child centroid and discover auto face detection expansion

crashable v1

.idea/deviceManager.xml

@@ -1,28 +1,6 @@
 <?xml version="1.0" encoding="UTF-8"?>
 <project version="4">
   <component name="DeviceTable">
-    <option name="collapsedNodes">
-      <list>
-        <CategoryListState>
-          <option name="categories">
-            <list>
-              <CategoryState>
-                <option name="attribute" value="Type" />
-                <option name="value" value="Virtual" />
-              </CategoryState>
-              <CategoryState>
-                <option name="attribute" value="Type" />
-                <option name="value" value="Virtual" />
-              </CategoryState>
-              <CategoryState>
-                <option name="attribute" value="Type" />
-                <option name="value" value="Virtual" />
-              </CategoryState>
-            </list>
-          </option>
-        </CategoryListState>
-      </list>
-    </option>
     <option name="columnSorters">
       <list>
         <ColumnSorterState>
@@ -38,6 +16,13 @@
         <option value="Type" />
         <option value="Type" />
         <option value="Type" />
+        <option value="Type" />
+        <option value="Type" />
+        <option value="Type" />
+        <option value="Type" />
+        <option value="Type" />
+        <option value="Type" />
+        <option value="Type" />
       </list>
     </option>
   </component>

app/src/main/java/com/placeholder/sherpai2/data/local/AppDatabase.kt

@@ -2,32 +2,22 @@ package com.placeholder.sherpai2.data.local
 
 import androidx.room.Database
 import androidx.room.RoomDatabase
+import androidx.sqlite.db.SupportSQLiteDatabase
+import androidx.room.migration.Migration
 import com.placeholder.sherpai2.data.local.dao.*
 import com.placeholder.sherpai2.data.local.entity.*
 
 /**
  * AppDatabase - Complete database for SherpAI2
  *
- * VERSION 7 - Added face detection cache to ImageEntity:
- * - hasFaces: Boolean?
- * - faceCount: Int?
- * - facesLastDetected: Long?
- * - faceDetectionVersion: Int?
+ * VERSION 8 - PHASE 2: Multi-centroid face models + age tagging
+ * - Added PersonEntity.isChild, siblingIds, familyGroupId
+ * - Changed FaceModelEntity.embedding → centroidsJson (multi-centroid)
+ * - Added PersonAgeTagEntity table for searchable age tags
  *
- * ENTITIES:
- * - YOUR EXISTING: Image, Tag, Event, junction tables
- * - NEW: PersonEntity (people in your app)
- * - NEW: FaceModelEntity (face embeddings, links to PersonEntity)
- * - NEW: PhotoFaceTagEntity (face detections, links to ImageEntity + FaceModelEntity)
- *
- * DEV MODE: Using destructive migration (fallbackToDestructiveMigration)
- * - Fresh install on every schema change
- * - No manual migrations needed during development
- *
- * PRODUCTION MODE: Add proper migrations before release
- * - See DatabaseMigration.kt for migration code
- * - Remove fallbackToDestructiveMigration()
- * - Add .addMigrations(MIGRATION_6_7)
+ * MIGRATION STRATEGY:
+ * - Development: fallbackToDestructiveMigration (fresh install)
+ * - Production: Add MIGRATION_7_8 before release
  */
 @Database(
     entities = [
@@ -42,16 +32,16 @@ import com.placeholder.sherpai2.data.local.entity.*
         PersonEntity::class,
         FaceModelEntity::class,
         PhotoFaceTagEntity::class,
+        PersonAgeTagEntity::class,        // NEW: Age tagging
 
         // ===== COLLECTIONS =====
         CollectionEntity::class,
         CollectionImageEntity::class,
         CollectionFilterEntity::class
     ],
-    version = 7,  // INCREMENTED for face detection cache
+    version = 8,  // INCREMENTED for Phase 2
     exportSchema = false
 )
-// No TypeConverters needed - embeddings stored as strings
 abstract class AppDatabase : RoomDatabase() {
 
     // ===== CORE DAOs =====
@@ -66,33 +56,111 @@ abstract class AppDatabase : RoomDatabase() {
     abstract fun personDao(): PersonDao
     abstract fun faceModelDao(): FaceModelDao
     abstract fun photoFaceTagDao(): PhotoFaceTagDao
+    abstract fun personAgeTagDao(): PersonAgeTagDao  // NEW
 
     // ===== COLLECTIONS DAO =====
     abstract fun collectionDao(): CollectionDao
 }
 
 /**
- * MIGRATION NOTES FOR PRODUCTION:
+ * MIGRATION 7 → 8 (Phase 2)
  *
- * When ready to ship to users, replace destructive migration with proper migration:
+ * Changes:
+ * 1. Add isChild, siblingIds, familyGroupId to persons table
+ * 2. Rename embedding → centroidsJson in face_models table
+ * 3. Create person_age_tags table
+ */
+val MIGRATION_7_8 = object : Migration(7, 8) {
+    override fun migrate(database: SupportSQLiteDatabase) {
+
+        // ===== STEP 1: Update persons table =====
+        database.execSQL("ALTER TABLE persons ADD COLUMN isChild INTEGER NOT NULL DEFAULT 0")
+        database.execSQL("ALTER TABLE persons ADD COLUMN siblingIds TEXT DEFAULT NULL")
+        database.execSQL("ALTER TABLE persons ADD COLUMN familyGroupId TEXT DEFAULT NULL")
+
+        // Create index on familyGroupId for sibling queries
+        database.execSQL("CREATE INDEX IF NOT EXISTS index_persons_familyGroupId ON persons(familyGroupId)")
+
+        // ===== STEP 2: Update face_models table =====
+        // Rename embedding column to centroidsJson
+        // SQLite doesn't support RENAME COLUMN directly, so we need to:
+        // 1. Create new table with new schema
+        // 2. Copy data (converting single embedding to centroid JSON)
+        // 3. Drop old table
+        // 4. Rename new table
+
+        // Create new table
+        database.execSQL("""
+            CREATE TABLE IF NOT EXISTS face_models_new (
+                id TEXT PRIMARY KEY NOT NULL,
+                personId TEXT NOT NULL,
+                centroidsJson TEXT NOT NULL,
+                trainingImageCount INTEGER NOT NULL,
+                averageConfidence REAL NOT NULL,
+                createdAt INTEGER NOT NULL,
+                updatedAt INTEGER NOT NULL,
+                lastUsed INTEGER,
+                isActive INTEGER NOT NULL,
+                FOREIGN KEY(personId) REFERENCES persons(id) ON DELETE CASCADE
+            )
+        """)
+
+        // Copy data, converting embedding to centroidsJson format
+        // This converts single embedding to a list with one centroid
+        database.execSQL("""
+            INSERT INTO face_models_new 
+            SELECT 
+                id,
+                personId,
+                '[{"embedding":' || REPLACE(REPLACE(embedding, ',', ','), ',', ',') || ',"effectiveTimestamp":' || createdAt || ',"ageAtCapture":null,"photoCount":' || trainingImageCount || ',"timeRangeMonths":12,"avgConfidence":' || averageConfidence || '}]' as centroidsJson,
+                trainingImageCount,
+                averageConfidence,
+                createdAt,
+                updatedAt,
+                lastUsed,
+                isActive
+            FROM face_models
+        """)
+
+        // Drop old table
+        database.execSQL("DROP TABLE face_models")
+
+        // Rename new table
+        database.execSQL("ALTER TABLE face_models_new RENAME TO face_models")
+
+        // Recreate index
+        database.execSQL("CREATE UNIQUE INDEX IF NOT EXISTS index_face_models_personId ON face_models(personId)")
+
+        // ===== STEP 3: Create person_age_tags table =====
+        database.execSQL("""
+            CREATE TABLE IF NOT EXISTS person_age_tags (
+                id TEXT PRIMARY KEY NOT NULL,
+                personId TEXT NOT NULL,
+                imageId TEXT NOT NULL,
+                ageAtCapture INTEGER NOT NULL,
+                tagValue TEXT NOT NULL,
+                confidence REAL NOT NULL,
+                createdAt INTEGER NOT NULL,
+                FOREIGN KEY(personId) REFERENCES persons(id) ON DELETE CASCADE,
+                FOREIGN KEY(imageId) REFERENCES images(imageId) ON DELETE CASCADE
+            )
+        """)
+
+        // Create indices for fast lookups
+        database.execSQL("CREATE INDEX IF NOT EXISTS index_person_age_tags_personId ON person_age_tags(personId)")
+        database.execSQL("CREATE INDEX IF NOT EXISTS index_person_age_tags_imageId ON person_age_tags(imageId)")
+        database.execSQL("CREATE INDEX IF NOT EXISTS index_person_age_tags_ageAtCapture ON person_age_tags(ageAtCapture)")
+        database.execSQL("CREATE INDEX IF NOT EXISTS index_person_age_tags_tagValue ON person_age_tags(tagValue)")
+    }
+}
+
+/**
+ * PRODUCTION MIGRATION NOTES:
  *
- * val MIGRATION_6_7 = object : Migration(6, 7) {
- *     override fun migrate(database: SupportSQLiteDatabase) {
- *         // Add face detection cache columns
- *         database.execSQL("ALTER TABLE images ADD COLUMN hasFaces INTEGER DEFAULT NULL")
- *         database.execSQL("ALTER TABLE images ADD COLUMN faceCount INTEGER DEFAULT NULL")
- *         database.execSQL("ALTER TABLE images ADD COLUMN facesLastDetected INTEGER DEFAULT NULL")
- *         database.execSQL("ALTER TABLE images ADD COLUMN faceDetectionVersion INTEGER DEFAULT NULL")
+ * Before shipping to users, update DatabaseModule to use migration:
  *
- *         // Create indices
- *         database.execSQL("CREATE INDEX IF NOT EXISTS index_images_hasFaces ON images(hasFaces)")
- *         database.execSQL("CREATE INDEX IF NOT EXISTS index_images_faceCount ON images(faceCount)")
- *     }
- * }
- *
- * Then in your database builder:
- * Room.databaseBuilder(context, AppDatabase::class.java, "database_name")
- *     .addMigrations(MIGRATION_6_7)  // Add this
+ * Room.databaseBuilder(context, AppDatabase::class.java, "sherpai.db")
+ *     .addMigrations(MIGRATION_7_8)  // Add this
  *     // .fallbackToDestructiveMigration()  // Remove this
  *     .build()
  */

