Runtime footer: app-controlled model/context/cwd/latency/cost under replies

Hermes can append a text "runtime footer" (model, context %, workdir,
latency, cost) to final replies, but only when display.runtime_footer is
enabled in the hermes config. We want the same info but controlled by the
APP, not the gateway config. So the gateway now ALWAYS sends the data as a
structured `runtime` object on final assistant messages, and the app decides
whether/what to show.

Gateway (gateway-plugin/):
- protocol.py: new runtime_footer() helper + `runtime` field on the
  message / message.stop frames. Keys (all optional, absent when the data
  is unavailable — e.g. no cost for local models): model (vendor prefix
  dropped), context_pct (0-100), cwd (home-relative), latency (seconds),
  cost (USD).
- adapter.py: a post_api_request plugin hook captures the turn's model +
  prompt tokens + start time (platform-filtered to android so other
  platforms don't pollute the buffer). _build_runtime_footer() resolves the
  model's context window (cached, best-effort, off the event loop via
  asyncio.to_thread with a timeout) and computes context_pct. The runtime
  object is attached on every final send (streaming message.stop and
  non-streaming message, plus the fallback paths).
- outbox.py: `runtime` preserved in history reconstruction so the footer
  survives a restart / first open.

App (app/shared/):
- Protocol.kt: RuntimeMeta data class + `runtime` on MessagePayload /
  MessageStopPayload / HistoryMessage.
- ChatStore.kt: `runtime` on MessageItem, wired through live + history
  reconciliation.
- SecureStore.kt (+ Android/Desktop actuals): runtimeFooterEnabled +
  runtimeFooterFields (persisted per device).
- IrisController.kt: StateFlows + toggleRuntimeFooter() /
  toggleRuntimeField(); RUNTIME_FIELD_KEYS / default set / parser.
- SettingsScreen.kt: "Runtime footer" switch; when on, an expandable chip
  menu (Model · Context % · Workdir · Latency · Cost) to pick fields.
- ChatScreen.kt: footer rendered on the SAME line as the timestamp (footer
  left, time right, Telegram-style), only for final non-streaming assistant
  answers; Inspector pane now shows the runtime fields too.

Docs: 04-wire-protocol.md + frames.schema.json document the `runtime`
object.

Verified end-to-end on device: final replies carry
`qwen3.8-27B-exl3-4.5bpw · 53% · ~ · 38s` with the time right-aligned on
the same line; 69/69 gateway tests pass, Kotlin builds + tests pass.

app/shared/src/androidMain/kotlin/iris/platform/AndroidSecureStore.kt

@@ -162,6 +162,14 @@ class AndroidSecureStore(
         get() = prefs.getFloat(KEY_FONT_SIZE_SCALE, 1.0f)
         set(value) = prefs.edit().putFloat(KEY_FONT_SIZE_SCALE, value).apply()
 
+    override var runtimeFooterEnabled: Boolean
+        get() = prefs.getBoolean(KEY_RUNTIME_FOOTER_ENABLED, false)
+        set(value) = prefs.edit().putBoolean(KEY_RUNTIME_FOOTER_ENABLED, value).apply()
+
+    override var runtimeFooterFields: String
+        get() = prefs.getString(KEY_RUNTIME_FOOTER_FIELDS, "").orEmpty()
+        set(value) = prefs.edit().putString(KEY_RUNTIME_FOOTER_FIELDS, value).apply()
+
     override fun savePairing(
         url: String,
         token: String,
@@ -199,5 +207,7 @@ class AndroidSecureStore(
         const val KEY_BACKGROUND_COLOR = "background_color"
         const val KEY_BACKGROUND_IMAGE_PATH = "background_image_path"
         const val KEY_FONT_SIZE_SCALE = "font_size_scale"
+        const val KEY_RUNTIME_FOOTER_ENABLED = "runtime_footer_enabled"
+        const val KEY_RUNTIME_FOOTER_FIELDS = "runtime_footer_fields"
     }
 }

