superduper
superduper copied to clipboard
superduper-io
Superduper: Integrate AI models and machine learning workflows with your database to implement custom AI applications, without moving your data. Including streaming inference, scalable model hosting,...
Currently: ```python class MongoQuery(Query): patterns = ['.*find', ...] def __getattr__(self, item): # returns the find query ... def _execute_find(self): # executes the find query return ... ``` Proposal: separate into...
- `_Fittable` - AOB
This is, in our experience, a useful thing to have. Once the template is installed, use one or more of the queries to look at the results.
Currently identifiers of queries are built automatically, and are far too long. To do: optimize this.
In the `Cluster` refactoring https://github.com/superduper-io/superduper/pull/2434/commits, we created a base class `BaseBackend` which prevents multiple simultaneous active versions of a `Component`. We need to provide error handling, user feedback and warnings...
Create a function which defaults to doing nothing (or very little) in the OSS implementation which performs actions needed in the places where we currently send requests to services.
Example: ```json { "_base": "?test", "_builds": { "dill_lazy": { "_path": "superduper.components.datatype.get_serializer", "method": "dill", "encodable": "lazy_artifact" }, "1c94cb0a8bbabb82c63be6ac52d3a1edf4fe5d79": { "_path": "superduper.components.datatype.LazyArtifact", "datatype": "?dill_lazy", "blob": "&:blob:s3://" }, "test": { "_path": "superduper.components.model.ObjectModel", "object":...
```python query = db[''].find().limit('') q = query.set_variables(vlimit=1, input_table='coco') ``` The above fails with weird query placement.
Metadata
Owner
Metadata
Superduper: Integrate AI models and machine learning workflows with your database to implement custom AI applications, without moving your data. Including streaming inference, scalable model hosting,...