README
¶
Model Data Extractor
Type: models-data-extractor
The Models Data Extractor converts the response from a models-data-source into endpoint attributes consumed by filters and scorers.
What it does
- Receives the parsed API response forwarded by
models-data-source. - Converts it into a
ModelDataCollection— a slice ofModelDataentries, each with:ID(string): model identifier (e.g."llama-3-8b").Parent(string, optional): base model the adapter derives from.
- Stores the collection as an attribute on the corresponding endpoint.
Attributes produced
ModelDataCollectionstored at attribute keyModelsAttributeKey("/v1/models") on each endpoint.
attr, ok := endpoint.GetAttributes().Get(models.ModelsAttributeKey)
if !ok || attr == nil {
return fmt.Errorf("no models found")
}
modelData, ok := attr.(models.ModelDataCollection)
Configuration
No configuration parameters.
Documentation
¶
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
Types ¶
type ModelExtractor ¶
type ModelExtractor struct {
// contains filtered or unexported fields
}
ModelExtractor implements the models extraction.
func NewModelExtractor ¶
func NewModelExtractor() *ModelExtractor
NewModelExtractor returns a new model extractor.
func (*ModelExtractor) Extract ¶
func (me *ModelExtractor) Extract(_ context.Context, in fwkdl.PollInput[*ModelResponse]) error
Extract stores the model list as an endpoint attribute.
func (*ModelExtractor) Produces ¶
func (me *ModelExtractor) Produces() map[fwkplugin.DataKey]any
Produces returns data produced by the producer.
func (*ModelExtractor) TypedName ¶
func (me *ModelExtractor) TypedName() fwkplugin.TypedName
TypedName returns the type and name of the ModelExtractor.
type ModelResponse ¶
type ModelResponse struct {
Object string `json:"object"`
Data []attrmodels.ModelData `json:"data"`
}
ModelResponse is the response from /v1/models API.
Source Files
Click to show internal directories.
Click to hide internal directories.