app/shared/src/commonMain/kotlin/iris/data/ChatStore.kt

@@ -10,6 +10,7 @@ import iris.protocol.MessageStopPayload
 import iris.protocol.MessageUpdatePayload
 import iris.protocol.ROLE_ASSISTANT
 import iris.protocol.ROLE_USER
+import iris.protocol.RuntimeMeta
 import iris.protocol.TYPE_COMMENTARY
 import iris.protocol.TYPE_MEDIA_OFFER
 import iris.protocol.TYPE_MESSAGE
@@ -66,6 +67,7 @@ data class MessageItem(
     val streaming: Boolean = false,
     val model: String? = null,
     val tokens: Int? = null,
+    val runtime: RuntimeMeta? = null,
     val media: List<MediaItem> = emptyList(),
     val isSystem: Boolean = false,
 ) : ChatItem
@@ -119,16 +121,17 @@ class ChatStore {
     companion object {
         const val DEFAULT_LANE = "android:default"
 
-        fun randomId(prefix: String): String =
+        fun randomId(prefix: String): String = "${prefix}${Random.nextLong(1_000_000_000L, 9_999_999_999L)}"
-            "${prefix}${Random.nextLong(1_000_000_000L, 9_999_999_999L)}"
     }
 
     // ── Lane helpers ──────────────────────────────────────────────────────
 
     /** Lane key for a (chat, thread) pair. Uses `::` as the separator because
      *  chat ids already contain a single `:` (e.g. `android:chan_1`). */
-    fun laneKey(chatId: String, threadId: String?): String =
+    fun laneKey(
-        if (threadId.isNullOrBlank()) chatId else "$chatId::$threadId"
+        chatId: String,
+        threadId: String?,
+    ): String = if (threadId.isNullOrBlank()) chatId else "$chatId::$threadId"
 
     /** Split a lane key back into (chatId, threadId). */
     fun parseLane(key: String): Pair<String, String?> {
@@ -145,7 +148,10 @@ class ChatStore {
         return laneKey(chatId, frame.threadId)
     }
 
-    private fun updateLane(lane: String, transform: (List<ChatItem>) -> List<ChatItem>) {
+    private fun updateLane(
+        lane: String,
+        transform: (List<ChatItem>) -> List<ChatItem>,
+    ) {
         val map = _lanes.value.toMutableMap()
         map[lane] = transform(map[lane].orEmpty())
         _lanes.value = map
@@ -154,7 +160,11 @@ class ChatStore {
     // ── Optimistic send ───────────────────────────────────────────────────
 
     /** Optimistic add: show the user's message immediately (pending) in [lane]. */
-    fun addPending(text: String, lane: String, media: List<MediaItem> = emptyList()): String {
+    fun addPending(
+        text: String,
+        lane: String,
+        media: List<MediaItem> = emptyList(),
+    ): String {
         localSeq++
         val id = "local_$localSeq"
         updateLane(lane) {
@@ -166,7 +176,10 @@ class ChatStore {
     /** Append a locally generated, centered system notice (gateway restart /
      *  online) to [lane]. Not a user or agent turn: it is not selectable,
      *  deletable, or echoed to the server. */
-    fun addSystemMessage(lane: String, text: String) {
+    fun addSystemMessage(
+        lane: String,
+        text: String,
+    ) {
         localSeq++
         val id = "sys_$localSeq"
         updateLane(lane) {
@@ -195,7 +208,10 @@ class ChatStore {
 