app/src/main/java/com/placeholder/sherpai2/data/local/dao/Personagetagdao.kt

@@ -0,0 +1,104 @@
+package com.placeholder.sherpai2.data.local.dao
+
+import androidx.room.*
+import com.placeholder.sherpai2.data.local.entity.PersonAgeTagEntity
+import kotlinx.coroutines.flow.Flow
+
+/**
+ * PersonAgeTagDao - Manage searchable age tags for children
+ *
+ * USAGE EXAMPLES:
+ * - Search "emma age 3" → getImageIdsForTag("emma_age3")
+ * - Find all photos of Emma at age 5 → getImageIdsForPersonAtAge(emmaId, 5)
+ * - Get age progression → getTagsForPerson(emmaId) sorted by age
+ */
+@Dao
+interface PersonAgeTagDao {
+
+    @Insert(onConflict = OnConflictStrategy.REPLACE)
+    suspend fun insertTag(tag: PersonAgeTagEntity)
+
+    @Insert(onConflict = OnConflictStrategy.REPLACE)
+    suspend fun insertTags(tags: List<PersonAgeTagEntity>)
+
+    /**
+     * Get all age tags for a person (sorted by age)
+     * Useful for age progression timeline
+     */
+    @Query("SELECT * FROM person_age_tags WHERE personId = :personId ORDER BY ageAtCapture ASC")
+    suspend fun getTagsForPerson(personId: String): List<PersonAgeTagEntity>
+
+    /**
+     * Get all age tags for an image
+     */
+    @Query("SELECT * FROM person_age_tags WHERE imageId = :imageId")
+    suspend fun getTagsForImage(imageId: String): List<PersonAgeTagEntity>
+
+    /**
+     * Search by tag value (e.g., "emma_age3")
+     * Returns all image IDs matching this tag
+     */
+    @Query("SELECT DISTINCT imageId FROM person_age_tags WHERE tagValue = :tagValue")
+    suspend fun getImageIdsForTag(tagValue: String): List<String>
+
+    /**
+     * Get images of a person at a specific age
+     */
+    @Query("SELECT DISTINCT imageId FROM person_age_tags WHERE personId = :personId AND ageAtCapture = :age")
+    suspend fun getImageIdsForPersonAtAge(personId: String, age: Int): List<String>
+
+    /**
+     * Get images of a person in an age range
+     */
+    @Query("""
+        SELECT DISTINCT imageId FROM person_age_tags 
+        WHERE personId = :personId 
+        AND ageAtCapture BETWEEN :minAge AND :maxAge
+        ORDER BY ageAtCapture ASC
+    """)
+    suspend fun getImageIdsForPersonAgeRange(personId: String, minAge: Int, maxAge: Int): List<String>
+
+    /**
+     * Get all unique ages for a person (for age picker UI)
+     */
+    @Query("SELECT DISTINCT ageAtCapture FROM person_age_tags WHERE personId = :personId ORDER BY ageAtCapture ASC")
+    suspend fun getAgesForPerson(personId: String): List<Int>
+
+    /**
+     * Delete all tags for a person
+     */
+    @Query("DELETE FROM person_age_tags WHERE personId = :personId")
+    suspend fun deleteTagsForPerson(personId: String)
+
+    /**
+     * Delete all tags for an image
+     */
+    @Query("DELETE FROM person_age_tags WHERE imageId = :imageId")
+    suspend fun deleteTagsForImage(imageId: String)
+
+    /**
+     * Get count of photos at each age (for statistics)
+     */
+    @Query("""
+        SELECT ageAtCapture, COUNT(DISTINCT imageId) as count 
+        FROM person_age_tags 
+        WHERE personId = :personId 
+        GROUP BY ageAtCapture 
+        ORDER BY ageAtCapture ASC
+    """)
+    suspend fun getPhotoCountByAge(personId: String): List<AgePhotoCount>
+
+    /**
+     * Flow version for reactive UI
+     */
+    @Query("SELECT * FROM person_age_tags WHERE personId = :personId ORDER BY ageAtCapture ASC")
+    fun getTagsForPersonFlow(personId: String): Flow<List<PersonAgeTagEntity>>
+}
+
+/**
+ * Data class for age photo count statistics
+ */
+data class AgePhotoCount(
+    val ageAtCapture: Int,
+    val count: Int
+)

app/src/main/java/com/placeholder/sherpai2/data/local/entity/Facerecognitionentities.kt

@@ -5,19 +5,24 @@ import androidx.room.Entity
 import androidx.room.ForeignKey
 import androidx.room.Index
 import androidx.room.PrimaryKey
+import org.json.JSONArray
+import org.json.JSONObject
 import java.util.UUID
 
 /**
- * PersonEntity - NO DEFAULT VALUES for KSP compatibility
+ * PersonEntity - ENHANCED with child tracking and sibling relationships
  */
 @Entity(
     tableName = "persons",
-    indices = [Index(value = ["name"])]
+    indices = [
+        Index(value = ["name"]),
+        Index(value = ["familyGroupId"])
+    ]
 )
 data class PersonEntity(
     @PrimaryKey
     @ColumnInfo(name = "id")
-    val id: String,                    // ← No default
+    val id: String,
 
     @ColumnInfo(name = "name")
     val name: String,
@@ -25,26 +30,48 @@ data class PersonEntity(
     @ColumnInfo(name = "dateOfBirth")
     val dateOfBirth: Long?,
 
+    @ColumnInfo(name = "isChild")
+    val isChild: Boolean,              // NEW: Auto-set based on age
+
+    @ColumnInfo(name = "siblingIds")
+    val siblingIds: String?,           // NEW: JSON list ["uuid1", "uuid2"]
+
+    @ColumnInfo(name = "familyGroupId")
+    val familyGroupId: String?,        // NEW: UUID for family unit
+
     @ColumnInfo(name = "relationship")
     val relationship: String?,
 
     @ColumnInfo(name = "createdAt")
-    val createdAt: Long,               // ← No default
+    val createdAt: Long,
 
     @ColumnInfo(name = "updatedAt")
-    val updatedAt: Long                // ← No default
+    val updatedAt: Long
 ) {
     companion object {
         fun create(
             name: String,
             dateOfBirth: Long? = null,
+            isChild: Boolean = false,
+            siblingIds: List<String> = emptyList(),
             relationship: String? = null
         ): PersonEntity {
             val now = System.currentTimeMillis()
+
+            // Create family group if siblings exist
+            val familyGroupId = if (siblingIds.isNotEmpty()) {
+                UUID.randomUUID().toString()
+            } else null
+
             return PersonEntity(
                 id = UUID.randomUUID().toString(),
                 name = name,
                 dateOfBirth = dateOfBirth,
+                isChild = isChild,
+                siblingIds = if (siblingIds.isNotEmpty()) {
+                    JSONArray(siblingIds).toString()
+                } else null,
+                familyGroupId = familyGroupId,
                 relationship = relationship,
                 createdAt = now,
                 updatedAt = now
@@ -52,6 +79,17 @@ data class PersonEntity(
         }
     }
 
+    fun getSiblingIds(): List<String> {
+        return if (siblingIds != null) {
+            try {
+                val jsonArray = JSONArray(siblingIds)
+                (0 until jsonArray.length()).map { jsonArray.getString(it) }
+            } catch (e: Exception) {
+                emptyList()
+            }
+        } else emptyList()
+    }
+
     fun getAge(): Int? {
         if (dateOfBirth == null) return null
         val now = System.currentTimeMillis()
@@ -74,7 +112,7 @@ data class PersonEntity(
 }
 
 /**
- * FaceModelEntity - NO DEFAULT VALUES
+ * FaceModelEntity - MULTI-CENTROID support for temporal tracking
  */
 @Entity(
     tableName = "face_models",
@@ -91,13 +129,13 @@ data class PersonEntity(
 data class FaceModelEntity(
     @PrimaryKey
     @ColumnInfo(name = "id")
-    val id: String,                    // ← No default
+    val id: String,
 
     @ColumnInfo(name = "personId")
     val personId: String,
 
-    @ColumnInfo(name = "embedding")
-    val embedding: String,
+    @ColumnInfo(name = "centroidsJson")
+    val centroidsJson: String,         // NEW: List<TemporalCentroid> as JSON
 
     @ColumnInfo(name = "trainingImageCount")
     val trainingImageCount: Int,
@@ -106,10 +144,10 @@ data class FaceModelEntity(
     val averageConfidence: Float,
 
     @ColumnInfo(name = "createdAt")
-    val createdAt: Long,               // ← No default
+    val createdAt: Long,
 
     @ColumnInfo(name = "updatedAt")
-    val updatedAt: Long,               // ← No default
+    val updatedAt: Long,
 
     @ColumnInfo(name = "lastUsed")
     val lastUsed: Long?,
@@ -118,17 +156,42 @@ data class FaceModelEntity(
     val isActive: Boolean
 ) {
     companion object {
+        /**
+         * Backwards compatible create() method
+         * Used by existing FaceRecognitionRepository code
+         */
         fun create(
             personId: String,
             embeddingArray: FloatArray,
             trainingImageCount: Int,
             averageConfidence: Float
+        ): FaceModelEntity {
+            return createFromEmbedding(personId, embeddingArray, trainingImageCount, averageConfidence)
+        }
+
+        /**
+         * Create from single embedding (backwards compatible)
+         */
+        fun createFromEmbedding(
+            personId: String,
+            embeddingArray: FloatArray,
+            trainingImageCount: Int,
+            averageConfidence: Float
         ): FaceModelEntity {
             val now = System.currentTimeMillis()
+            val centroid = TemporalCentroid(
+                embedding = embeddingArray.toList(),
+                effectiveTimestamp = now,
+                ageAtCapture = null,
+                photoCount = trainingImageCount,
+                timeRangeMonths = 12,
+                avgConfidence = averageConfidence
+            )
+
             return FaceModelEntity(
                 id = UUID.randomUUID().toString(),
                 personId = personId,
-                embedding = embeddingArray.joinToString(","),
+                centroidsJson = serializeCentroids(listOf(centroid)),
                 trainingImageCount = trainingImageCount,
                 averageConfidence = averageConfidence,
                 createdAt = now,
@@ -137,15 +200,106 @@ data class FaceModelEntity(
                 isActive = true
             )
         }
+
+        /**
+         * Create from multiple centroids (temporal tracking)
+         */
+        fun createFromCentroids(
+            personId: String,
+            centroids: List<TemporalCentroid>,
+            trainingImageCount: Int,
+            averageConfidence: Float
+        ): FaceModelEntity {
+            val now = System.currentTimeMillis()
+            return FaceModelEntity(
+                id = UUID.randomUUID().toString(),
+                personId = personId,
+                centroidsJson = serializeCentroids(centroids),
+                trainingImageCount = trainingImageCount,
+                averageConfidence = averageConfidence,
+                createdAt = now,
+                updatedAt = now,
+                lastUsed = null,
+                isActive = true
+            )
         }
 
+        /**
+         * Serialize list of centroids to JSON
+         */
+        private fun serializeCentroids(centroids: List<TemporalCentroid>): String {
+            val jsonArray = JSONArray()
+            centroids.forEach { centroid ->
+                val jsonObj = JSONObject()
+                jsonObj.put("embedding", JSONArray(centroid.embedding))
+                jsonObj.put("effectiveTimestamp", centroid.effectiveTimestamp)
+                jsonObj.put("ageAtCapture", centroid.ageAtCapture)
+                jsonObj.put("photoCount", centroid.photoCount)
+                jsonObj.put("timeRangeMonths", centroid.timeRangeMonths)
+                jsonObj.put("avgConfidence", centroid.avgConfidence)
+                jsonArray.put(jsonObj)
+            }
+            return jsonArray.toString()
+        }
+
+        /**
+         * Deserialize JSON to list of centroids
+         */
+        private fun deserializeCentroids(json: String): List<TemporalCentroid> {
+            val jsonArray = JSONArray(json)
+            return (0 until jsonArray.length()).map { i ->
+                val jsonObj = jsonArray.getJSONObject(i)
+                val embeddingArray = jsonObj.getJSONArray("embedding")
+                val embedding = (0 until embeddingArray.length()).map { j ->
+                    embeddingArray.getDouble(j).toFloat()
+                }
+                TemporalCentroid(
+                    embedding = embedding,
+                    effectiveTimestamp = jsonObj.getLong("effectiveTimestamp"),
+                    ageAtCapture = if (jsonObj.isNull("ageAtCapture")) null else jsonObj.getDouble("ageAtCapture").toFloat(),
+                    photoCount = jsonObj.getInt("photoCount"),
+                    timeRangeMonths = jsonObj.getInt("timeRangeMonths"),
+                    avgConfidence = jsonObj.getDouble("avgConfidence").toFloat()
+                )
+            }
+        }
+    }
+
+    fun getCentroids(): List<TemporalCentroid> {
+        return try {
+            FaceModelEntity.deserializeCentroids(centroidsJson)
+        } catch (e: Exception) {
+            emptyList()
+        }
+    }
+
+    // Backwards compatibility: get first centroid as single embedding
     fun getEmbeddingArray(): FloatArray {
-        return embedding.split(",").map { it.toFloat() }.toFloatArray()
+        val centroids = getCentroids()
+        return if (centroids.isNotEmpty()) {
+            centroids.first().getEmbeddingArray()
+        } else {
+            FloatArray(192) // Empty embedding
+        }
     }
 }
 
 /**
- * PhotoFaceTagEntity - NO DEFAULT VALUES
+ * TemporalCentroid - Represents a face appearance at a specific time period
+ */
+data class TemporalCentroid(
+    val embedding: List<Float>,        // 192D vector
+    val effectiveTimestamp: Long,      // Center of time window
+    val ageAtCapture: Float?,          // Age in years (for children)
+    val photoCount: Int,               // Number of photos in this cluster
+    val timeRangeMonths: Int,          // Width of time window (e.g., 6 months)
+    val avgConfidence: Float           // Quality indicator
+) {
+    fun getEmbeddingArray(): FloatArray = embedding.toFloatArray()
+}
+
+/**
+ * PhotoFaceTagEntity - Unchanged
  */
 @Entity(
     tableName = "photo_face_tags",
@@ -172,7 +326,7 @@ data class FaceModelEntity(
 data class PhotoFaceTagEntity(
     @PrimaryKey
     @ColumnInfo(name = "id")
-    val id: String,                    // ← No default
+    val id: String,
 
