langflow-ai/langflow · error · HTTPException

Error ingesting via connector.

Error message

Error ingesting via connector.

What it means

Generic 500 raised by the connector-based knowledge-base ingestion endpoint (POST /api/v1/knowledge_bases/{kb_name}/ingest with a connector source). Any non-HTTP exception thrown while scheduling or launching the ingestion job (job creation, model_selection handling, job_service interaction) is caught, logged server-side as 'Error ingesting via connector to KB: <e>', and re-raised as this 500 with the original exception chained. The client message intentionally hides the root cause; the server log has it.

Source

Thrown at src/backend/base/langflow/api/v1/knowledge_bases.py:1926

            kb_path=kb_path,
            files_data=None,
            chunk_size=payload.chunk_size,
            chunk_overlap=payload.chunk_overlap,
            separator=payload.separator,
            source_name=payload.source_name,
            current_user=current_user,
            model_selection=model_selection,
            task_job_id=job_id,
            job_service=job_service,
            source=source,
        )
        return TaskResponse(id=str(job_id), href=f"/task/{job_id}")

    except HTTPException:
        raise
    except Exception as e:
        await logger.aerror("Error ingesting via connector to KB: %s", e)
        raise HTTPException(status_code=500, detail="Error ingesting via connector.") from e


@router.get("/{kb_name}/runs", status_code=HTTPStatus.OK)
async def list_ingestion_runs(
    kb_name: str,
    current_user: CurrentActiveUser,
    page: Annotated[int, Query(ge=1)] = 1,
    limit: Annotated[int, Query(ge=1, le=100)] = 50,
) -> PaginatedIngestionRunResponse:
    """Paginated list of ingestion runs for a KB (newest first).

    Scoped to the requesting user so one account can't observe
    another's run history. Returns counter-only rows; the UI fetches
    the detail endpoint for the drill-down.
    """
    _kb_guard = await _guard_kb_action(current_user=current_user, action=KnowledgeBaseAction.READ, kb_name=kb_name)
    # Verify the KB path exists + traversal-safe before exposing run
    # history — otherwise a crafted ``kb_name`` could be used to probe

View on GitHub (pinned to 976ec789d2)

Solutions

1. Read the server log line 'Error ingesting via connector to KB: ...' — the chained exception names the real cause
2. Verify the connector's credentials/configuration in the KB settings and re-test the connector standalone
3. Check the job service / task queue is reachable and healthy (other ingestion endpoints work?)
4. Reproduce with a minimal 1-file connector ingestion to isolate payload vs infrastructure issues

try:
    resp = await client.post(f"/api/v1/knowledge_bases/{kb}/ingest", json=payload)
except httpx.HTTPStatusError as e:
    if e.response.status_code == 500 and "connector" in e.response.json()["detail"]:
        # root cause only in server logs; surface retry option to user
        raise IngestionStartError(kb) from e
    raise

When it happens

Trigger: POST /api/v1/knowledge_bases/{kb_name}/ingest in connector mode where the code between job setup and TaskResponse creation raises: job_service failures, invalid connector configuration, DB errors inserting the job row, or an invalid model_selection that survives earlier validation.

Common situations: Connector credentials missing/expired (e.g. bad API key for the connector provider), job/queue backend unreachable or misconfigured, database locked or down, deploying a new connector type without registering its job handler.

Related errors

• Error ingesting files to knowledge base.
• Error deleting knowledge base.
• Error deleting knowledge bases.
• No ingestion job found for the knowledge base {kb_name}
• Cannot cancel job with status '{job_status}'

AI-assisted analysis of langflow-ai/langflow@976ec789d2 (2026-08-14). Data as JSON: /api/errors/c151ed0aaf5facff. Report an issue: GitHub.