Capture Loguru log args as parameters

[sentrivana]

Right now the Loguru Sentry logs integration doesn't capture args, kwargs provided to the logger, only the already formatted message. This is because Loguru pre-formats the message and we only have access to the already formatted message in the handler.

We've submitted a feature request to Loguru to potentially get access to the args & kwargs but maybe there's another way to get this to work?

Monkeypatching isn't straightforward because all the logic is in one function and the log record is also created there.

Goals:

• support extras
• support contextualize
• update the Logs docs
• update the Loguru docs

[lookpl]

Hi, I found a workaround without touching Loguru code - adding a custom sink with patched loguru_sentry_logs_handler. It appears that Sentry already has an access to extras in record. I'm not sure if my patch covers all cases, but at least I have an access to attributes in the dashboard.

def patched_loguru_sentry_logs_handler(message):
    ... (copy original code from loguru_sentry_logs_handler)

    # PATCH: Add extra parameters from the log record as structured attributes
    extra = record.get("extra", {})
    if extra:
        for key, value in extra.items():
            # Convert the value to string to ensure it's serializable
            try:
                if isinstance(value, (str, int, float, bool)):
                    attrs[f"sentry.message.parameter.{key}"] = value
                else:
                    # For complex objects, convert to string representation
                    attrs[f"sentry.message.parameter.{key}"] = str(value)
            except Exception:
                # Fallback for any serialization issues
                attrs[f"sentry.message.parameter.{key}"] = f"<{type(value).__name__}>"

    # continue with updated attrs
    client._capture_experimental_log(
        {
            "severity_text": otel_severity_text,
            "severity_number": otel_severity_number,
            "body": record["message"],
            "attributes": attrs,
            "time_unix_nano": int(record["time"].timestamp() * 1e9),
            "trace_id": None,
        }
    )

Then, I've added a new sink to Loguru after all configurations.

if LoguruIntegration.sentry_logs_level is not None:
    logger.add(
        patched_loguru_sentry_logs_handler,
        level=LoggingLevels.INFO.value,
    )

Also, some adjustments are needed to remove auto-registered Sentry sink from Loguru. My quick workaround was to set extremely high logging level so it's never called, while my sink method has a lower threshold.

[antonpirker]

Maybe related: https://discord.com/channels/621778831602221064/1415404794151178250/1415404794151178250

[Rishabhg71]

Maybe related: https://discord.com/channels/621778831602221064/1415404794151178250/1415404794151178250

My issue might be similar but isn't exactly the same. I'm contextualizing the logger with specific values using the following code:

with logger.contextualize(conversation_id=conversation_id, session_start_time=time.monotonic()):
    ...

Even though I'm using the correct values, my log format displays both values: <cyan>[{extra[conversation_id]}]</cyan> | <yellow>[{extra[elapsed_seconds]}]</yellow> | <level>{level: <8}</level> | <level>{message}</level>{exception}. However, Sentry does not capture the formatted logs and only shows the log message.