     @ColumnInfo(name = "imageId")
     val imageId: String,
@@ -190,7 +344,7 @@ data class PhotoFaceTagEntity(
     val embedding: String,
 
     @ColumnInfo(name = "detectedAt")
-    val detectedAt: Long,              // ← No default
+    val detectedAt: Long,
 
     @ColumnInfo(name = "verifiedByUser")
     val verifiedByUser: Boolean,
@@ -229,3 +383,73 @@ data class PhotoFaceTagEntity(
         return embedding.split(",").map { it.toFloat() }.toFloatArray()
     }
 }
+
+/**
+ * PersonAgeTagEntity - NEW: Searchable age tags
+ */
+@Entity(
+    tableName = "person_age_tags",
+    foreignKeys = [
+        ForeignKey(
+            entity = PersonEntity::class,
+            parentColumns = ["id"],
+            childColumns = ["personId"],
+            onDelete = ForeignKey.CASCADE
+        ),
+        ForeignKey(
+            entity = ImageEntity::class,
+            parentColumns = ["imageId"],
+            childColumns = ["imageId"],
+            onDelete = ForeignKey.CASCADE
+        )
+    ],
+    indices = [
+        Index(value = ["personId"]),
+        Index(value = ["imageId"]),
+        Index(value = ["ageAtCapture"]),
+        Index(value = ["tagValue"])
+    ]
+)
+data class PersonAgeTagEntity(
+    @PrimaryKey
+    @ColumnInfo(name = "id")
+    val id: String,
+
+    @ColumnInfo(name = "personId")
+    val personId: String,
+
+    @ColumnInfo(name = "imageId")
+    val imageId: String,
+
+    @ColumnInfo(name = "ageAtCapture")
+    val ageAtCapture: Int,
+
+    @ColumnInfo(name = "tagValue")
+    val tagValue: String,             // e.g., "emma_age3"
+
+    @ColumnInfo(name = "confidence")
+    val confidence: Float,
+
+    @ColumnInfo(name = "createdAt")
+    val createdAt: Long
+) {
+    companion object {
+        fun create(
+            personId: String,
+            personName: String,
+            imageId: String,
+            ageAtCapture: Int,
+            confidence: Float
+        ): PersonAgeTagEntity {
+            return PersonAgeTagEntity(
+                id = UUID.randomUUID().toString(),
+                personId = personId,
+                imageId = imageId,
+                ageAtCapture = ageAtCapture,
+                tagValue = "${personName.lowercase().replace(" ", "_")}_age$ageAtCapture",
+                confidence = confidence,
+                createdAt = System.currentTimeMillis()
+            )
+        }
+    }
+}

app/src/main/java/com/placeholder/sherpai2/di/DatabaseModule.kt

@@ -3,6 +3,7 @@ package com.placeholder.sherpai2.di
 import android.content.Context
 import androidx.room.Room
 import com.placeholder.sherpai2.data.local.AppDatabase
+import com.placeholder.sherpai2.data.local.MIGRATION_7_8
 import com.placeholder.sherpai2.data.local.dao.*
 import dagger.Module
 import dagger.Provides
@@ -14,9 +15,9 @@ import javax.inject.Singleton
 /**
  * DatabaseModule - Provides database and ALL DAOs
  *
- * DEVELOPMENT CONFIGURATION:
- * - fallbackToDestructiveMigration enabled
- * - No migrations required
+ * PHASE 2 UPDATES:
+ * - Added PersonAgeTagDao
+ * - Added migration v7→v8 (commented out for development)
  */
 @Module
 @InstallIn(SingletonComponent::class)
@@ -34,7 +35,12 @@ object DatabaseModule {
             AppDatabase::class.java,
             "sherpai.db"
         )
+            // DEVELOPMENT MODE: Destructive migration (fresh install on schema change)
             .fallbackToDestructiveMigration()
+
+            // PRODUCTION MODE: Uncomment this and remove fallbackToDestructiveMigration()
+            // .addMigrations(MIGRATION_7_8)
+
             .build()
 
     // ===== CORE DAOs =====
@@ -77,7 +83,12 @@ object DatabaseModule {
     fun providePhotoFaceTagDao(db: AppDatabase): PhotoFaceTagDao =
         db.photoFaceTagDao()
 
+    @Provides
+    fun providePersonAgeTagDao(db: AppDatabase): PersonAgeTagDao =  // NEW
+        db.personAgeTagDao()
+
     // ===== COLLECTIONS DAOs =====
+
     @Provides
     fun provideCollectionDao(db: AppDatabase): CollectionDao =
         db.collectionDao()

app/src/main/java/com/placeholder/sherpai2/domain/clustering/Faceclusteringservice.kt

