infiniflow/ragflow · error · RuntimeError

HTTP {response.status_code}: {response.text}

Error message

HTTP {response.status_code}: {response.text}

What it means

Raised by SelfManagedProvider.execute_code() when the POST to {endpoint}/run returns any non-200 HTTP status. The full response body (response.text) is embedded in the RuntimeError, so the message carries the remote sandbox service's error detail — e.g. 400 bad payload, 413 code too large, 500 executor crash, 502/503 from a proxy in front of the sandbox service.

Source

Thrown at agent/sandbox/providers/self_managed.py:153

        # Normalize language
        normalized_lang = self._normalize_language(language)

        # Prepare request
        code_b64 = base64.b64encode(code.encode("utf-8")).decode("utf-8")
        payload = {"code_b64": code_b64, "language": normalized_lang, "arguments": arguments or {}}

        url = f"{self.endpoint}/run"
        exec_timeout = timeout or self.timeout

        start_time = time.time()

        try:
            response = requests.post(url, json=payload, timeout=exec_timeout, headers={"Content-Type": "application/json"})

            execution_time = time.time() - start_time

            if response.status_code != 200:
                raise RuntimeError(f"HTTP {response.status_code}: {response.text}")

            result = response.json()
            structured_result = result.get("result") or {}

            return ExecutionResult(
                stdout=result.get("stdout", ""),
                stderr=result.get("stderr", ""),
                exit_code=result.get("exit_code", 0),
                execution_time=execution_time,
                metadata={
                    "status": result.get("status"),
                    "time_used_ms": result.get("time_used_ms"),
                    "memory_used_kb": result.get("memory_used_kb"),
                    "detail": result.get("detail"),
                    "instance_id": instance_id,
                    "artifacts": result.get("artifacts", []),
                    "result_present": structured_result.get("present", False),
                    "result_value": structured_result.get("value"),

View on GitHub (pinned to 554fb1133a)

Solutions

1. Read the embedded response body in the exception message — it is the remote service's own error and usually names the real cause.
2. Verify the endpoint config points at the sandbox service's /run route (not a UI or proxy path) and the service is up (curl the health endpoint).
3. For 413, shrink the code payload or raise the proxy body-size limit; for 401/403 fix credentials/auth config.
4. Retry once on transient 502/503/504 (service restart), but fix persistent 4xx/5xx rather than retrying.

Example fix

# before
result = provider.execute_code(instance_id, code, "python")  # raises HTTP 500: ...

# after
try:
    result = provider.execute_code(instance_id, code, "python")
except RuntimeError as e:
    if "HTTP 5" in str(e) or "HTTP 502" in str(e):
        raise  # surface, or retry after service check
    raise

import requests
cfg = requests.get(f"{endpoint}/health", timeout=5)
assert cfg.status_code == 200, f"sandbox service unhealthy: {cfg.status_code}"

try:
    result = provider.execute_code(instance_id, code, "python")
except RuntimeError as e:
    msg = str(e)
    if msg.startswith("HTTP 5") or msg.startswith("HTTP 503"):
        # transient: backoff and retry once
        ...
    elif msg.startswith("HTTP 4"):
        raise SandboxRequestRejected(msg) from e  # do not retry 4xx
    else:
        raise

When it happens

Trigger: Sandbox service rejecting the request payload (invalid language, oversized code_b64); executor container crashing mid-run (500); the endpoint URL pointing at a reverse proxy that returns 404/502; auth middleware returning 401/403; service overloaded returning 503.

Common situations: Wrong SANDBOX endpoint/port in config (404 from a generic web server); payload too large for a proxy's client_max_body_size (413); sandbox executor bugs surfacing as 500; service starting up or restarting during the call.

Related errors

• Execution timed out after {exec_timeout} seconds
• HTTP request failed: {str(e)}
• Provider not initialized. Call initialize() first.
• Failed to create UCloud Agent Sandbox: {exc}
• UCloud Agent Sandbox execution failed: {exc}

AI-assisted analysis of infiniflow/ragflow@554fb1133a (2026-08-15). Data as JSON: /api/errors/91f6400069c74966. Report an issue: GitHub.