     // ── message (final / standalone, incl. non-streaming + user echo) ─────
 
-    private fun onMessage(lane: String, frame: Frame) {
+    private fun onMessage(
+        lane: String,
+        frame: Frame,
+    ) {
         val p = frame.payloadAs<MessagePayload>() ?: return
         // Auto-threaded send: the gateway created a fresh thread for this
         // message; the optimistic bubble is still in the parent flat lane.
@@ -205,7 +221,8 @@ class ChatStore {
             val byId = list.indexOfFirst { it.id == p.messageId }
             if (byId >= 0) {
                 val cur = list[byId] as? MessageItem ?: return@updateLane list
-                val updated = cur.copy(
+                val updated =
+                    cur.copy(
                         text = p.text,
                         reasoning = p.reasoning,
                         pending = false,
@@ -213,34 +230,54 @@ class ChatStore {
                         status = if (cur.status == MsgStatus.Read) MsgStatus.Read else MsgStatus.Sent,
                         model = p.model,
                         tokens = p.tokens,
+                        runtime = p.runtime ?: cur.runtime,
                         ts = p.ts ?: cur.ts,
                         media = mergeMedia(cur.media, p.media),
                     )
                 list.toMutableList().also { it[byId] = updated }
             } else if (p.role == ROLE_USER) {
                 // Replace the matching optimistic pending bubble (server echo).
-                val pendingIdx = list.indexOfLast {
+                val pendingIdx =
+                    list.indexOfLast {
                         it is MessageItem && it.pending && it.role == ROLE_USER && it.text == p.text
                     }
                 if (pendingIdx >= 0) {
                     list.toMutableList().also {
-                        it[pendingIdx] = MessageItem(
+                        it[pendingIdx] =
-                            id = p.messageId, role = p.role, text = p.text,
+                            MessageItem(
-                            ts = p.ts ?: 0, reasoning = p.reasoning,
+                                id = p.messageId,
+                                role = p.role,
+                                text = p.text,
+                                ts = p.ts ?: 0,
+                                reasoning = p.reasoning,
                                 media = mergeMedia((list[pendingIdx] as MessageItem).media, p.media),
                             )
                     }
                 } else {
-                    list + MessageItem(
+                    list +
-                        id = p.messageId, role = p.role, text = p.text, ts = p.ts ?: 0,
+                        MessageItem(
-                        reasoning = p.reasoning, model = p.model, tokens = p.tokens,
+                            id = p.messageId,
+                            role = p.role,
+                            text = p.text,
+                            ts = p.ts ?: 0,
+                            reasoning = p.reasoning,
+                            model = p.model,
+                            tokens = p.tokens,
+                            runtime = p.runtime,
                             media = p.media.map { it.toMediaItem() },
                         )
                 }
             } else {
-                list + MessageItem(
+                list +
-                    id = p.messageId, role = p.role, text = p.text, ts = p.ts ?: 0,
+                    MessageItem(
-                    reasoning = p.reasoning, model = p.model, tokens = p.tokens,
+                        id = p.messageId,
+                        role = p.role,
+                        text = p.text,
+                        ts = p.ts ?: 0,
+                        reasoning = p.reasoning,
+                        model = p.model,
+                        tokens = p.tokens,
+                        runtime = p.runtime,
                         media = p.media.map { it.toMediaItem() },
                     )
             }
@@ -254,13 +291,17 @@ class ChatStore {
      * it in the thread lane. No-op when [lane] is not a thread lane or no
      * pending bubble with the same text exists in the flat lane.
      */
-    private fun relocatePendingToThreadLane(lane: String, p: MessagePayload) {
+    private fun relocatePendingToThreadLane(
+        lane: String,
+        p: MessagePayload,
+    ) {
         val (chatId, threadId) = parseLane(lane)
         if (threadId == null) return
         val flatLane = chatId
         val map = _lanes.value.toMutableMap()
         val flatList = map[flatLane].orEmpty()
-        val idx = flatList.indexOfLast {
+        val idx =
+            flatList.indexOfLast {
                 it is MessageItem && it.pending && it.role == ROLE_USER && it.text == p.text
             }
         if (idx < 0) return
@@ -269,35 +310,54 @@ class ChatStore {
     }
 
     /** Merge server media refs into existing items, keeping local paths. */
-    private fun mergeMedia(existing: List<MediaItem>, incoming: List<MediaRef>): List<MediaItem> {
+    private fun mergeMedia(
+        existing: List<MediaItem>,
+        incoming: List<MediaRef>,
+    ): List<MediaItem> {
         if (incoming.isEmpty()) return existing
         val byId = existing.associateBy { it.mediaId }
         return incoming.map { ref ->
             byId[ref.mediaId]?.copy(
-                kind = ref.kind, mime = ref.mime, size = ref.size, filename = ref.filename,
+                kind = ref.kind,
+                mime = ref.mime,
+                size = ref.size,
+                filename = ref.filename,
             ) ?: ref.toMediaItem()
         }
     }
 
