Instrument `print`

[teocns]

captures all stdout

[braelyn-ai]

capture log output over time and tie to spans on flush. (what new logs have been printed since the last span was sent)

[braelyn-ai]

if we can instrument the python logger, we can capture much more value. then we need to find/build a way to translate logs from logger to OTEL_log format. then send that to the collector

[braelyn-ai]

I'm going to look into storing logs in otel_logs vs otel_traces. both methods have merit, what is the best for us?

[teocns]

it will automatically store logs in the table if you configure a LoggingProvider in the SDK (alongside MeterProvider), nothing else to do

[braelyn-ai]

https://github.com/open-telemetry/opentelemetry-python/blob/e5a93079d173c5385d37085db6339d3bccc54736/docs/examples/logs/README.rst#L2

otel-python has support for logs but they are experimental. i believe strongly we should use the otel_logs table and record logs independent of spans. instead, we can maybe have a trace_id based on the active trace in the sdk at the time of the log

[braelyn-ai]

example with otel-python

[braelyn-ai]

after conversation with @areibman, we're going to halt an OTEL implementation for now. Existing work done for OTEL is on the branch otel_logs and can easily be picked back up.

in the meantime, instead of using OTEL, we're simply creating a tmp logfile with the trace_id as the name, then after the trace finishes, we upload as multi-part to the api, which in turn stores it in S3