@@ -0,0 +1,465 @@
+package com.placeholder.sherpai2.domain.clustering
+
+import android.content.Context
+import android.graphics.Bitmap
+import android.graphics.BitmapFactory
+import android.net.Uri
+import com.google.mlkit.vision.common.InputImage
+import com.google.mlkit.vision.face.FaceDetection
+import com.google.mlkit.vision.face.FaceDetectorOptions
+import com.placeholder.sherpai2.data.local.dao.ImageDao
+import com.placeholder.sherpai2.data.local.entity.ImageEntity
+import com.placeholder.sherpai2.ml.FaceNetModel
+import dagger.hilt.android.qualifiers.ApplicationContext
+import kotlinx.coroutines.Dispatchers
+import kotlinx.coroutines.async
+import kotlinx.coroutines.awaitAll
+import kotlinx.coroutines.coroutineScope
+import kotlinx.coroutines.sync.Semaphore
+import kotlinx.coroutines.withContext
+import java.util.concurrent.atomic.AtomicInteger
+import javax.inject.Inject
+import javax.inject.Singleton
+import kotlin.math.sqrt
+
+/**
+ * FaceClusteringService - Auto-discover people in photo library
+ *
+ * STRATEGY:
+ * 1. Load all images with faces (from cache)
+ * 2. Detect faces and generate embeddings (parallel)
+ * 3. DBSCAN clustering on embeddings
+ * 4. Co-occurrence analysis (faces in same photo)
+ * 5. Return high-quality clusters (10-100 people typical)
+ *
+ * PERFORMANCE:
+ * - Uses face detection cache (only ~30% of photos)
+ * - Parallel processing (12 concurrent)
+ * - Smart sampling (don't need ALL faces for clustering)
+ * - Result: ~2-5 minutes for 10,000 photo library
+ */
+@Singleton
+class FaceClusteringService @Inject constructor(
+    @ApplicationContext private val context: Context,
+    private val imageDao: ImageDao
+) {
+
+    private val semaphore = Semaphore(12)
+
+    /**
+     * Main clustering entry point
+     *
+     * @param maxFacesToCluster Limit for performance (default 2000)
+     * @param onProgress Progress callback (current, total, message)
+     */
+    suspend fun discoverPeople(
+        maxFacesToCluster: Int = 2000,
+        onProgress: (Int, Int, String) -> Unit = { _, _, _ -> }
+    ): ClusteringResult = withContext(Dispatchers.Default) {
+
+        onProgress(0, 100, "Loading images with faces...")
+
+        // Step 1: Get images with faces (cached, fast!)
+        val imagesWithFaces = imageDao.getImagesWithFaces()
+
+        if (imagesWithFaces.isEmpty()) {
+            // Check if face cache is populated at all
+            val totalImages = withContext(Dispatchers.IO) {
+                imageDao.getImageCount()
+            }
+
+            if (totalImages == 0) {
+                return@withContext ClusteringResult(
+                    clusters = emptyList(),
+                    totalFacesAnalyzed = 0,
+                    processingTimeMs = 0,
+                    errorMessage = "No photos in library. Please wait for photo ingestion to complete."
+                )
+            }
+
+            // Images exist but no face cache - need to run PopulateFaceDetectionCacheUseCase first
+            return@withContext ClusteringResult(
+                clusters = emptyList(),
+                totalFacesAnalyzed = 0,
+                processingTimeMs = 0,
+                errorMessage = "Face detection cache not ready. Please wait for initial face scan to complete (check MainActivity progress bar)."
+            )
+        }
+
+        onProgress(10, 100, "Analyzing ${imagesWithFaces.size} photos...")
+
+        val startTime = System.currentTimeMillis()
+
+        // Step 2: Detect faces and generate embeddings (parallel)
+        val allFaces = detectFacesInImages(
+            images = imagesWithFaces.take(1000), // Smart limit: don't need all photos
+            onProgress = { current, total ->
+                onProgress(10 + (current * 40 / total), 100, "Detecting faces... $current/$total")
+            }
+        )
+
+        if (allFaces.isEmpty()) {
+            return@withContext ClusteringResult(
+                clusters = emptyList(),
+                totalFacesAnalyzed = 0,
+                processingTimeMs = System.currentTimeMillis() - startTime
+            )
+        }
+
+        onProgress(50, 100, "Clustering ${allFaces.size} faces...")
+
+        // Step 3: DBSCAN clustering on embeddings
+        val rawClusters = performDBSCAN(
+            faces = allFaces.take(maxFacesToCluster),
+            epsilon = 0.30f,  // BALANCED: Not too strict, not too loose
+            minPoints = 5      // Minimum 5 photos to form a cluster
+        )
+
+        onProgress(70, 100, "Analyzing relationships...")
+
+        // Step 4: Build co-occurrence graph
+        val coOccurrenceGraph = buildCoOccurrenceGraph(rawClusters)
+
+        onProgress(80, 100, "Selecting representative faces...")
+
+        // Step 5: Select representative faces for each cluster
+        val clusters = rawClusters.map { cluster ->
+            FaceCluster(
+                clusterId = cluster.clusterId,
+                faces = cluster.faces,
+                representativeFaces = selectRepresentativeFaces(cluster.faces, count = 6),
+                photoCount = cluster.faces.map { it.imageId }.distinct().size,
+                averageConfidence = cluster.faces.map { it.confidence }.average().toFloat(),
+                estimatedAge = estimateAge(cluster.faces),
+                potentialSiblings = findPotentialSiblings(cluster, rawClusters, coOccurrenceGraph)
+            )
+        }.sortedByDescending { it.photoCount } // Most frequent first
+
+        onProgress(100, 100, "Found ${clusters.size} people!")
+
+        ClusteringResult(
+            clusters = clusters,
+            totalFacesAnalyzed = allFaces.size,
+            processingTimeMs = System.currentTimeMillis() - startTime
+        )
+    }
+
+    /**
+     * Detect faces in images and generate embeddings (parallel)
+     */
+    private suspend fun detectFacesInImages(
+        images: List<ImageEntity>,
+        onProgress: (Int, Int) -> Unit
+    ): List<DetectedFaceWithEmbedding> = coroutineScope {
+
+        val detector = com.google.mlkit.vision.face.FaceDetection.getClient(
+            com.google.mlkit.vision.face.FaceDetectorOptions.Builder()
+                .setPerformanceMode(com.google.mlkit.vision.face.FaceDetectorOptions.PERFORMANCE_MODE_ACCURATE)
+                .setMinFaceSize(0.15f)
+                .build()
+        )
+
+        val faceNetModel = FaceNetModel(context)
+        val allFaces = mutableListOf<DetectedFaceWithEmbedding>()
+        val processedCount = java.util.concurrent.atomic.AtomicInteger(0)
+
+        try {
+            val jobs = images.map { image ->
+                async {
+                    semaphore.acquire()
+                    try {
+                        val faces = detectFacesInImage(image, detector, faceNetModel)
+                        val current = processedCount.incrementAndGet()
+                        if (current % 10 == 0) {
+                            onProgress(current, images.size)
+                        }
+                        faces
+                    } finally {
+                        semaphore.release()
+                    }
+                }
+            }
+
+            jobs.awaitAll().flatten().also {
+                allFaces.addAll(it)
+            }
+
+        } finally {
+            detector.close()
+            faceNetModel.close()
+        }
+
+        allFaces
+    }
+
+    private suspend fun detectFacesInImage(
+        image: ImageEntity,
+        detector: com.google.mlkit.vision.face.FaceDetector,
+        faceNetModel: FaceNetModel
+    ): List<DetectedFaceWithEmbedding> = withContext(Dispatchers.IO) {
+
+        try {
+            val uri = Uri.parse(image.imageUri)
+            val bitmap = loadBitmapDownsampled(uri, 512) ?: return@withContext emptyList()
+
+            val mlImage = com.google.mlkit.vision.common.InputImage.fromBitmap(bitmap, 0)
+            val faces = com.google.android.gms.tasks.Tasks.await(detector.process(mlImage))
+
+            val result = faces.mapNotNull { face ->
+                try {
+                    val faceBitmap = Bitmap.createBitmap(
+                        bitmap,
+                        face.boundingBox.left.coerceIn(0, bitmap.width - 1),
+                        face.boundingBox.top.coerceIn(0, bitmap.height - 1),
+                        face.boundingBox.width().coerceAtMost(bitmap.width - face.boundingBox.left),
+                        face.boundingBox.height().coerceAtMost(bitmap.height - face.boundingBox.top)
+                    )
+
+                    val embedding = faceNetModel.generateEmbedding(faceBitmap)
+                    faceBitmap.recycle()
+
+                    DetectedFaceWithEmbedding(
+                        imageId = image.imageId,
+                        imageUri = image.imageUri,
+                        capturedAt = image.capturedAt,
+                        embedding = embedding,
+                        boundingBox = face.boundingBox,
+                        confidence = 1.0f // Placeholder
+                    )
+                } catch (e: Exception) {
+                    null
+                }
+            }
+
+            bitmap.recycle()
+            result
+
+        } catch (e: Exception) {
+            emptyList()
+        }
+    }
+
+    /**
+     * DBSCAN clustering algorithm
+     */
+    private fun performDBSCAN(
+        faces: List<DetectedFaceWithEmbedding>,
+        epsilon: Float,
+        minPoints: Int
+    ): List<RawCluster> {
+
+        val visited = mutableSetOf<Int>()
+        val clusters = mutableListOf<RawCluster>()
+        var clusterId = 0
+
+        for (i in faces.indices) {
+            if (i in visited) continue
+
+            val neighbors = findNeighbors(i, faces, epsilon)
+
+            if (neighbors.size < minPoints) {
+                visited.add(i)
+                continue // Noise point
+            }
+
+            // Start new cluster
+            val cluster = mutableListOf<DetectedFaceWithEmbedding>()
+            val queue = ArrayDeque(neighbors)
+            visited.add(i)
+            cluster.add(faces[i])
+
+            while (queue.isNotEmpty()) {
+                val pointIdx = queue.removeFirst()
+                if (pointIdx in visited) continue
+
+                visited.add(pointIdx)
+                cluster.add(faces[pointIdx])
+
+                val pointNeighbors = findNeighbors(pointIdx, faces, epsilon)
+                if (pointNeighbors.size >= minPoints) {
+                    queue.addAll(pointNeighbors.filter { it !in visited })
+                }
+            }
+
+            if (cluster.size >= minPoints) {
+                clusters.add(RawCluster(clusterId++, cluster))
+            }
+        }
+
+        return clusters
+    }
+
+    private fun findNeighbors(
+        pointIdx: Int,
+        faces: List<DetectedFaceWithEmbedding>,
+        epsilon: Float
+    ): List<Int> {
+        val point = faces[pointIdx]
+        return faces.indices.filter { i ->
+            i != pointIdx && cosineSimilarity(point.embedding, faces[i].embedding) > (1 - epsilon)
+        }
+    }
+
+    private fun cosineSimilarity(a: FloatArray, b: FloatArray): Float {
+        var dotProduct = 0f
+        var normA = 0f
+        var normB = 0f
+
+        for (i in a.indices) {
+            dotProduct += a[i] * b[i]
+            normA += a[i] * a[i]
+            normB += b[i] * b[i]
+        }
+
+        return dotProduct / (sqrt(normA) * sqrt(normB))
+    }
+
+    /**
+     * Build co-occurrence graph (faces appearing in same photos)
+     */
+    private fun buildCoOccurrenceGraph(clusters: List<RawCluster>): Map<Int, Map<Int, Int>> {
+        val graph = mutableMapOf<Int, MutableMap<Int, Int>>()
+
+        for (i in clusters.indices) {
+            graph[i] = mutableMapOf()
+            val imageIds = clusters[i].faces.map { it.imageId }.toSet()
+
+            for (j in clusters.indices) {
+                if (i == j) continue
+
+                val sharedImages = clusters[j].faces.count { it.imageId in imageIds }
+                if (sharedImages > 0) {
+                    graph[i]!![j] = sharedImages
+                }
+            }
+        }
+
+        return graph
+    }
+
+    private fun findPotentialSiblings(
+        cluster: RawCluster,
+        allClusters: List<RawCluster>,
+        coOccurrenceGraph: Map<Int, Map<Int, Int>>
+    ): List<Int> {
+        val clusterIdx = allClusters.indexOf(cluster)
+        if (clusterIdx == -1) return emptyList()
+
+        val siblings = coOccurrenceGraph[clusterIdx]
+            ?.filter { (_, count) -> count >= 5 } // At least 5 shared photos
+            ?.keys
+            ?.toList()
+            ?: emptyList()
+
+        return siblings
+    }
+
+    /**
+     * Select diverse representative faces for UI display
+     */
+    private fun selectRepresentativeFaces(
+        faces: List<DetectedFaceWithEmbedding>,
+        count: Int
+    ): List<DetectedFaceWithEmbedding> {
+        if (faces.size <= count) return faces
+
+        // Time-based sampling: spread across different dates
+        val sortedByTime = faces.sortedBy { it.capturedAt }
+        val step = faces.size / count
+
+        return (0 until count).map { i ->
+            sortedByTime[i * step]
+        }
+    }
+
+    /**
+     * Estimate if cluster represents a child (based on photo timestamps)
+     */
+    private fun estimateAge(faces: List<DetectedFaceWithEmbedding>): AgeEstimate {
+        val timestamps = faces.map { it.capturedAt }.sorted()
+        val span = timestamps.last() - timestamps.first()
+        val spanYears = span / (365.25 * 24 * 60 * 60 * 1000)
+
+        // If face appearance changes over 3+ years, likely a child
+        return if (spanYears > 3.0) {
+            AgeEstimate.CHILD
+        } else {
+            AgeEstimate.UNKNOWN
+        }
+    }
+
+    private fun loadBitmapDownsampled(uri: Uri, maxDim: Int): Bitmap? {
+        return try {
+            val opts = BitmapFactory.Options().apply { inJustDecodeBounds = true }
+            context.contentResolver.openInputStream(uri)?.use {
+                BitmapFactory.decodeStream(it, null, opts)
+            }
+
+            var sample = 1
+            while (opts.outWidth / sample > maxDim || opts.outHeight / sample > maxDim) {
+                sample *= 2
+            }
+
+            val finalOpts = BitmapFactory.Options().apply {
+                inSampleSize = sample
+                inPreferredConfig = Bitmap.Config.RGB_565
+            }
+
+            context.contentResolver.openInputStream(uri)?.use {
+                BitmapFactory.decodeStream(it, null, finalOpts)
+            }
+        } catch (e: Exception) {
+            null
+        }
+    }
+}
+
+// ==================
+// DATA CLASSES
+// ==================
+
+data class DetectedFaceWithEmbedding(
+    val imageId: String,
+    val imageUri: String,
+    val capturedAt: Long,
+    val embedding: FloatArray,
+    val boundingBox: android.graphics.Rect,
+    val confidence: Float
+) {
+    override fun equals(other: Any?): Boolean {
+        if (this === other) return true
+        if (javaClass != other?.javaClass) return false
+        other as DetectedFaceWithEmbedding
+        return imageId == other.imageId
+    }
+
+    override fun hashCode(): Int = imageId.hashCode()
+}
+
+data class RawCluster(
+    val clusterId: Int,
+    val faces: List<DetectedFaceWithEmbedding>
+)
+
+data class FaceCluster(
+    val clusterId: Int,
+    val faces: List<DetectedFaceWithEmbedding>,
+    val representativeFaces: List<DetectedFaceWithEmbedding>,
+    val photoCount: Int,
+    val averageConfidence: Float,
+    val estimatedAge: AgeEstimate,
+    val potentialSiblings: List<Int>
+)
+
+data class ClusteringResult(
+    val clusters: List<FaceCluster>,
+    val totalFacesAnalyzed: Int,
+    val processingTimeMs: Long,
+    val errorMessage: String? = null
+)
+
+enum class AgeEstimate {
+    CHILD,     // Appearance changes significantly over time
+    ADULT,     // Stable appearance
+    UNKNOWN    // Not enough data
+}

app/src/main/java/com/placeholder/sherpai2/domain/training/Clustertrainingservice.kt

