Enable exporting spans to snowflake stage if a TruLensSnowflakeSpanExporter is provided
#1708
Conversation
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
sfc-gh-gtokernliang
force-pushed
the
garett/SNOW-1854070
branch
from
83c4d06 to
c6a8462
Compare
sfc-gh-gtokernliang
force-pushed
the
garett/SNOW-1871920
branch
2 times, most recently
from
8d2cece to
b4bbdbf
Compare
sfc-gh-gtokernliang
changed the title
SnowflakeConnector is provided
dosubot
bot
added
the
size:L
tests/unit/static/golden/test_otel_instrument__test_instrument_decorator.csv
Outdated
Show resolved
Hide resolved
| @@ -972,6 +970,22 @@ async def __aexit__(self, exc_type, exc_value, exc_tb): | |||
|
|
|||
| return | |||
|
|
|||
| def __call__(self, *, run_name: str, input_id: str): | |||
| if not self.session.experimental_feature( | |||
There was a problem hiding this comment.
sorry can you explain to me what's going on in this function?
There was a problem hiding this comment.
haha - this might not have been the best way to achieve this.
The goal is to enable syntax that looks like:
with custom_app(run_name='run_name', input_id='input_id') as recording:
test_app.respond_to_query(input)to achieve this - custom_app(run_name='run_name', input_id='input_id') should return an object that is a recording context (in our case a class with the __enter__ and __exit__ methods), and that the setting of the run_name and input_id shouldn't taint the object itself.
hence - the approach here is, when custom_app is called, it:
- Checks that OTEL tracing is enabled
- creates a clone of itself using
model_construct+model_dumpthat is classed toOTELRecordingContext(see https://github.com/truera/trulens/pull/1708/files/5a555840da38b280a0b3c3cbabb40cbd8a3e93c4#diff-25f7f4954d77bc71539bcd8d1e7b3133552596b468bf53a6962b636b176de682R96) - then performs the normal recording context duties
sfc-gh-gtokernliang
force-pushed
the
garett/add-utils
branch
from
61d87a8 to
9e5b60b
Compare
| if not core_instruments.Instrument._have_context(): | ||
| raise RuntimeError(core_endpoint._NO_CONTEXT_WARNING) | ||
|
|
||
| def _prevent_invalid_otel_syntax(self): |
There was a problem hiding this comment.
why did you name this function this? Like what do you mean by otel tracing syntax?
There was a problem hiding this comment.
basically trying to force the user to use syntax that looks like:
with app():
...or
with app(run_name=run_name, input_id=input_id):
...as opposed to the conventional
with app:
...this is because we need to be able to return a different object in order to avoid the possibility of polluting the App
sfc-gh-gtokernliang
changed the title
SnowflakeConnector is providedTruLensSnowflakeSpanExporter is provided
| # For use as a context manager. | ||
| def __enter__(self): | ||
| if not core_instruments.Instrument._have_context(): | ||
| raise RuntimeError(core_endpoint._NO_CONTEXT_WARNING) |
There was a problem hiding this comment.
we want to force the user to use syntax that looks like:
with app():
...which will allow us to return a wrapping class around app to hold the OTEL properties, to avoid the risk of OTEL properties being mixed between invocations
| if not core_instruments.Instrument._have_context(): | ||
| raise RuntimeError(core_endpoint._NO_CONTEXT_WARNING) | ||
|
|
||
| def _prevent_invalid_otel_syntax(self): |
There was a problem hiding this comment.
basically trying to force the user to use syntax that looks like:
with app():
...or
with app(run_name=run_name, input_id=input_id):
...as opposed to the conventional
with app:
...this is because we need to be able to return a different object in order to avoid the possibility of polluting the App
dosubot
bot
added
size:XXL
| # There might be nested records - in those scenarios use the outer one. | ||
| if not get_baggage(SpanAttributes.RECORD_ID): | ||
| self.attach_to_context(SpanAttributes.RECORD_ID, otel_record_id) | ||
|
|
||
| self.attach_to_context(SpanAttributes.APP_NAME, self.app.app_name) |
There was a problem hiding this comment.
Can you do the same as RECORD_ID For the other attributes? Since I guess the APP_NAME and other stuff aren't thread safe since if you call the child LLM app then continue working the baggage will get changed on you.
There was a problem hiding this comment.
changed attach_to_context to check for existence + ensure that value is non null
Description
Enable exporting spans to a snowflake stage if
SnowflakeConnectoris provided.Other details good to know for developers
N/A
Type of change
not work as expected)
Important
Add
TruLensSnowflakeSpanExporterto export spans to Snowflake stage and refactor OTEL tracing inTruSession.TruLensSnowflakeSpanExporterinexporter/snowflake.pyto export spans to Snowflake stage._export_to_snowflake_stagemethod for exporting spans as protobuf files.span_contextandtokensfromAppclass inapp.py._prevent_invalid_otel_syntaxinapp.py.convert_to_any_valuetoexporter/utils.py.TruSessionto handleSpanExporterandTracerProviderinsession.py.experimental_force_flushmethod toTruSessionfor flushing spans.test_exporter.pyandtest_otel_instrument.pyto reflect new exporter functionality.api.trulens.3.11.yamlandapi.trulens_eval.3.11.yaml.This description was created by
for 4e27038. It will automatically update as commits are pushed.