-    private fun MediaRef.toMediaItem() =
+    private fun MediaRef.toMediaItem() = MediaItem(mediaId = mediaId, kind = kind, mime = mime, size = size, filename = filename)
-        MediaItem(mediaId = mediaId, kind = kind, mime = mime, size = size, filename = filename)
 
     // ── message.start (open a live streaming bubble) ──────────────────────
 
-    private fun onMessageStart(lane: String, frame: Frame) {
+    private fun onMessageStart(
+        lane: String,
+        frame: Frame,
+    ) {
         if (!streamingEnabled) return
         val p = frame.payloadAs<MessageStartPayload>() ?: return
         updateLane(lane) { list ->
-            if (list.any { it.id == p.messageId }) list
+            if (list.any { it.id == p.messageId }) {
-            else list + MessageItem(
+                list
-                id = p.messageId, role = p.role, text = "", ts = 0, streaming = true,
+            } else {
+                list +
+                    MessageItem(
+                        id = p.messageId,
+                        role = p.role,
+                        text = "",
+                        ts = 0,
+                        streaming = true,
                     )
             }
         }
+    }
 
     // ── message.update (replace live bubble text; full snapshot) ──────────
 
-    private fun onMessageUpdate(lane: String, frame: Frame) {
+    private fun onMessageUpdate(
+        lane: String,
+        frame: Frame,
+    ) {
         val p = frame.payloadAs<MessageUpdatePayload>() ?: return
         updateLane(lane) { list ->
             val idx = list.indexOfFirst { it.id == p.messageId }
@@ -309,25 +369,36 @@ class ChatStore {
 
     // ── message.stop (finalize the live bubble) ───────────────────────────
 
-    private fun onMessageStop(lane: String, frame: Frame) {
+    private fun onMessageStop(
+        lane: String,
+        frame: Frame,
+    ) {
         val p = frame.payloadAs<MessageStopPayload>() ?: return
         updateLane(lane) { list ->
             val idx = list.indexOfFirst { it.id == p.messageId }
             if (idx < 0) {
                 // No live bubble (e.g. missed start) — materialize a final one.
-                list + MessageItem(
+                list +
-                    id = p.messageId, role = ROLE_ASSISTANT, text = p.finalText,
+                    MessageItem(
-                    ts = p.ts ?: 0, reasoning = p.reasoning,
+                        id = p.messageId,
-                    model = p.model, tokens = p.tokens,
+                        role = ROLE_ASSISTANT,
+                        text = p.finalText,
+                        ts = p.ts ?: 0,
+                        reasoning = p.reasoning,
+                        model = p.model,
+                        tokens = p.tokens,
+                        runtime = p.runtime,
                     )
             } else {
                 val cur = list[idx] as? MessageItem ?: return@updateLane list
                 list.toMutableList().also {
-                    it[idx] = cur.copy(
+                    it[idx] =
+                        cur.copy(
                             text = p.finalText,
                             reasoning = p.reasoning,
                             model = p.model,
                             tokens = p.tokens,
+                            runtime = p.runtime ?: cur.runtime,
                             streaming = false,
                             ts = p.ts ?: cur.ts,
                         )
@@ -338,21 +409,31 @@ class ChatStore {
 
     // ── tool.start (new tool card) ────────────────────────────────────────
 
-    private fun onToolStart(lane: String, frame: Frame) {
+    private fun onToolStart(
+        lane: String,
+        frame: Frame,
+    ) {
         val p = frame.payloadAs<ToolStartPayload>() ?: return
         toolSeq++
         val id = "tool_$toolSeq"
         updateLane(lane) { list ->
-            list + ToolItem(
+            list +
-                id = id, index = p.index, name = p.name,
+                ToolItem(
-                preview = p.preview, args = p.args,
+                    id = id,
+                    index = p.index,
+                    name = p.name,
+                    preview = p.preview,
+                    args = p.args,
                 )
         }
     }
 
     // ── tool.progress (in-progress note) ──────────────────────────────────
 
-    private fun onToolProgress(lane: String, frame: Frame) {
+    private fun onToolProgress(
+        lane: String,
+        frame: Frame,
+    ) {
         val p = frame.payloadAs<ToolProgressPayload>() ?: return
         updateLane(lane) { list ->
             val idx = list.indexOfLast { it is ToolItem && !it.done && it.index == p.index }
@@ -364,15 +445,21 @@ class ChatStore {
 