@@ -0,0 +1,234 @@
+package com.placeholder.sherpai2.domain.training
+
+import android.content.Context
+import android.graphics.BitmapFactory
+import android.net.Uri
+import com.placeholder.sherpai2.data.local.dao.FaceModelDao
+import com.placeholder.sherpai2.data.local.dao.PersonDao
+import com.placeholder.sherpai2.data.local.entity.FaceModelEntity
+import com.placeholder.sherpai2.data.local.entity.PersonEntity
+import com.placeholder.sherpai2.data.local.entity.TemporalCentroid
+import com.placeholder.sherpai2.domain.clustering.FaceCluster
+import com.placeholder.sherpai2.ml.FaceNetModel
+import dagger.hilt.android.qualifiers.ApplicationContext
+import kotlinx.coroutines.Dispatchers
+import kotlinx.coroutines.withContext
+import javax.inject.Inject
+import javax.inject.Singleton
+import kotlin.math.abs
+
+/**
+ * ClusterTrainingService - Train multi-centroid face models from clusters
+ *
+ * STRATEGY:
+ * 1. For children: Create multiple temporal centroids (one per age period)
+ * 2. For adults: Create single centroid (stable appearance)
+ * 3. Use K-Means clustering on timestamps to find age groups
+ * 4. Calculate centroid for each time period
+ */
+@Singleton
+class ClusterTrainingService @Inject constructor(
+    @ApplicationContext private val context: Context,
+    private val personDao: PersonDao,
+    private val faceModelDao: FaceModelDao
+) {
+
+    private val faceNetModel by lazy { FaceNetModel(context) }
+
+    /**
+     * Train a person from an auto-discovered cluster
+     *
+     * @return PersonId on success
+     */
+    suspend fun trainFromCluster(
+        cluster: FaceCluster,
+        name: String,
+        dateOfBirth: Long?,
+        isChild: Boolean,
+        siblingClusterIds: List<Int>,
+        onProgress: (Int, Int, String) -> Unit = { _, _, _ -> }
+    ): String = withContext(Dispatchers.Default) {
+
+        onProgress(0, 100, "Creating person...")
+
+        // Step 1: Create PersonEntity
+        val person = PersonEntity.create(
+            name = name,
+            dateOfBirth = dateOfBirth,
+            isChild = isChild,
+            siblingIds = emptyList(), // Will update after siblings are created
+            relationship = if (isChild) "Child" else null
+        )
+
+        withContext(Dispatchers.IO) {
+            personDao.insert(person)
+        }
+
+        onProgress(20, 100, "Analyzing face variations...")
+
+        // Step 2: Generate embeddings for all faces in cluster
+        val facesWithEmbeddings = cluster.faces.mapNotNull { face ->
+            try {
+                val bitmap = context.contentResolver.openInputStream(Uri.parse(face.imageUri))?.use {
+                    BitmapFactory.decodeStream(it)
+                } ?: return@mapNotNull null
+
+                // Generate embedding
+                val embedding = faceNetModel.generateEmbedding(bitmap)
+                bitmap.recycle()
+
+                Triple(face.imageUri, face.capturedAt, embedding)
+            } catch (e: Exception) {
+                null
+            }
+        }
+
+        if (facesWithEmbeddings.isEmpty()) {
+            throw Exception("Failed to process any faces from cluster")
+        }
+
+        onProgress(50, 100, "Creating face model...")
+
+        // Step 3: Create centroids based on whether person is a child
+        val centroids = if (isChild && dateOfBirth != null) {
+            createTemporalCentroidsForChild(
+                facesWithEmbeddings = facesWithEmbeddings,
+                dateOfBirth = dateOfBirth
+            )
+        } else {
+            createSingleCentroid(facesWithEmbeddings)
+        }
+
+        onProgress(80, 100, "Saving model...")
+
+        // Step 4: Calculate average confidence
+        val avgConfidence = centroids.map { it.avgConfidence }.average().toFloat()
+
+        // Step 5: Create FaceModelEntity
+        val faceModel = FaceModelEntity.createFromCentroids(
+            personId = person.id,
+            centroids = centroids,
+            trainingImageCount = cluster.faces.size,
+            averageConfidence = avgConfidence
+        )
+
+        withContext(Dispatchers.IO) {
+            faceModelDao.insertFaceModel(faceModel)
+        }
+
+        onProgress(100, 100, "Complete!")
+
+        person.id
+    }
+
+    /**
+     * Create temporal centroids for a child
+     * Groups faces by age and creates one centroid per age period
+     */
+    private fun createTemporalCentroidsForChild(
+        facesWithEmbeddings: List<Triple<String, Long, FloatArray>>,
+        dateOfBirth: Long
+    ): List<TemporalCentroid> {
+
+        // Group faces by age (in years)
+        val facesByAge = facesWithEmbeddings.groupBy { (_, capturedAt, _) ->
+            val ageMs = capturedAt - dateOfBirth
+            val ageYears = (ageMs / (365.25 * 24 * 60 * 60 * 1000)).toInt()
+            ageYears.coerceIn(0, 18) // Cap at 18 years
+        }
+
+        // Create one centroid per age group
+        return facesByAge.map { (age, faces) ->
+            val embeddings = faces.map { it.third }
+            val avgEmbedding = averageEmbeddings(embeddings)
+            val avgTimestamp = faces.map { it.second }.average().toLong()
+
+            // Calculate confidence (how similar faces are to each other)
+            val confidences = embeddings.map { emb ->
+                cosineSimilarity(avgEmbedding, emb)
+            }
+            val avgConfidence = confidences.average().toFloat()
+
+            TemporalCentroid(
+                embedding = avgEmbedding.toList(),
+                effectiveTimestamp = avgTimestamp,
+                ageAtCapture = age.toFloat(),
+                photoCount = faces.size,
+                timeRangeMonths = 12, // 1 year window
+                avgConfidence = avgConfidence
+            )
+        }.sortedBy { it.ageAtCapture }
+    }
+
+    /**
+     * Create single centroid for an adult (stable appearance)
+     */
+    private fun createSingleCentroid(
+        facesWithEmbeddings: List<Triple<String, Long, FloatArray>>
+    ): List<TemporalCentroid> {
+
+        val embeddings = facesWithEmbeddings.map { it.third }
+        val avgEmbedding = averageEmbeddings(embeddings)
+        val avgTimestamp = facesWithEmbeddings.map { it.second }.average().toLong()
+
+        val confidences = embeddings.map { emb ->
+            cosineSimilarity(avgEmbedding, emb)
+        }
+        val avgConfidence = confidences.average().toFloat()
+
+        return listOf(
+            TemporalCentroid(
+                embedding = avgEmbedding.toList(),
+                effectiveTimestamp = avgTimestamp,
+                ageAtCapture = null,
+                photoCount = facesWithEmbeddings.size,
+                timeRangeMonths = 24, // 2 year window for adults
+                avgConfidence = avgConfidence
+            )
+        )
+    }
+
+    /**
+     * Average multiple embeddings into one
+     */
+    private fun averageEmbeddings(embeddings: List<FloatArray>): FloatArray {
+        val size = embeddings.first().size
+        val avg = FloatArray(size) { 0f }
+
+        embeddings.forEach { embedding ->
+            for (i in embedding.indices) {
+                avg[i] += embedding[i]
+            }
+        }
+
+        val count = embeddings.size.toFloat()
+        for (i in avg.indices) {
+            avg[i] /= count
+        }
+
+        // Normalize to unit length
+        val norm = kotlin.math.sqrt(avg.map { it * it }.sum())
+        return avg.map { it / norm }.toFloatArray()
+    }
+
+    /**
+     * Calculate cosine similarity between two embeddings
+     */
+    private fun cosineSimilarity(a: FloatArray, b: FloatArray): Float {
+        var dotProduct = 0f
+        var normA = 0f
+        var normB = 0f
+
+        for (i in a.indices) {
+            dotProduct += a[i] * b[i]
+            normA += a[i] * a[i]
+            normB += b[i] * b[i]
+        }
+
+        return dotProduct / (kotlin.math.sqrt(normA) * kotlin.math.sqrt(normB))
+    }
+
+    fun cleanup() {
+        faceNetModel.close()
+    }
+}

app/src/main/java/com/placeholder/sherpai2/domain/usecase/Populatefacedetectioncacheusecase.kt

@@ -176,7 +176,7 @@ class PopulateFaceDetectionCacheUseCase @Inject constructor(
                     faceCount = update.faceCount
                 )
             } catch (e: Exception) {
-                // Skip failed updates
+                // Skip failed updates //todo
             }
         }
     }

