Wire model-config-datasource into the runner and helm chart.

cmd/runner/runner.go

@@ -47,6 +47,7 @@ import (
 	"github.com/llm-d/llm-d-inference-payload-processor/pkg/framework/interface/datalayer/datasource"
 	"github.com/llm-d/llm-d-inference-payload-processor/pkg/framework/interface/plugin"
 	"github.com/llm-d/llm-d-inference-payload-processor/pkg/framework/interface/requesthandling"
+	modelconfigcollector "github.com/llm-d/llm-d-inference-payload-processor/pkg/framework/plugins/datalayer/modelconfigcollector"
 	requestmetadata "github.com/llm-d/llm-d-inference-payload-processor/pkg/framework/plugins/datalayer/requestmetadata"
 	"github.com/llm-d/llm-d-inference-payload-processor/pkg/framework/plugins/modelselector/picker/maxscore"
 	"github.com/llm-d/llm-d-inference-payload-processor/pkg/framework/plugins/modelselector/picker/random"
@@ -277,6 +278,7 @@ func (r *Runner) registerInTreePlugins() {
 	plugin.Register(bodyfieldtoheader.BodyFieldToHeaderPluginType, bodyfieldtoheader.BodyFieldToHeaderPluginFactory)
 	plugin.Register(basemodelextractor.BaseModelToHeaderPluginType, basemodelextractor.BaseModelToHeaderPluginFactory)
 	plugin.Register(requestmetadata.PluginType, requestmetadata.ExtractorFactory)
+	plugin.Register(modelconfigcollector.PluginType, modelconfigcollector.DatasourceFactory)
 	// register model selector plugins
 	plugin.Register(random.RandomPickerType, random.RandomPickerFactory)
 	plugin.Register(maxscore.MaxScorePickerType, maxscore.MaxScorePickerFactory)

config/charts/payload-processor/README.md

@@ -86,6 +86,17 @@ provider:
 
 > **Tip:** Only enable events your plugins need. Each extra event adds a network hop between the proxy and the payload processor.
 
+### Configure Model List (model-config-datasource)
+
+`payloadProcessor.listModels` is a list of model names that the IPP should be aware of. When set, the chart renders a `models.json` file into the ConfigMap, which is mounted at `/config/models.json` inside the pod. The `model-config-datasource` plugin reads this file to populate the IPP datastore and watches it for live updates on ConfigMap remounts.
+
+```yaml
+payloadProcessor:
+  listModels:
+    - meta-llama/Llama-3.1-8B-Instruct
+    - meta-llama/Llama-3.1-70B-Instruct
+```
+
 ### Uninstall
 
 Run the following command to uninstall the chart:
@@ -111,6 +122,7 @@ The following table list the configurable parameters of the chart.
 | `payloadProcessor.image.pullPolicy`       | Image pull policy for the container. Possible values: `Always`, `IfNotPresent`, or `Never`. Defaults to `Always`.                                                                                                                                                                                |
 | `payloadProcessor.flags`                  | map of flags which are passed through to the payload processor. Refer to [runner.go](https://github.com/llm-d/llm-d-inference-payload-processor/blob/main/cmd/payload-processor/runner/runner.go) for complete list.                                                                                                     |
 | `payloadProcessor.plugins`   | Custom ordered plugins array to set for the payload processor. Each plugin has fields: type, name and optionally json (which represents parameters of the plugin). If not specified, the payload processor will use by default the `body-field-to-header` to extract the `model` field, and `base-model-to-header` (in that order). |
+| `payloadProcessor.listModels`             | Optional list of model names. When set, a `models.json` file is rendered into the ConfigMap and mounted at `/config/models.json` in the pod. Used by the `model-config-datasource` plugin to seed the IPP datastore with known models. |
 | `provider.name`              | Name of the Inference Gateway implementation being used. Possible values: `istio`, `gke`. Defaults to `none`.                                                                                                                                                                                    |
 | `provider.supportedEvents.requestHeaders` | Enable Request Headers event. Defaults to `true`. |
 | `provider.supportedEvents.requestBody` | Enable Request Body event. Defaults to `true`. |

config/charts/payload-processor/templates/config.yaml

@@ -26,4 +26,11 @@ data:
     kind: PayloadProcessorConfig
     {{- .Values.payloadProcessor.customConfig | toYaml | nindent 4 }}
   {{- end }}
+  {{- if .Values.payloadProcessor.listModels }}
+  {{- $models := list }}
+  {{- range .Values.payloadProcessor.listModels }}
+  {{- $models = append $models (dict "name" .) }}
+  {{- end }}
+  models.json: {{ dict "models" $models | toJson | quote }}
+  {{- end }}
 ---