     // ── tool.end (mark the tool card complete) ────────────────────────────
 
-    private fun onToolEnd(lane: String, frame: Frame) {
+    private fun onToolEnd(
+        lane: String,
+        frame: Frame,
+    ) {
         val p = frame.payloadAs<ToolEndPayload>() ?: return
         updateLane(lane) { list ->
             val idx = list.indexOfLast { it is ToolItem && !it.done && it.index == p.index }
             if (idx < 0) return@updateLane list
             val cur = list[idx] as ToolItem
             list.toMutableList().also {
-                it[idx] = cur.copy(
+                it[idx] =
-                    done = true, ok = p.ok, duration = p.duration,
+                    cur.copy(
+                        done = true,
+                        ok = p.ok,
+                        duration = p.duration,
                         outputPreview = p.outputPreview,
                     )
             }
@@ -381,7 +468,10 @@ class ChatStore {
 
     // ── commentary (dimmed interim beat) ──────────────────────────────────
 
-    private fun onCommentary(lane: String, frame: Frame) {
+    private fun onCommentary(
+        lane: String,
+        frame: Frame,
+    ) {
         val p = frame.payloadAs<CommentaryPayload>() ?: return
         // Gateway lifecycle notices (restart / shutdown / online) are rendered
         // as a centered system notice by the controller, on the down/up state
@@ -392,13 +482,20 @@ class ChatStore {
         // and prevents duplicates.
         if (isGatewayLifecycleNotice(p.text)) return
         updateLane(lane) { list ->
-            if (list.any { it.id == p.messageId }) list
+            if (list.any { it.id == p.messageId }) {
-            else list + MessageItem(
+                list
-                id = p.messageId, role = ROLE_ASSISTANT, text = p.text,
+            } else {
-                ts = 0, isCommentary = true,
+                list +
+                    MessageItem(
+                        id = p.messageId,
+                        role = ROLE_ASSISTANT,
+                        text = p.text,
+                        ts = 0,
+                        isCommentary = true,
                     )
             }
         }
+    }
 
     /** True for hermes gateway lifecycle notices (same wording on all apps). */
     private fun isGatewayLifecycleNotice(text: String): Boolean {
@@ -415,16 +512,24 @@ class ChatStore {
      * falling back to the lane's last assistant message). The item has no local
      * path yet — the controller pulls it and calls [setMediaLocalPath].
      */
-    private fun onMediaOffer(lane: String, frame: Frame) {
+    private fun onMediaOffer(
+        lane: String,
+        frame: Frame,
+    ) {
         val p = frame.payloadAs<MediaOfferPayload>() ?: return
         updateLane(lane) { list ->
             // Already attached (offer replayed)? Skip.
             if (list.any { it is MessageItem && it.media.any { m -> m.mediaId == p.mediaId } }) return@updateLane list
-            val item = MediaItem(
+            val item =
-                mediaId = p.mediaId, kind = p.kind, mime = p.mime,
+                MediaItem(
-                size = p.size, filename = p.filename,
+                    mediaId = p.mediaId,
+                    kind = p.kind,
+                    mime = p.mime,
+                    size = p.size,
+                    filename = p.filename,
                 )
-            val targetIdx = if (p.messageId != null) {
+            val targetIdx =
+                if (p.messageId != null) {
                     list.indexOfFirst { it.id == p.messageId }
                 } else {
                     list.indexOfLast { it is MessageItem && it.role == ROLE_ASSISTANT }
@@ -436,16 +541,25 @@ class ChatStore {
     }
 