app/src/main/java/com/placeholder/sherpai2/ui/discover/Discoverpeoplescreen.kt

@@ -0,0 +1,687 @@
+package com.placeholder.sherpai2.ui.discover
+
+import android.graphics.BitmapFactory
+import android.net.Uri
+import androidx.compose.foundation.Image
+import androidx.compose.foundation.background
+import androidx.compose.foundation.clickable
+import androidx.compose.foundation.layout.*
+import androidx.compose.foundation.lazy.LazyColumn
+import androidx.compose.foundation.lazy.grid.GridCells
+import androidx.compose.foundation.lazy.grid.LazyVerticalGrid
+import androidx.compose.foundation.lazy.grid.items
+import androidx.compose.foundation.lazy.items
+import androidx.compose.foundation.shape.CircleShape
+import androidx.compose.foundation.shape.RoundedCornerShape
+import androidx.compose.material.icons.Icons
+import androidx.compose.material.icons.filled.*
+import androidx.compose.material3.*
+import androidx.compose.runtime.*
+import androidx.compose.ui.Alignment
+import androidx.compose.ui.Modifier
+import androidx.compose.ui.draw.clip
+import androidx.compose.ui.graphics.asImageBitmap
+import androidx.compose.ui.layout.ContentScale
+import androidx.compose.ui.platform.LocalContext
+import androidx.compose.ui.text.font.FontWeight
+import androidx.compose.ui.text.style.TextAlign
+import androidx.compose.ui.unit.dp
+import androidx.hilt.navigation.compose.hiltViewModel
+import androidx.lifecycle.compose.collectAsStateWithLifecycle
+import com.placeholder.sherpai2.domain.clustering.AgeEstimate
+import com.placeholder.sherpai2.domain.clustering.FaceCluster
+import java.text.SimpleDateFormat
+import java.util.*
+
+/**
+ * DiscoverPeopleScreen - Beautiful auto-clustering UI
+ *
+ * FLOW:
+ * 1. Hero CTA: "Discover People in Your Photos"
+ * 2. Auto-clustering progress (2-5 min)
+ * 3. Grid of discovered people
+ * 4. Tap cluster → Name person + metadata
+ * 5. Background deep scan starts
+ */
+@OptIn(ExperimentalMaterial3Api::class)
+@Composable
+fun DiscoverPeopleScreen(
+    viewModel: DiscoverPeopleViewModel = hiltViewModel()
+) {
+    val uiState by viewModel.uiState.collectAsStateWithLifecycle()
+
+    // NO SCAFFOLD - MainScreen already has TopAppBar
+    Box(modifier = Modifier.fillMaxSize()) {
+        when (val state = uiState) {
+            is DiscoverUiState.Idle -> IdleScreen(
+                onStartDiscovery = { viewModel.startDiscovery() }
+            )
+
+            is DiscoverUiState.Clustering -> ClusteringProgressScreen(
+                progress = state.progress,
+                total = state.total,
+                message = state.message
+            )
+
+            is DiscoverUiState.NamingReady -> ClusterGridScreen(
+                result = state.result,
+                onClusterClick = { cluster ->
+                    viewModel.selectCluster(cluster)
+                }
+            )
+
+            is DiscoverUiState.NamingCluster -> NamingDialog(
+                cluster = state.selectedCluster,
+                suggestedSiblings = state.suggestedSiblings,
+                onConfirm = { name, dob, isChild, siblings ->
+                    viewModel.confirmClusterName(
+                        cluster = state.selectedCluster,
+                        name = name,
+                        dateOfBirth = dob,
+                        isChild = isChild,
+                        selectedSiblings = siblings
+                    )
+                },
+                onDismiss = { viewModel.cancelNaming() }
+            )
+
+            is DiscoverUiState.NoPeopleFound -> EmptyStateScreen(
+                message = state.message
+            )
+
+            is DiscoverUiState.Error -> ErrorScreen(
+                message = state.message,
+                onRetry = { viewModel.startDiscovery() }
+            )
+        }
+    }
+}
+
+/**
+ * Idle state - Hero CTA to start discovery
+ */
+@Composable
+fun IdleScreen(
+    onStartDiscovery: () -> Unit
+) {
+    Column(
+        modifier = Modifier
+            .fillMaxSize()
+            .padding(32.dp),
+        horizontalAlignment = Alignment.CenterHorizontally,
+        verticalArrangement = Arrangement.Center
+    ) {
+        Icon(
+            imageVector = Icons.Default.AutoAwesome,
+            contentDescription = null,
+            modifier = Modifier.size(120.dp),
+            tint = MaterialTheme.colorScheme.primary
+        )
+
+        Spacer(Modifier.height(24.dp))
+
+        Text(
+            text = "Discover People",
+            style = MaterialTheme.typography.headlineLarge,
+            fontWeight = FontWeight.Bold,
+            textAlign = TextAlign.Center
+        )
+
+        Spacer(Modifier.height(16.dp))
+
+        Text(
+            text = "Let AI automatically find and group faces in your photos. " +
+                    "You'll name them, and we'll tag all their photos.",
+            style = MaterialTheme.typography.bodyLarge,
+            textAlign = TextAlign.Center,
+            color = MaterialTheme.colorScheme.onSurfaceVariant
+        )
+
+        Spacer(Modifier.height(32.dp))
+
+        Button(
+            onClick = onStartDiscovery,
+            modifier = Modifier
+                .fillMaxWidth()
+                .height(56.dp),
+            colors = ButtonDefaults.buttonColors(
+                containerColor = MaterialTheme.colorScheme.primary
+            )
+        ) {
+            Icon(
+                imageVector = Icons.Default.AutoAwesome,
+                contentDescription = null,
+                modifier = Modifier.size(24.dp)
+            )
+            Spacer(Modifier.width(8.dp))
+            Text(
+                text = "Start Discovery",
+                style = MaterialTheme.typography.titleMedium,
+                fontWeight = FontWeight.Bold
+            )
+        }
+
+        Spacer(Modifier.height(16.dp))
+
+        Card(
+            modifier = Modifier.fillMaxWidth(),
+            colors = CardDefaults.cardColors(
+                containerColor = MaterialTheme.colorScheme.surfaceVariant
+            )
+        ) {
+            Column(
+                modifier = Modifier.padding(16.dp),
+                verticalArrangement = Arrangement.spacedBy(8.dp)
+            ) {
+                InfoRow(Icons.Default.Speed, "Fast: Analyzes ~1000 photos in 2-5 minutes")
+                InfoRow(Icons.Default.Security, "Private: Everything stays on your device")
+                InfoRow(Icons.Default.AutoAwesome, "Smart: Groups faces automatically")
+            }
+        }
+    }
+}
+
+@Composable
+fun InfoRow(icon: androidx.compose.ui.graphics.vector.ImageVector, text: String) {
+    Row(
+        verticalAlignment = Alignment.CenterVertically,
+        horizontalArrangement = Arrangement.spacedBy(12.dp)
+    ) {
+        Icon(
+            imageVector = icon,
+            contentDescription = null,
+            tint = MaterialTheme.colorScheme.primary,
+            modifier = Modifier.size(20.dp)
+        )
+        Text(
+            text = text,
+            style = MaterialTheme.typography.bodyMedium
+        )
+    }
+}
+
+/**
+ * Clustering progress screen
+ */
+@Composable
+fun ClusteringProgressScreen(
+    progress: Int,
+    total: Int,
+    message: String
+) {
+    Column(
+        modifier = Modifier
+            .fillMaxSize()
+            .padding(32.dp),
+        horizontalAlignment = Alignment.CenterHorizontally,
+        verticalArrangement = Arrangement.Center
+    ) {
+        CircularProgressIndicator(
+            modifier = Modifier.size(80.dp),
+            strokeWidth = 6.dp
+        )
+
+        Spacer(Modifier.height(32.dp))
+
+        Text(
+            text = "Discovering People...",
+            style = MaterialTheme.typography.headlineSmall,
+            fontWeight = FontWeight.Bold
+        )
+
+        Spacer(Modifier.height(16.dp))
+
+        LinearProgressIndicator(
+            progress = { if (total > 0) progress.toFloat() / total.toFloat() else 0f },
+            modifier = Modifier.fillMaxWidth(),
+        )
+
+        Spacer(Modifier.height(8.dp))
+
+        Text(
+            text = message,
+            style = MaterialTheme.typography.bodyMedium,
+            color = MaterialTheme.colorScheme.onSurfaceVariant
+        )
+
+        Spacer(Modifier.height(24.dp))
+
+        Text(
+            text = "This will take 2-5 minutes. You can leave and come back later.",
+            style = MaterialTheme.typography.bodySmall,
+            textAlign = TextAlign.Center,
+            color = MaterialTheme.colorScheme.onSurfaceVariant
+        )
+    }
+}
+
+/**
+ * Grid of discovered clusters
+ */
+@Composable
+fun ClusterGridScreen(
+    result: com.placeholder.sherpai2.domain.clustering.ClusteringResult,
+    onClusterClick: (FaceCluster) -> Unit
+) {
+    Column(
+        modifier = Modifier
+            .fillMaxSize()
+            .padding(16.dp)
+    ) {
+        Text(
+            text = "Found ${result.clusters.size} People",
+            style = MaterialTheme.typography.headlineSmall,
+            fontWeight = FontWeight.Bold
+        )
+
+        Spacer(Modifier.height(8.dp))
+
+        Text(
+            text = "Tap to name each person. We'll then tag all their photos.",
+            style = MaterialTheme.typography.bodyMedium,
+            color = MaterialTheme.colorScheme.onSurfaceVariant
+        )
+
+        Spacer(Modifier.height(16.dp))
+
+        LazyVerticalGrid(
+            columns = GridCells.Fixed(2),
+            horizontalArrangement = Arrangement.spacedBy(12.dp),
+            verticalArrangement = Arrangement.spacedBy(12.dp)
+        ) {
+            items(result.clusters) { cluster ->
+                ClusterCard(
+                    cluster = cluster,
+                    onClick = { onClusterClick(cluster) }
+                )
+            }
+        }
+    }
+}
+
+/**
+ * Single cluster card
+ */
+@Composable
+fun ClusterCard(
+    cluster: FaceCluster,
+    onClick: () -> Unit
+) {
+    val context = LocalContext.current
+
+    Card(
+        modifier = Modifier
+            .fillMaxWidth()
+            .clickable(onClick = onClick),
+        elevation = CardDefaults.cardElevation(defaultElevation = 2.dp)
+    ) {
+        Column {
+            // Face grid (2x3)
+            LazyVerticalGrid(
+                columns = GridCells.Fixed(3),
+                modifier = Modifier.height(180.dp),
+                userScrollEnabled = false
+            ) {
+                items(cluster.representativeFaces.take(6)) { face ->
+                    val bitmap = remember(face.imageUri) {
+                        try {
+                            context.contentResolver.openInputStream(Uri.parse(face.imageUri))?.use {
+                                BitmapFactory.decodeStream(it)
+                            }
+                        } catch (e: Exception) {
+                            null
+                        }
+                    }
+
+                    if (bitmap != null) {
+                        Image(
+                            bitmap = bitmap.asImageBitmap(),
+                            contentDescription = null,
+                            modifier = Modifier
+                                .fillMaxWidth()
+                                .aspectRatio(1f),
+                            contentScale = ContentScale.Crop
+                        )
+                    } else {
+                        Box(
+                            modifier = Modifier
+                                .fillMaxWidth()
+                                .aspectRatio(1f)
+                                .background(MaterialTheme.colorScheme.surfaceVariant),
+                            contentAlignment = Alignment.Center
+                        ) {
+                            Icon(
+                                imageVector = Icons.Default.Person,
+                                contentDescription = null,
+                                tint = MaterialTheme.colorScheme.onSurfaceVariant
+                            )
+                        }
+                    }
+                }
+            }
+
+            // Info
+            Column(
+                modifier = Modifier.padding(12.dp)
+            ) {
+                Row(
+                    modifier = Modifier.fillMaxWidth(),
+                    horizontalArrangement = Arrangement.SpaceBetween,
+                    verticalAlignment = Alignment.CenterVertically
+                ) {
+                    Text(
+                        text = "${cluster.photoCount} photos",
+                        style = MaterialTheme.typography.titleMedium,
+                        fontWeight = FontWeight.Bold
+                    )
+
+                    if (cluster.estimatedAge == AgeEstimate.CHILD) {
+                        Surface(
+                            shape = RoundedCornerShape(12.dp),
+                            color = MaterialTheme.colorScheme.primaryContainer
+                        ) {
+                            Text(
+                                text = "Child",
+                                modifier = Modifier.padding(horizontal = 8.dp, vertical = 4.dp),
+                                style = MaterialTheme.typography.labelSmall,
+                                color = MaterialTheme.colorScheme.onPrimaryContainer
+                            )
+                        }
+                    }
+                }
+
+                if (cluster.potentialSiblings.isNotEmpty()) {
+                    Spacer(Modifier.height(4.dp))
+                    Text(
+                        text = "Appears with ${cluster.potentialSiblings.size} other person(s)",
+                        style = MaterialTheme.typography.bodySmall,
+                        color = MaterialTheme.colorScheme.onSurfaceVariant
+                    )
+                }
+            }
+        }
+    }
+}
+
+/**
+ * Naming dialog
+ */
+@Composable
+fun NamingDialog(
+    cluster: FaceCluster,
+    suggestedSiblings: List<FaceCluster>,
+    onConfirm: (String, Long?, Boolean, List<Int>) -> Unit,
+    onDismiss: () -> Unit
+) {
+    var name by remember { mutableStateOf("") }
+    var isChild by remember { mutableStateOf(cluster.estimatedAge == AgeEstimate.CHILD) }
+    var dateOfBirth by remember { mutableStateOf<Long?>(null) }
+    var selectedSiblings by remember { mutableStateOf<Set<Int>>(emptySet()) }
+    var showDatePicker by remember { mutableStateOf(false) }
+    val context = LocalContext.current
+
+    // Date picker dialog
+    if (showDatePicker) {
+        val calendar = java.util.Calendar.getInstance()
+        if (dateOfBirth != null) {
+            calendar.timeInMillis = dateOfBirth!!
+        }
+
+        val datePickerDialog = android.app.DatePickerDialog(
+            context,
+            { _, year, month, dayOfMonth ->
+                val cal = java.util.Calendar.getInstance()
+                cal.set(year, month, dayOfMonth)
+                dateOfBirth = cal.timeInMillis
+                showDatePicker = false
+            },
+            calendar.get(java.util.Calendar.YEAR),
+            calendar.get(java.util.Calendar.MONTH),
+            calendar.get(java.util.Calendar.DAY_OF_MONTH)
+        )
+
+        datePickerDialog.setOnDismissListener {
+            showDatePicker = false
+        }
+
+        DisposableEffect(Unit) {
+            datePickerDialog.show()
+            onDispose {
+                datePickerDialog.dismiss()
+            }
+        }
+    }
+
+    AlertDialog(
+        onDismissRequest = onDismiss,
+        title = {
+            Text("Name This Person")
+        },
+        text = {
+            Column(
+                verticalArrangement = Arrangement.spacedBy(16.dp)
+            ) {
+                // FACE PREVIEW - Show 6 representative faces
+                Text(
+                    text = "${cluster.photoCount} photos found",
+                    style = MaterialTheme.typography.bodyMedium,
+                    color = MaterialTheme.colorScheme.onSurfaceVariant
+                )
+
+                LazyVerticalGrid(
+                    columns = GridCells.Fixed(3),
+                    modifier = Modifier.height(180.dp),
+                    horizontalArrangement = Arrangement.spacedBy(4.dp),
+                    verticalArrangement = Arrangement.spacedBy(4.dp)
+                ) {
+                    items(cluster.representativeFaces.take(6)) { face ->
+                        val bitmap = remember(face.imageUri) {
+                            try {
+                                context.contentResolver.openInputStream(Uri.parse(face.imageUri))?.use {
+                                    BitmapFactory.decodeStream(it)
+                                }
+                            } catch (e: Exception) {
+                                null
+                            }
+                        }
+
+                        if (bitmap != null) {
+                            Image(
+                                bitmap = bitmap.asImageBitmap(),
+                                contentDescription = null,
+                                modifier = Modifier
+                                    .fillMaxWidth()
+                                    .aspectRatio(1f)
+                                    .clip(RoundedCornerShape(8.dp)),
+                                contentScale = ContentScale.Crop
+                            )
+                        } else {
+                            Box(
+                                modifier = Modifier
+                                    .fillMaxWidth()
+                                    .aspectRatio(1f)
+                                    .clip(RoundedCornerShape(8.dp))
+                                    .background(MaterialTheme.colorScheme.surfaceVariant),
+                                contentAlignment = Alignment.Center
+                            ) {
+                                Icon(
+                                    imageVector = Icons.Default.Person,
+                                    contentDescription = null,
+                                    tint = MaterialTheme.colorScheme.onSurfaceVariant
+                                )
+                            }
+                        }
+                    }
+                }
+
+                HorizontalDivider()
+
+                // Name input
+                OutlinedTextField(
+                    value = name,
+                    onValueChange = { name = it },
+                    label = { Text("Name") },
+                    singleLine = true,
+                    modifier = Modifier.fillMaxWidth()
+                )
+
+                // Is child toggle
+                Row(
+                    modifier = Modifier.fillMaxWidth(),
+                    horizontalArrangement = Arrangement.SpaceBetween,
+                    verticalAlignment = Alignment.CenterVertically
+                ) {
+                    Text("This person is a child")
+                    Switch(
+                        checked = isChild,
+                        onCheckedChange = { isChild = it }
+                    )
+                }
+
+                // Date of birth (if child)
+                if (isChild) {
+                    OutlinedButton(
+                        onClick = { showDatePicker = true },
+                        modifier = Modifier.fillMaxWidth()
+                    ) {
+                        Icon(Icons.Default.CalendarToday, null)
+                        Spacer(Modifier.width(8.dp))
+                        Text(
+                            if (dateOfBirth != null) {
+                                SimpleDateFormat("MMM dd, yyyy", Locale.getDefault())
+                                    .format(Date(dateOfBirth!!))
+                            } else {
+                                "Set Date of Birth"
+                            }
+                        )
+                    }
+                }
+
+                // Suggested siblings
+                if (suggestedSiblings.isNotEmpty()) {
+                    Text(
+                        "Appears with these people (select siblings):",
+                        style = MaterialTheme.typography.labelMedium
+                    )
+
+                    suggestedSiblings.forEach { sibling ->
+                        Row(
+                            modifier = Modifier.fillMaxWidth(),
+                            verticalAlignment = Alignment.CenterVertically
+                        ) {
+                            Checkbox(
+                                checked = sibling.clusterId in selectedSiblings,
+                                onCheckedChange = { checked ->
+                                    selectedSiblings = if (checked) {
+                                        selectedSiblings + sibling.clusterId
+                                    } else {
+                                        selectedSiblings - sibling.clusterId
+                                    }
+                                }
+                            )
+                            Text("Person ${sibling.clusterId + 1} (${sibling.photoCount} photos)")
+                        }
+                    }
+                }
+            }
+        },
+        confirmButton = {
+            TextButton(
+                onClick = {
+                    onConfirm(
+                        name,
+                        dateOfBirth,
+                        isChild,
+                        selectedSiblings.toList()
+                    )
+                },
+                enabled = name.isNotBlank()
+            ) {
+                Text("Save & Train")
+            }
+        },
+        dismissButton = {
+            TextButton(onClick = onDismiss) {
+                Text("Cancel")
+            }
+        }
+    )
+
+    // TODO: Add DatePickerDialog when showDatePicker is true
+}
+
+/**
+ * Empty state screen
+ */
+@Composable
+fun EmptyStateScreen(message: String) {
+    Column(
+        modifier = Modifier
+            .fillMaxSize()
+            .padding(32.dp),
+        horizontalAlignment = Alignment.CenterHorizontally,
+        verticalArrangement = Arrangement.Center
+    ) {
+        Icon(
+            imageVector = Icons.Default.PersonOff,
+            contentDescription = null,
+            modifier = Modifier.size(80.dp),
+            tint = MaterialTheme.colorScheme.onSurfaceVariant
+        )
+
+        Spacer(Modifier.height(16.dp))
+
+        Text(
+            text = message,
+            style = MaterialTheme.typography.bodyLarge,
+            textAlign = TextAlign.Center
+        )
+    }
+}
+
+/**
+ * Error screen
+ */
+@Composable
+fun ErrorScreen(
+    message: String,
+    onRetry: () -> Unit
+) {
+    Column(
+        modifier = Modifier
+            .fillMaxSize()
+            .padding(32.dp),
+        horizontalAlignment = Alignment.CenterHorizontally,
+        verticalArrangement = Arrangement.Center
+    ) {
+        Icon(
+            imageVector = Icons.Default.Error,
+            contentDescription = null,
+            modifier = Modifier.size(80.dp),
+            tint = MaterialTheme.colorScheme.error
+        )
+
+        Spacer(Modifier.height(16.dp))
+
+        Text(
+            text = "Oops!",
+            style = MaterialTheme.typography.headlineMedium,
+            fontWeight = FontWeight.Bold
+        )
+
+        Spacer(Modifier.height(8.dp))
+
+        Text(
+            text = message,
+            style = MaterialTheme.typography.bodyLarge,
+            textAlign = TextAlign.Center,
+            color = MaterialTheme.colorScheme.onSurfaceVariant
+        )
+
+        Spacer(Modifier.height(24.dp))
+
+        Button(onClick = onRetry) {
+            Text("Try Again")
+        }
+    }
+}

