feat: plumb Dataset start/end times through graphQL #194
Conversation
nate-mar
changed the title
src/phoenix/core/model.py
Outdated
| from phoenix.datasets import Dataset | ||
| from phoenix.datasets.schema import EmbeddingFeatures | ||
| from phoenix.core.datasets import Dataset | ||
| from phoenix.core.datasets.schema import EmbeddingFeatures |
There was a problem hiding this comment.
I think I sorta wanted the code that the user interacts with be unnested and not be part of core. Core being part of the application. I think voxel has a similar organization. Makes for easier discovery in the notebook. But we would dissuade people from importing core in the notebook
There was a problem hiding this comment.
Ah ok gotcha -- so datasets is the API interface so to speak, and users should never need to interact with anything in core.
There was a problem hiding this comment.
Yeah that was the idea. That the user interacts with datasets and metrics, phoenix serves core via an api.
nate-mar
changed the title
app/schema.graphql
Outdated
| startTime: DateTime | ||
| endTime: DateTime |
There was a problem hiding this comment.
| startTime: DateTime | |
| endTime: DateTime | |
| startTime: DateTime! | |
| endTime: DateTime! |
Let's make these non-maybe
| readonly endTime: any | null; | ||
| readonly name: string; | ||
| readonly startTime: any | null; |
There was a problem hiding this comment.
update relay.config and these will become strings.
There was a problem hiding this comment.
err, what do I update relay.config with? do I need to add something?
app/relay.config.js
Outdated
| @@ -7,5 +7,6 @@ module.exports = { | |||
| noFutureProofEnums: true, | |||
| customScalars: { | |||
| GlobalID: "String", | |||
| "DateTime": "string", | |||
There was a problem hiding this comment.
It's mapping to another base primitive so it should be uppercase String
src/phoenix/datasets/dataset.py
Outdated
| def start_time(self) -> datetime: | ||
| """Returns the datetime of the earliest inference in the dataset""" | ||
| ts_col_name: str = cast(str, self.schema.timestamp_column_name) | ||
| dt: datetime = self.__dataframe[ts_col_name].min() |
There was a problem hiding this comment.
Since this never changes you can use dynamic programming to compute once
There was a problem hiding this comment.
I'm assuming you're referring to the caching aspect of dp? how about just using the @cached_property annotation? or do we have a preexisting convention for how we'd like to do this?
There was a problem hiding this comment.
Oh that's great. Very cool
src/phoenix/datasets/dataset.py
Outdated
| @property | ||
| def end_time(self) -> datetime: | ||
| """Returns the datetime of the latest inference in the dataset""" | ||
| ts_col_name: str = cast(str, self.schema.timestamp_column_name) |
There was a problem hiding this comment.
Prefer dynamic programming to compute once
src/phoenix/datasets/dataset.py
Outdated
| """Returns the datetime of the latest inference in the dataset""" | ||
| ts_col_name: str = cast(str, self.schema.timestamp_column_name) | ||
| dt: datetime = self.__dataframe[ts_col_name].max() | ||
| return dt |
There was a problem hiding this comment.
Been advocating for non abbreviated variable names right now https://youtu.be/-J3wNP6u5YU
Resolves #172