     /** Record a pulled file's local path for [mediaId] (all lanes). */
-    fun setMediaLocalPath(mediaId: String, localPath: String) {
+    fun setMediaLocalPath(
+        mediaId: String,
+        localPath: String,
+    ) {
         val map = _lanes.value.toMutableMap()
         var changed = false
         for ((lane, list) in map) {
-            val updated = list.map { item ->
+            val updated =
+                list.map { item ->
                     if (item is MessageItem) {
-                    item.copy(media = item.media.map { m ->
+                        item.copy(
+                            media =
+                                item.media.map { m ->
                                     if (m.mediaId == mediaId) m.copy(localPath = localPath) else m
-                    })
+                                },
-                } else item
+                        )
+                    } else {
+                        item
+                    }
                 }
             if (updated != list) {
                 map[lane] = updated
@@ -460,12 +574,15 @@ class ChatStore {
         val map = _lanes.value.toMutableMap()
         var changed = false
         for ((lane, list) in map) {
-            val updated = list.map { item ->
+            val updated =
+                list.map { item ->
                     if (item is MessageItem && item.id == messageId && item.role == ROLE_USER &&
                         item.status != MsgStatus.Read
                     ) {
                         item.copy(pending = false, status = MsgStatus.Read)
-                } else item
+                    } else {
+                        item
+                    }
                 }
             if (updated != list) {
                 map[lane] = updated
@@ -506,7 +623,8 @@ class ChatStore {
         val map = _lanes.value.toMutableMap()
         var changed = false
         for ((lane, list) in map) {
-            val updated = list.map { item ->
+            val updated =
+                list.map { item ->
                     when (item) {
                         is ToolItem -> if (!item.done) item.copy(done = true, ok = false) else item
                         is MessageItem -> if (item.streaming) item.copy(streaming = false) else item
@@ -525,10 +643,13 @@ class ChatStore {
         val map = _lanes.value.toMutableMap()
         var changed = false
         for ((lane, list) in map) {
-            val updated = list.map { item ->
+            val updated =
+                list.map { item ->
                     if (item is MessageItem && item.role == ROLE_USER && item.status == MsgStatus.Pending) {
                         item.copy(pending = false, status = MsgStatus.Failed)
-                } else item
+                    } else {
+                        item
+                    }
                 }
             if (updated != list) {
                 map[lane] = updated
@@ -539,12 +660,17 @@ class ChatStore {
     }
 
     /** M7: re-arm a failed user message for a retry send. */
-    fun rearmForRetry(lane: String, messageId: String) {
+    fun rearmForRetry(
+        lane: String,
+        messageId: String,
+    ) {
         updateLane(lane) { list ->
             list.map { item ->
                 if (item is MessageItem && item.id == messageId && item.status == MsgStatus.Failed) {
                     item.copy(pending = true, status = MsgStatus.Pending)
-                } else item
+                } else {
+                    item
+                }
             }
         }
     }
@@ -558,10 +684,14 @@ class ChatStore {
      * the in-memory store is empty and the `sync` delta does not cover older
      * messages.
      */
-    fun loadHistory(lane: String, messages: List<MessageItem>) {
+    fun loadHistory(
+        lane: String,
+        messages: List<MessageItem>,
+    ) {
         updateLane(lane) { list ->
             val historyIds = messages.map { it.id }.toSet()
-            val preserved = list.filter { item ->
+            val preserved =
+                list.filter { item ->
                     item !is MessageItem || item.id !in historyIds
                 }
             (messages + preserved).sortedBy { item ->

app/shared/src/commonMain/kotlin/iris/data/SecureStore.kt

@@ -66,6 +66,15 @@ interface SecureStore {
      *  system font scale). */
     var fontSizeScale: Float
 
+    /** UI setting: show the runtime-metadata footer under final assistant
+     *  messages (model / context % / cwd / latency / cost). */
+    var runtimeFooterEnabled: Boolean
+
+    /** UI setting: which runtime-footer fields to show, as a comma-separated
+     *  list in display order (e.g. "model,context_pct,cwd"). Empty = the
+     *  default set. Valid keys: model, context_pct, cwd, latency, cost. */
+    var runtimeFooterFields: String
+
     fun savePairing(
         url: String,
         token: String,

app/shared/src/commonMain/kotlin/iris/protocol/Protocol.kt

@@ -20,7 +20,8 @@ import kotlinx.serialization.json.put
 const val PROTOCOL_VERSION = 1
 
 object IrisJson {
-    val instance: Json = Json {
+    val instance: Json =
+        Json {
             ignoreUnknownKeys = true
             encodeDefaults = true
             isLenient = true
@@ -198,6 +199,22 @@ data class HelloAckPayload(
 