app/src/main/java/com/placeholder/sherpai2/ui/discover/Discoverpeopleviewmodel.kt

@@ -0,0 +1,222 @@
+package com.placeholder.sherpai2.ui.discover
+
+import androidx.lifecycle.ViewModel
+import androidx.lifecycle.viewModelScope
+import com.placeholder.sherpai2.domain.clustering.ClusteringResult
+import com.placeholder.sherpai2.domain.clustering.FaceCluster
+import com.placeholder.sherpai2.domain.clustering.FaceClusteringService
+import com.placeholder.sherpai2.domain.training.ClusterTrainingService
+import dagger.hilt.android.lifecycle.HiltViewModel
+import kotlinx.coroutines.flow.MutableStateFlow
+import kotlinx.coroutines.flow.StateFlow
+import kotlinx.coroutines.flow.asStateFlow
+import kotlinx.coroutines.launch
+import javax.inject.Inject
+
+/**
+ * DiscoverPeopleViewModel - Manages auto-clustering and naming flow
+ *
+ * PHASE 2: Now includes multi-centroid training from clusters
+ *
+ * STATE FLOW:
+ * 1. Idle → User taps "Discover People"
+ * 2. Clustering → Auto-analyzing faces (2-5 min)
+ * 3. NamingReady → Shows clusters, user names them
+ * 4. Training → Creating multi-centroid face model
+ * 5. Complete → Ready to scan library
+ */
+@HiltViewModel
+class DiscoverPeopleViewModel @Inject constructor(
+    private val clusteringService: FaceClusteringService,
+    private val trainingService: ClusterTrainingService
+) : ViewModel() {
+
+    private val _uiState = MutableStateFlow<DiscoverUiState>(DiscoverUiState.Idle)
+    val uiState: StateFlow<DiscoverUiState> = _uiState.asStateFlow()
+
+    // Track which clusters have been named
+    private val namedClusterIds = mutableSetOf<Int>()
+
+    /**
+     * Start auto-clustering process
+     */
+    fun startDiscovery() {
+        viewModelScope.launch {
+            try {
+                // Clear named clusters for new discovery
+                namedClusterIds.clear()
+
+                _uiState.value = DiscoverUiState.Clustering(0, 100, "Starting...")
+
+                val result = clusteringService.discoverPeople(
+                    onProgress = { current, total, message ->
+                        _uiState.value = DiscoverUiState.Clustering(current, total, message)
+                    }
+                )
+
+                // Check for errors
+                if (result.errorMessage != null) {
+                    _uiState.value = DiscoverUiState.Error(result.errorMessage)
+                    return@launch
+                }
+
+                if (result.clusters.isEmpty()) {
+                    _uiState.value = DiscoverUiState.NoPeopleFound(
+                        "No faces found in your library. Make sure face detection cache is populated."
+                    )
+                } else {
+                    _uiState.value = DiscoverUiState.NamingReady(result)
+                }
+
+            } catch (e: Exception) {
+                _uiState.value = DiscoverUiState.Error(
+                    e.message ?: "Failed to discover people"
+                )
+            }
+        }
+    }
+
+    /**
+     * User selected a cluster to name
+     */
+    fun selectCluster(cluster: FaceCluster) {
+        val currentState = _uiState.value
+        if (currentState is DiscoverUiState.NamingReady) {
+            _uiState.value = DiscoverUiState.NamingCluster(
+                result = currentState.result,
+                selectedCluster = cluster,
+                suggestedSiblings = currentState.result.clusters.filter {
+                    it.clusterId in cluster.potentialSiblings
+                }
+            )
+        }
+    }
+
+    /**
+     * User confirmed name and metadata for a cluster
+     *
+     * CREATES:
+     * 1. PersonEntity with all metadata (name, DOB, siblings)
+     * 2. Multi-centroid FaceModelEntity (temporal tracking for children)
+     * 3. Removes cluster from display
+     */
+    fun confirmClusterName(
+        cluster: FaceCluster,
+        name: String,
+        dateOfBirth: Long?,
+        isChild: Boolean,
+        selectedSiblings: List<Int>
+    ) {
+        viewModelScope.launch {
+            try {
+                val currentState = _uiState.value
+                if (currentState !is DiscoverUiState.NamingCluster) return@launch
+
+                // Train person from cluster
+                val personId = trainingService.trainFromCluster(
+                    cluster = cluster,
+                    name = name,
+                    dateOfBirth = dateOfBirth,
+                    isChild = isChild,
+                    siblingClusterIds = selectedSiblings,
+                    onProgress = { current, total, message ->
+                        _uiState.value = DiscoverUiState.Clustering(current, total, message)
+                    }
+                )
+
+                // Mark cluster as named
+                namedClusterIds.add(cluster.clusterId)
+
+                // Filter out named clusters
+                val remainingClusters = currentState.result.clusters
+                    .filter { it.clusterId !in namedClusterIds }
+
+                if (remainingClusters.isEmpty()) {
+                    // All clusters named! Show success
+                    _uiState.value = DiscoverUiState.NoPeopleFound(
+                        "All people have been named! 🎉\n\nGo to 'People' to see your trained models."
+                    )
+                } else {
+                    // Return to naming screen with remaining clusters
+                    _uiState.value = DiscoverUiState.NamingReady(
+                        result = currentState.result.copy(clusters = remainingClusters)
+                    )
+                }
+
+            } catch (e: Exception) {
+                _uiState.value = DiscoverUiState.Error(
+                    e.message ?: "Failed to create person: ${e.message}"
+                )
+            }
+        }
+    }
+
+    /**
+     * Cancel naming and go back to cluster list
+     */
+    fun cancelNaming() {
+        val currentState = _uiState.value
+        if (currentState is DiscoverUiState.NamingCluster) {
+            _uiState.value = DiscoverUiState.NamingReady(
+                result = currentState.result
+            )
+        }
+    }
+
+    /**
+     * Reset to idle state
+     */
+    fun reset() {
+        _uiState.value = DiscoverUiState.Idle
+    }
+}
+
+/**
+ * UI States for Discover People flow
+ */
+sealed class DiscoverUiState {
+
+    /**
+     * Initial state - user hasn't started discovery
+     */
+    object Idle : DiscoverUiState()
+
+    /**
+     * Auto-clustering in progress
+     */
+    data class Clustering(
+        val progress: Int,
+        val total: Int,
+        val message: String
+    ) : DiscoverUiState()
+
+    /**
+     * Clustering complete, ready for user to name people
+     */
+    data class NamingReady(
+        val result: ClusteringResult
+    ) : DiscoverUiState()
+
+    /**
+     * User is naming a specific cluster
+     */
+    data class NamingCluster(
+        val result: ClusteringResult,
+        val selectedCluster: FaceCluster,
+        val suggestedSiblings: List<FaceCluster>
+    ) : DiscoverUiState()
+
+    /**
+     * No people found in library
+     */
+    data class NoPeopleFound(
+        val message: String
+    ) : DiscoverUiState()
+
+    /**
+     * Error occurred
+     */
+    data class Error(
+        val message: String
+    ) : DiscoverUiState()
+}

app/src/main/java/com/placeholder/sherpai2/ui/navigation/AppDestinations.kt

@@ -53,25 +53,25 @@ sealed class AppDestinations(
     // FACE RECOGNITION
     // ==================
 
+    data object Discover : AppDestinations(
+        route = AppRoutes.DISCOVER,
+        icon = Icons.Default.AutoAwesome,
+        label = "Discover",
+        description = "Find people in your photos"
+    )
+
     data object Inventory : AppDestinations(
         route = AppRoutes.INVENTORY,
         icon = Icons.Default.Face,
-        label = "People Models",
-        description = "Existing Face Detection Models"
+        label = "People",
+        description = "Manage recognized people"
     )
 
     data object Train : AppDestinations(
         route = AppRoutes.TRAIN,
         icon = Icons.Default.ModelTraining,
-        label = "Create Model",
-        description = "Create a new Person Model"
-    )
-
-    data object Models : AppDestinations(
-        route = AppRoutes.MODELS,
-        icon = Icons.Default.SmartToy,
-        label = "Generative",
-        description = "AI Creation"
+        label = "Train Model",
+        description = "Create a new person model"
     )
 
     // ==================
@@ -117,9 +117,9 @@ val photoDestinations = listOf(
 
 // Face recognition section
 val faceRecognitionDestinations = listOf(
+    AppDestinations.Discover,  // ✨ NEW: Auto-cluster discovery
     AppDestinations.Inventory,
-    AppDestinations.Train,
-    AppDestinations.Models
+    AppDestinations.Train
 )
 
 // Organization section
@@ -145,9 +145,9 @@ fun getDestinationByRoute(route: String?): AppDestinations? {
         AppRoutes.SEARCH -> AppDestinations.Search
         AppRoutes.EXPLORE -> AppDestinations.Explore
         AppRoutes.COLLECTIONS -> AppDestinations.Collections
+        AppRoutes.DISCOVER -> AppDestinations.Discover
         AppRoutes.INVENTORY -> AppDestinations.Inventory
         AppRoutes.TRAIN -> AppDestinations.Train
-        AppRoutes.MODELS -> AppDestinations.Models
         AppRoutes.TAGS -> AppDestinations.Tags
         AppRoutes.UTILITIES -> AppDestinations.UTILITIES
         AppRoutes.SETTINGS -> AppDestinations.Settings

app/src/main/java/com/placeholder/sherpai2/ui/navigation/AppNavHost.kt

@@ -18,6 +18,7 @@ import com.placeholder.sherpai2.ui.album.AlbumViewScreen
 import com.placeholder.sherpai2.ui.album.AlbumViewModel
 import com.placeholder.sherpai2.ui.collections.CollectionsScreen
 import com.placeholder.sherpai2.ui.collections.CollectionsViewModel
+import com.placeholder.sherpai2.ui.discover.DiscoverPeopleScreen
 import com.placeholder.sherpai2.ui.explore.ExploreScreen
 import com.placeholder.sherpai2.ui.imagedetail.ImageDetailScreen
 import com.placeholder.sherpai2.ui.modelinventory.PersonInventoryScreen
@@ -32,15 +33,12 @@ import com.placeholder.sherpai2.ui.trainingprep.TrainingPhotoSelectorScreen
 import com.placeholder.sherpai2.ui.utilities.PhotoUtilitiesScreen
 import java.net.URLDecoder
 import java.net.URLEncoder
+import com.placeholder.sherpai2.ui.navigation.AppRoutes
 
