feat(server): wire real inference into ProfileAdaptationService

apps/server/src/main/kotlin/com/correx/apps/server/Main.kt

@@ -57,9 +57,11 @@ import com.correx.core.toolintent.rules.ExecInterpreterRule
 import com.correx.core.toolintent.rules.NetworkHostRule
 import com.correx.core.toolintent.rules.PathContainmentRule
 import com.correx.apps.server.freestyle.FreestyleDriver
+import com.correx.apps.server.inference.summarizeWithInference
 import com.correx.apps.server.personalization.ProfileAdaptationService
 import com.correx.core.artifacts.kind.DefaultArtifactKindRegistry
 import com.correx.core.events.types.ArtifactId
+import com.correx.core.events.types.SessionId
 import com.correx.infrastructure.InfrastructureModule
 import com.correx.infrastructure.inference.DefaultProviderRegistry
 import com.correx.infrastructure.workflow.ExecutionPlanCompiler
@@ -320,7 +322,7 @@ fun main() {
         if (correxConfig.personalization.enabled && correxConfig.personalization.learn) {
             ProfileAdaptationService(
                 decisionJournalRepository = decisionJournalRepository,
-                summarize = { _ -> "# no changes\n# (stub: wire real LLM summarizer)" },
+                summarize = { prompt -> summarizeWithInference(prompt, SessionId("system"), inferenceRouter) },
             )
         } else null
 

apps/server/src/main/kotlin/com/correx/apps/server/inference/InferenceSummarizer.kt

@@ -0,0 +1,57 @@
+package com.correx.apps.server.inference
+
+import com.correx.core.context.model.CompressionMetadata
+import com.correx.core.context.model.ContextEntry
+import com.correx.core.context.model.ContextLayer
+import com.correx.core.context.model.ContextPack
+import com.correx.core.context.model.EntryRole
+import com.correx.core.events.types.ContextEntryId
+import com.correx.core.events.types.ContextPackId
+import com.correx.core.events.types.InferenceRequestId
+import com.correx.core.events.types.SessionId
+import com.correx.core.events.types.StageId
+import com.correx.core.inference.GenerationConfig
+import com.correx.core.inference.InferenceRequest
+import com.correx.core.inference.InferenceRouter
+import com.correx.core.inference.ModelCapability
+import com.correx.core.inference.ResponseFormat
+import java.util.UUID
+
+suspend fun summarizeWithInference(
+    prompt: String,
+    sessionId: SessionId,
+    inferenceRouter: InferenceRouter,
+): String {
+    val stageId = StageId("summarizer")
+    val provider = inferenceRouter.route(stageId, setOf(ModelCapability.General))
+    val contextPack = ContextPack(
+        id = ContextPackId(UUID.randomUUID().toString()),
+        sessionId = sessionId,
+        stageId = stageId,
+        layers = mapOf(
+            ContextLayer.L0 to listOf(
+                ContextEntry(
+                    id = ContextEntryId(UUID.randomUUID().toString()),
+                    layer = ContextLayer.L0,
+                    content = prompt,
+                    sourceType = "summarizer",
+                    sourceId = "summarizer",
+                    tokenEstimate = prompt.length / 4,
+                    role = EntryRole.USER,
+                )
+            )
+        ),
+        budgetUsed = prompt.length / 4,
+        budgetLimit = 512,
+        compressionMetadata = CompressionMetadata(),
+    )
+    val request = InferenceRequest(
+        requestId = InferenceRequestId(UUID.randomUUID().toString()),
+        sessionId = sessionId,
+        stageId = stageId,
+        contextPack = contextPack,
+        generationConfig = GenerationConfig(temperature = 0.3, topP = 0.9, maxTokens = 512),
+        responseFormat = ResponseFormat.Text,
+    )
+    return provider.infer(request).text
+}