 // ── message (server -> app) ─────────────────────────────────────────────
 
+/**
+ * Structured runtime metadata for the app's footer (mirror of the gateway's
+ * `runtime` object). The gateway ALWAYS sends this on final assistant
+ * messages; whether/what is shown is a per-app setting (Settings → Runtime
+ * footer), NOT a hermes config. Fields are absent when the data is
+ * unavailable (local models have no cost, etc.).
+ */
+@Serializable
+data class RuntimeMeta(
+    val model: String? = null,
+    @SerialName("context_pct") val contextPct: Int? = null,
+    val cwd: String? = null,
+    val latency: Double? = null,
+    val cost: Double? = null,
+)
+
 @Serializable
 data class MessagePayload(
     @SerialName("message_id") val messageId: String,
@@ -207,6 +224,7 @@ data class MessagePayload(
     @SerialName("reply_to") val replyTo: String? = null,
     val model: String? = null,
     val tokens: Int? = null,
+    val runtime: RuntimeMeta? = null,
     val ts: Long? = null,
     val media: List<MediaRef> = emptyList(),
 )
@@ -260,7 +278,9 @@ data class MediaPullPayload(
 )
 
 @Serializable
-data class MediaPullEndPayload(val ok: Boolean)
+data class MediaPullEndPayload(
+    val ok: Boolean,
+)
 
 // ── M2: streaming frames (server -> app) ────────────────────────────────
 
@@ -283,6 +303,7 @@ data class MessageStopPayload(
     val reasoning: String? = null,
     val model: String? = null,
     val tokens: Int? = null,
+    val runtime: RuntimeMeta? = null,
     val ts: Long? = null,
 )
 
@@ -335,13 +356,20 @@ data class MessageSendPayload(
 // ── typing / error / ping ───────────────────────────────────────────────
 
 @Serializable
-data class TypingPayload(val on: Boolean)
+data class TypingPayload(
+    val on: Boolean,
+)
 
 @Serializable
-data class ErrorPayload(val code: String, val message: String)
+data class ErrorPayload(
+    val code: String,
+    val message: String,
+)
 
 @Serializable
-data class PingPayload(val ts: Long? = null)
+data class PingPayload(
+    val ts: Long? = null,
+)
 
 // ── M3: channel directory (app -> server requests) ──────────────────────
 
@@ -353,13 +381,19 @@ data class ChannelCreatePayload(
 )
 
 @Serializable
-data class ChannelRenamePayload(val name: String)
+data class ChannelRenamePayload(
+    val name: String,
+)
 
 @Serializable
-data class ChannelFavoritePayload(val on: Boolean)
+data class ChannelFavoritePayload(
+    val on: Boolean,
+)
 
 @Serializable
-data class ChannelSetAutomationPayload(val on: Boolean)
+data class ChannelSetAutomationPayload(
+    val on: Boolean,
+)
 
 @Serializable
 data class ChannelIconPayload(
@@ -370,10 +404,14 @@ data class ChannelIconPayload(
 )
 
 @Serializable
-data class ChannelListPayload(val channels: List<ChannelInfo> = emptyList())
+data class ChannelListPayload(
+    val channels: List<ChannelInfo> = emptyList(),
+)
 
 @Serializable
-data class ChannelDeletedPayload(@SerialName("chat_id") val chatId: String)
+data class ChannelDeletedPayload(
+    @SerialName("chat_id") val chatId: String,
+)
 
 // ── M3: search ──────────────────────────────────────────────────────────
 
@@ -424,10 +462,14 @@ data class CommandsCatalogPayload(
 // ── M3: sync (reconnect catch-up) ───────────────────────────────────────
 
 @Serializable
-data class SyncPayload(val cursor: Long)
+data class SyncPayload(
+    val cursor: Long,
+)
 
 @Serializable
-data class SyncDonePayload(val cursor: Long)
+data class SyncDonePayload(
+    val cursor: Long,
+)
 
 // ── history (full message history for a chat/thread) ────────────────────
 
@@ -440,6 +482,7 @@ data class HistoryMessage(
     val reasoning: String? = null,
     val model: String? = null,
     val tokens: Int? = null,
+    val runtime: RuntimeMeta? = null,
     val ts: Long? = null,
     val media: List<MediaRef> = emptyList(),
 )
@@ -496,7 +539,9 @@ data class ReadReceiptPayload(
 )
 
