HKUDS/DeepTutor · error · LLMAPIError

Network error connecting to {binding} API: {e}

Error message

Network error connecting to {binding} API: {e}

What it means

The generic branch of the aiohttp.ClientError handler in _openai_complete: any client-side transport failure that is not a 'forcibly closed' reset (DNS failure, connection refused, SSL errors, timeouts surfacing as ClientError) is wrapped as LLMAPIError with status_code=0 and the original exception text. The `from e` chain preserves the underlying cause for debugging.

Source

Thrown at deeptutor/services/llm/cloud_provider.py:442

                                )
                    else:
                        raise LLMAPIError(
                            f"OpenAI API error: {error_text}",
                            status_code=resp.status,
                            provider=binding or "openai",
                        )
        except aiohttp.ClientError as e:
            # Handle connection errors with more specific messages
            if "forcibly closed" in str(e).lower() or "10054" in str(e):
                raise LLMAPIError(
                    f"Connection to {binding} API was forcibly closed. "
                    "This may indicate network issues or server-side problems. "
                    "Please check your internet connection and try again.",
                    status_code=0,
                    provider=binding or "openai",
                ) from e
            else:
                raise LLMAPIError(
                    f"Network error connecting to {binding} API: {e}",
                    status_code=0,
                    provider=binding or "openai",
                ) from e

    if content is not None:
        # Clean thinking tags from response using unified utility
        return clean_thinking_tags(content, binding, model)

    raise LLMConfigError("Cloud completion failed: no valid configuration")


async def _openai_stream(
    model: str,
    prompt: str,
    system_prompt: str,
    api_key: str | None,
    base_url: str | None,

View on GitHub (pinned to 3e82f13042)

Solutions

1. Inspect the chained cause (`e.__cause__`) to identify DNS/refused/SSL precisely.
2. Verify the service is up: curl the base_url; for local backends, start Ollama/vLLM first.
3. Unset or correct HTTP_PROXY/HTTPS_PROXY if a proxy intercepts the call.
4. For SSL issues on private gateways, install the CA cert into the trust store rather than disabling verification.

Example fix

// before
resp = await complete(prompt=p, model="llama3", base_url="http://localhost:11434")

# after
# ensure Ollama is running first: `ollama serve`
resp = await complete(prompt=p, model="llama3", base_url="http://localhost:11434/v1")

import socket
host = urlparse(base_url or "https://api.openai.com").hostname
socket.getaddrinfo(host, 443)  # DNS sanity check before the call

try:
    out = await complete(prompt=p, model=m, base_url=u)
except LLMAPIError as e:
    if e.status_code == 0 and e.__cause__ is not None:
        log.warning("transport failure: %r", e.__cause__)
        # distinguish DNS vs refused vs TLS from the cause type
    raise

When it happens

Trigger: DNS cannot resolve the provider host; endpoint port not listening (local Ollama/vLLM not started); TLS certificate verification failure; proxy misconfiguration since trust_env=True honors HTTP_PROXY/HTTPS_PROXY.

Common situations: Local LLM server not running before the call; corporate proxy env vars routing the API call wrongly; self-signed certs on a private gateway; container missing CA certificates.

Related errors

• Connection to {binding} API was forcibly closed. This may in
• OpenAI SDK connection error: {exc}
• result.error or "Could not connect to the LightRAG server."
• Channel onboarding provider request failed ({type(exc).__nam
• This model is not assigned to your account.

AI-assisted analysis of HKUDS/DeepTutor@3e82f13042 (2026-08-27). Data as JSON: /api/errors/2ffcf4a6a9cba0bd. Report an issue: GitHub.