epic-12: after epic audit and init commit

infrastructure/inference/commons/build.gradle

@@ -0,0 +1,19 @@
+plugins {
+    id 'java-library'
+    id 'org.jetbrains.kotlin.jvm'
+    id 'org.jetbrains.kotlin.plugin.serialization'
+}
+
+ext {
+    ktor_version = '3.0.3'
+}
+
+dependencies {
+    implementation project(':core:inference')
+    implementation project(':core:events')
+    implementation project(':core:context')
+    implementation "io.ktor:ktor-client-core:$ktor_version"
+    implementation "io.ktor:ktor-client-cio:$ktor_version"
+    implementation "io.ktor:ktor-client-content-negotiation:$ktor_version"
+    implementation "io.ktor:ktor-serialization-kotlinx-json:$ktor_version"
+}

infrastructure/inference/commons/src/main/kotlin/com/correx/infrastructure/inference/commons/ManagedInferenceProvider.kt

@@ -0,0 +1,26 @@
+package com.correx.infrastructure.inference.commons
+
+import com.correx.core.events.types.ProviderId
+import com.correx.core.inference.InferenceProvider
+import com.correx.core.inference.Tokenizer
+
+/**
+ * A wrapper around [InferenceProvider] that tracks model lifecycle.
+ * Ensures the provider is tied to a specific [ModelManager] instance and [ModelDescriptor].
+ */
+interface ManagedInferenceProvider : InferenceProvider {
+    /**
+     * Unload the managed model.
+     */
+    suspend fun unload()
+
+    /**
+     * Check if the underlying model is still loaded.
+     */
+    fun isLoaded(): Boolean
+
+    /**
+     * Load a different model, replacing this one.
+     */
+    suspend fun load(newDescriptor: ModelDescriptor): ManagedInferenceProvider
+}

infrastructure/inference/commons/src/main/kotlin/com/correx/infrastructure/inference/commons/ModelDescriptor.kt

@@ -0,0 +1,15 @@
+package com.correx.infrastructure.inference.commons
+
+import com.correx.core.inference.CapabilityScore
+
+/**
+ * Descriptor for a model instance with its configuration.
+ */
+data class ModelDescriptor(
+    val modelId: String,
+    val modelPath: String,
+    val residencyMode: ResidencyMode,
+    val idleTimeoutMs: Long = 60_000L,
+    val contextSize: Int = 8192,
+    val capabilities: Set<CapabilityScore> = emptySet(),
+)

infrastructure/inference/commons/src/main/kotlin/com/correx/infrastructure/inference/commons/ModelLoadException.kt

@@ -0,0 +1,11 @@
+package com.correx.infrastructure.inference.commons
+
+/**
+ * Thrown when model loading fails during ModelManager.load().
+ * No event is emitted on failure per event sourcing principles.
+ */
+class ModelLoadException(
+    override val message: String,
+    val modelId: String,
+    override val cause: Throwable? = null,
+) : RuntimeException(message, cause)

infrastructure/inference/commons/src/main/kotlin/com/correx/infrastructure/inference/commons/ModelManager.kt

@@ -0,0 +1,36 @@
+package com.correx.infrastructure.inference.commons
+
+import com.correx.core.inference.InferenceProvider
+import com.correx.core.inference.ProviderHealth
+
+/**
+ * Manages the lifecycle of models.
+ */
+interface ModelManager {
+    /**
+     * Loads a model and returns an InferenceProvider instance.
+     * 
+     * @param descriptor Model configuration
+     * @return An InferenceProvider for the loaded model
+     * @throws ModelLoadException if a different model is already loaded
+     */
+    suspend fun load(descriptor: ModelDescriptor): InferenceProvider
+
+    /**
+     * Unloads the currently loaded model.
+     * 
+     * @param modelId The model ID to unload
+     * @throws ModelLoadException if no model is currently loaded
+     */
+    suspend fun unload(modelId: String)
+
+    /**
+     * Returns the descriptor of the currently loaded model, or null if none.
+     */
+    fun currentModel(): ModelDescriptor?
+
+    /**
+     * Health check for the current model.
+     */
+    fun healthCheck(): ProviderHealth
+}

infrastructure/inference/commons/src/main/kotlin/com/correx/infrastructure/inference/commons/ResidencyMode.kt

@@ -0,0 +1,10 @@
+package com.correx.infrastructure.inference.commons
+
+/**
+ * Residency modes for model lifecycle management.
+ */
+enum class ResidencyMode {
+    PERSISTENT,  // never unload
+    DYNAMIC,     // unload after idle timeout
+    EPHEMERAL,   // unload immediately after infer completes
+}