 /**
- * AppNavHost - UPDATED with TrainingPhotoSelector integration
+ * AppNavHost - UPDATED with Discover People screen
  *
- * Changes:
- * - Replaced ImageSelectorScreen with TrainingPhotoSelectorScreen
- * - Shows ONLY photos with faces (hasFaces=true)
- * - Multi-select photo gallery for training
- * - Filters 10,000 photos → ~500 with faces for fast selection
+ * NEW: Replaces placeholder "Models" screen with auto-clustering face discovery
  */
 @Composable
 fun AppNavHost(
@@ -185,6 +183,22 @@ fun AppNavHost(
         // FACE RECOGNITION SYSTEM
         // ==========================================
 
+        /**
+         * DISCOVER PEOPLE SCREEN - ✨ NEW!
+         *
+         * Auto-clustering face discovery with spoon-feed naming flow:
+         * 1. Auto-clusters all faces in library (2-5 min)
+         * 2. Shows beautiful grid of discovered people
+         * 3. User taps to name each person
+         * 4. Captures: name, DOB, sibling relationships
+         * 5. Triggers deep background scan with age tagging
+         *
+         * Replaces: Old "Models" placeholder screen
+         */
+        composable(AppRoutes.DISCOVER) {
+            DiscoverPeopleScreen()
+        }
+
         /**
          * PERSON INVENTORY SCREEN
          */
@@ -197,7 +211,7 @@ fun AppNavHost(
         }
 
         /**
-         * TRAINING FLOW - UPDATED with TrainingPhotoSelector
+         * TRAINING FLOW - Manual training (still available)
          */
         composable(AppRoutes.TRAIN) { entry ->
             val trainViewModel: TrainViewModel = hiltViewModel()
@@ -235,15 +249,7 @@ fun AppNavHost(
         }
 
         /**
-         * TRAINING PHOTO SELECTOR - NEW: Custom gallery with face filtering
-         *
-         * Replaces native photo picker with custom selector that:
-         * - Shows ONLY photos with hasFaces=true
-         * - Multi-select with visual feedback
-         * - Face count badges on each photo
-         * - Enforces minimum 15 photos
-         *
-         * Result: User browses ~500 photos instead of 10,000!
+         * TRAINING PHOTO SELECTOR - Custom gallery with face filtering
          */
         composable(AppRoutes.TRAINING_PHOTO_SELECTOR) {
             TrainingPhotoSelectorScreen(
@@ -261,12 +267,12 @@ fun AppNavHost(
         }
 
         /**
-         * MODELS SCREEN
+         * MODELS SCREEN - DEPRECATED, kept for backwards compat
          */
         composable(AppRoutes.MODELS) {
             DummyScreen(
                 title = "AI Models",
-                subtitle = "Manage face recognition models"
+                subtitle = "Use 'Discover' instead"
             )
         }
 

app/src/main/java/com/placeholder/sherpai2/ui/navigation/AppRoutes.kt

@@ -17,9 +17,10 @@ object AppRoutes {
     const val IMAGE_DETAIL = "IMAGE_DETAIL"
 
     // Face recognition
+    const val DISCOVER = "discover"  // ✨ NEW: Auto-cluster face discovery
     const val INVENTORY = "inv"
     const val TRAIN = "train"
-    const val MODELS = "models"
+    const val MODELS = "models"  // DEPRECATED - kept for reference only
 
     // Organization
     const val TAGS = "tags"
@@ -30,7 +31,7 @@ object AppRoutes {
 
     // Internal training flow screens
     const val IMAGE_SELECTOR = "Image Selection"  // DEPRECATED - kept for reference only
-    const val TRAINING_PHOTO_SELECTOR = "training_photo_selector"  // NEW: Face-filtered gallery
+    const val TRAINING_PHOTO_SELECTOR = "training_photo_selector"  // Face-filtered gallery
     const val CROP_SCREEN = "CROP_SCREEN"
     const val TRAINING_SCREEN = "TRAINING_SCREEN"
     const val ScanResultsScreen = "First Scan Results"

app/src/main/java/com/placeholder/sherpai2/ui/presentation/AppDrawerContent.kt

@@ -21,7 +21,7 @@ import com.placeholder.sherpai2.ui.navigation.AppRoutes
 
 /**
  * SLIMMED DOWN AppDrawer - 280dp width, inline logo, cleaner sections
- * NOW WITH: Scrollable support for small phones + Collections item
+ * UPDATED: Discover People feature with sparkle icon ✨
  */
 @OptIn(ExperimentalMaterial3Api::class)
 @Composable
@@ -109,7 +109,7 @@ fun AppDrawerContent(
                 val photoItems = listOf(
                     DrawerItem(AppRoutes.SEARCH, "Search", Icons.Default.Search),
                     DrawerItem(AppRoutes.EXPLORE, "Explore", Icons.Default.Explore),
-                    DrawerItem(AppRoutes.COLLECTIONS, "Collections", Icons.Default.Collections)  // NEW!
+                    DrawerItem(AppRoutes.COLLECTIONS, "Collections", Icons.Default.Collections)
                 )
 
                 photoItems.forEach { item ->
@@ -126,9 +126,9 @@ fun AppDrawerContent(
                 DrawerSection(title = "Face Recognition")
 
                 val faceItems = listOf(
+                    DrawerItem(AppRoutes.DISCOVER, "Discover", Icons.Default.AutoAwesome), // ✨ UPDATED!
                     DrawerItem(AppRoutes.INVENTORY, "People", Icons.Default.Face),
-                    DrawerItem(AppRoutes.TRAIN, "Create Person", Icons.Default.ModelTraining),
-                    DrawerItem(AppRoutes.MODELS, "Models", Icons.Default.SmartToy)
+                    DrawerItem(AppRoutes.TRAIN, "Train Model", Icons.Default.ModelTraining)
                 )
 
                 faceItems.forEach { item ->

app/src/main/java/com/placeholder/sherpai2/ui/presentation/MainScreen.kt

@@ -15,7 +15,7 @@ import com.placeholder.sherpai2.ui.navigation.AppRoutes
 import kotlinx.coroutines.launch
 
 /**
- * Clean main screen - NO duplicate FABs, Collections support
+ * Clean main screen - NO duplicate FABs, Collections support, Discover People
  */
 @OptIn(ExperimentalMaterial3Api::class)
 @Composable
@@ -98,7 +98,6 @@ fun MainScreen() {
                                     )
                                 }
                             }
-                            // NOTE: Removed TAGS action - TagManagementScreen has its own inline FAB
                         }
                     },
                     colors = TopAppBarDefaults.topAppBarColors(
@@ -109,7 +108,6 @@ fun MainScreen() {
                     )
                 )
             }
-            // NOTE: NO floatingActionButton here - individual screens manage their own FABs inline
         ) { paddingValues ->
             AppNavHost(
                 navController = navController,
@@ -126,10 +124,11 @@ private fun getScreenTitle(route: String): String {
     return when (route) {
         AppRoutes.SEARCH -> "Search"
         AppRoutes.EXPLORE -> "Explore"
-        AppRoutes.COLLECTIONS -> "Collections"  // NEW!
+        AppRoutes.COLLECTIONS -> "Collections"
+        AppRoutes.DISCOVER -> "Discover People" // ✨ NEW!
         AppRoutes.INVENTORY -> "People"
         AppRoutes.TRAIN -> "Train New Person"
-        AppRoutes.MODELS -> "AI Models"
+        AppRoutes.MODELS -> "AI Models" // Deprecated, but keep for backwards compat
         AppRoutes.TAGS -> "Tag Management"
         AppRoutes.UTILITIES -> "Photo Util."
         AppRoutes.SETTINGS -> "Settings"
@@ -144,7 +143,8 @@ private fun getScreenSubtitle(route: String): String? {
     return when (route) {
         AppRoutes.SEARCH -> "Find photos by tags, people, or date"
         AppRoutes.EXPLORE -> "Browse your collection"
-        AppRoutes.COLLECTIONS -> "Your photo collections"  // NEW!
+        AppRoutes.COLLECTIONS -> "Your photo collections"
+        AppRoutes.DISCOVER -> "Auto-find faces in your library" // ✨ NEW!
         AppRoutes.INVENTORY -> "Trained face models"
         AppRoutes.TRAIN -> "Add a new person to recognize"
         AppRoutes.TAGS -> "Organize your photo collection"

app/src/main/java/com/placeholder/sherpai2/workers/Faceclusteringworker.kt

@@ -0,0 +1,113 @@
+package com.placeholder.sherpai2.workers
+
+import android.content.Context
+import androidx.hilt.work.HiltWorker
+import androidx.work.*
+import com.placeholder.sherpai2.domain.clustering.FaceClusteringService
+import dagger.assisted.Assisted
+import dagger.assisted.AssistedInject
+import kotlinx.coroutines.Dispatchers
+import kotlinx.coroutines.withContext
+
+/**
+ * FaceClusteringWorker - Background face clustering with persistence
+ *
+ * BENEFITS:
+ * - Survives app restarts
+ * - Runs even when app is backgrounded
+ * - Progress updates via WorkManager Data
+ * - Results saved to shared preferences
+ *
+ * USAGE:
+ * val workRequest = OneTimeWorkRequestBuilder<FaceClusteringWorker>()
+ *     .setConstraints(...)
+ *     .build()
+ * WorkManager.getInstance(context).enqueue(workRequest)
+ */
+@HiltWorker
+class FaceClusteringWorker @AssistedInject constructor(
+    @Assisted private val context: Context,
+    @Assisted workerParams: WorkerParameters,
+    private val clusteringService: FaceClusteringService
+) : CoroutineWorker(context, workerParams) {
+
+    companion object {
+        const val WORK_NAME = "face_clustering_discovery"
+        const val KEY_PROGRESS_CURRENT = "progress_current"
+        const val KEY_PROGRESS_TOTAL = "progress_total"
+        const val KEY_PROGRESS_MESSAGE = "progress_message"
+        const val KEY_CLUSTER_COUNT = "cluster_count"
+        const val KEY_FACE_COUNT = "face_count"
+        const val KEY_RESULT_JSON = "result_json"
+    }
+
+    override suspend fun doWork(): Result = withContext(Dispatchers.Default) {
+        try {
+            // Check if we should stop (work cancelled)
+            if (isStopped) {
+                return@withContext Result.failure()
+            }
+
+            withContext(Dispatchers.Main) {
+                setProgress(
+                    workDataOf(
+                        KEY_PROGRESS_CURRENT to 0,
+                        KEY_PROGRESS_TOTAL to 100,
+                        KEY_PROGRESS_MESSAGE to "Starting discovery..."
+                    )
+                )
+            }
+
+            // Run clustering
+            val result = clusteringService.discoverPeople(
+                onProgress = { current, total, message ->
+                    if (!isStopped) {
+                        kotlinx.coroutines.runBlocking {
+                            withContext(Dispatchers.Main) {
+                                setProgress(
+                                    workDataOf(
+                                        KEY_PROGRESS_CURRENT to current,
+                                        KEY_PROGRESS_TOTAL to total,
+                                        KEY_PROGRESS_MESSAGE to message
+                                    )
+                                )
+                            }
+                        }
+                    }
+                }
+            )
+
+            // Save result to SharedPreferences for ViewModel to read
+            val prefs = context.getSharedPreferences("face_clustering", Context.MODE_PRIVATE)
+            prefs.edit().apply {
+                putInt(KEY_CLUSTER_COUNT, result.clusters.size)
+                putInt(KEY_FACE_COUNT, result.totalFacesAnalyzed)
+                putLong("timestamp", System.currentTimeMillis())
+                // Don't serialize full result - too complex without proper setup
+                // Phase 2 will handle proper result persistence
+                apply()
+            }
+
+            // Success!
+            Result.success(
+                workDataOf(
+                    KEY_CLUSTER_COUNT to result.clusters.size,
+                    KEY_FACE_COUNT to result.totalFacesAnalyzed
+                )
+            )
+
+        } catch (e: Exception) {
+            // Save error state
+            val prefs = context.getSharedPreferences("face_clustering", Context.MODE_PRIVATE)
+            prefs.edit().apply {
+                putString("error", e.message ?: "Unknown error")
+                putLong("timestamp", System.currentTimeMillis())
+                apply()
+            }
+
+            Result.failure(
+                workDataOf("error" to (e.message ?: "Unknown error"))
+            )
+        }
+    }
+}