exists in your installed version; if not, align llama-index version with the project's lockfile.
If you do not use OpenAI Responses, ensure the module is not imported (check profile wiring / lazy import guards).
Rebuild containers so the new dependency is baked in.

Example fix

# before
uv sync --inexact

# after
uv sync --inexact --extra llm-openai

Defensive patterns

Strategy: try-catch

Validate before calling

def openai_responses_available() -> bool:
    try:
        import llama_index.llms.openai  # noqa: F401
        return hasattr(llama_index.llms.openai, 'OpenAIResponses')
    except ImportError:
        return False

Try / catch

try:
    from private_gpt.components.llm.custom.openairesponses import PatchedOpenAIResponsesLLM
except ImportError as e:
    if 'llm-openai' in str(e):
        raise RuntimeError('run: uv sync --inexact --extra llm-openai') from e

Prevention

Because this loads at import time, a missing extra crashes any importer — keep module imports lazy behind profile selection.
Pin llama-index to the lockfile version; the AttributeError branch indicates a version mismatch.

When it happens

Trigger: Installing or running the app without the llm-openai extra while any code path imports private_gpt.components.llm.custom.openairesponses — commonly when the configured LLM profile uses OpenAI Responses.

Common situations: Slim installs for local/Ollama-only deployments that still import the module; CI linting importing all modules; llama-index version where OpenAIResponses was renamed/removed (AttributeError branch).

Related errors

AI-assisted analysis of zylon-ai/private-gpt@4a030776a3 (2026-08-15). Data as JSON: /api/errors/c9269ef8c2cf6f00. Report an issue: GitHub.