mem0ai/mem0 · error · NetworkError

NET_CONNECT

NET_CONNECT

Error message

Connection failed: {str(e)}

What it means

The client-side translation of httpx.ConnectError into NetworkError with error_code NET_CONNECT, produced by the api_error_handler decorator. It means the TCP/TLS connection to the Mem0 API could not be established at all; the message embeds httpx's original error string (e.g. DNS failure, connection refused).

Source

Thrown at mem0/client/utils.py:79

        status_code=e.response.status_code,
        response_text=response_text,
        details=error_details,
        debug_info=debug_info,
    )


def _handle_request_error(e):
    logger.error(f"Request error occurred: {e}")

    if isinstance(e, httpx.TimeoutException):
        raise NetworkError(
            message=f"Request timed out: {str(e)}",
            error_code="NET_TIMEOUT",
            suggestion="Please check your internet connection and try again",
            debug_info={"error_type": "timeout", "original_error": str(e)},
        )
    elif isinstance(e, httpx.ConnectError):
        raise NetworkError(
            message=f"Connection failed: {str(e)}",
            error_code="NET_CONNECT",
            suggestion="Please check your internet connection and try again",
            debug_info={"error_type": "connection", "original_error": str(e)},
        )
    else:
        raise NetworkError(
            message=f"Network request failed: {str(e)}",
            error_code="NET_GENERIC",
            suggestion="Please check your internet connection and try again",
            debug_info={"error_type": "request", "original_error": str(e)},
        )


def api_error_handler(func):
    """Decorator to handle API errors consistently.

    This decorator catches HTTP and request errors and converts them to

View on GitHub (pinned to 001c235229)

Solutions

1. Verify network egress to the API host (curl https://api.mem0.ai) from the same environment
2. Fix any custom base_url typo or remove it to use the default
3. Set HTTPS_PROXY/HTTP_PROXY if behind a corporate proxy
4. If offline intentionally, switch to the open-source self-hosted Memory class instead of the hosted client

Example fix

# before
client = MemoryClient(base_url="https://api.mem0.ai/", api_key=API_KEY)  # blocked network

# after
export HTTPS_PROXY=http://proxy.corp:8080
client = MemoryClient(api_key=API_KEY)

import socket

try:
    socket.getaddrinfo("api.mem0.ai", 443)
except OSError as e:
    raise RuntimeError(f"cannot reach api.mem0.ai: {e}; check DNS/proxy/egress")

def is_connect_error(err) -> bool:
    return getattr(err, "error_code", None) == "NET_CONNECT"

try:
    client.get_all(user_id=uid)
except NetworkError as e:
    if e.error_code == "NET_CONNECT":
        logger.error("no route to API: %s", e.debug_info.get("original_error"))
        raise SystemExit(2)  # config/network problem; retrying won't help
    raise

When it happens

Trigger: Any decorated API call when DNS resolution fails, the host is unreachable, a proxy refuses the connection, or TLS handshake fails. Commonly '[Errno -2] Name or service not known' or 'Connection refused' inside the message.

Common situations: Typos or custom base_url overrides; running in sandboxes/containers without DNS or egress; corporate proxies requiring HTTP_PROXY/HTTPS_PROXY env vars; firewall blocking api.mem0.ai; offline development.

Related errors

• Invalid response format from ping endpoint
• LiteLLM failed: ${message}
• NET_TIMEOUT
• NET_GENERIC
• HTTP ${resp.status}: ${detail}

AI-assisted analysis of mem0ai/mem0@001c235229 (2026-08-15). Data as JSON: /api/errors/7fa99db74abd10e7. Report an issue: GitHub.