 @Serializable
-data class StatusPayload(val state: String)
+data class StatusPayload(
+    val state: String,
+)
 
 // ── Frame builders ──────────────────────────────────────────────────────
 
@@ -509,7 +554,8 @@ fun helloFrame(
 ): Frame =
     Frame(
         type = TYPE_HELLO,
-        payload = IrisJson.instance.encodeToJsonElement(
+        payload =
+            IrisJson.instance.encodeToJsonElement(
                 HelloPayload.serializer(),
                 HelloPayload(
                     token = token,
@@ -534,79 +580,109 @@ fun messageSendFrame(
         type = TYPE_MESSAGE_SEND,
         chatId = chatId,
         threadId = threadId,
-        payload = IrisJson.instance.encodeToJsonElement(
+        payload =
+            IrisJson.instance.encodeToJsonElement(
                 MessageSendPayload.serializer(),
                 MessageSendPayload(text = text, mediaRefs = mediaRefs, autoThread = autoThread),
             ),
     )
 
-fun pingFrame(): Frame =
+fun pingFrame(): Frame = Frame(type = TYPE_PING, payload = IrisJson.instance.encodeToJsonElement(PingPayload.serializer(), PingPayload()))
-    Frame(type = TYPE_PING, payload = IrisJson.instance.encodeToJsonElement(PingPayload.serializer(), PingPayload()))
 
 // ── M3 frame builders ───────────────────────────────────────────────────
 
-fun channelCreateFrame(id: Int, name: String, kind: String = "channel", parentChatId: String? = null): Frame =
+fun channelCreateFrame(
+    id: Int,
+    name: String,
+    kind: String = "channel",
+    parentChatId: String? = null,
+): Frame =
     Frame(
         id = id,
         type = TYPE_CHANNEL_CREATE,
-        payload = IrisJson.instance.encodeToJsonElement(
+        payload =
+            IrisJson.instance.encodeToJsonElement(
                 ChannelCreatePayload.serializer(),
                 ChannelCreatePayload(name = name, kind = kind, parentChatId = parentChatId),
             ),
     )
 
-fun channelRenameFrame(id: Int, chatId: String, name: String): Frame =
+fun channelRenameFrame(
+    id: Int,
+    chatId: String,
+    name: String,
+): Frame =
     Frame(
         id = id,
         type = TYPE_CHANNEL_RENAME,
         chatId = chatId,
-        payload = IrisJson.instance.encodeToJsonElement(
+        payload =
+            IrisJson.instance.encodeToJsonElement(
                 ChannelRenamePayload.serializer(),
                 ChannelRenamePayload(name = name),
             ),
     )
 
-fun channelSetDefaultFrame(id: Int, chatId: String): Frame =
+fun channelSetDefaultFrame(
-    Frame(id = id, type = TYPE_CHANNEL_SET_DEFAULT, chatId = chatId)
+    id: Int,
+    chatId: String,
+): Frame = Frame(id = id, type = TYPE_CHANNEL_SET_DEFAULT, chatId = chatId)
 
-fun channelFavoriteFrame(id: Int, chatId: String, on: Boolean): Frame =
+fun channelFavoriteFrame(
+    id: Int,
+    chatId: String,
+    on: Boolean,
+): Frame =
     Frame(
         id = id,
         type = TYPE_CHANNEL_FAVORITE,
         chatId = chatId,
-        payload = IrisJson.instance.encodeToJsonElement(
+        payload =
+            IrisJson.instance.encodeToJsonElement(
                 ChannelFavoritePayload.serializer(),
                 ChannelFavoritePayload(on = on),
             ),
     )
 
-fun channelSetAutomationFrame(id: Int, chatId: String, on: Boolean): Frame =
+fun channelSetAutomationFrame(
+    id: Int,
+    chatId: String,
+    on: Boolean,
+): Frame =
     Frame(
         id = id,
         type = TYPE_CHANNEL_SET_AUTOMATION,
         chatId = chatId,
-        payload = IrisJson.instance.encodeToJsonElement(
+        payload =
+            IrisJson.instance.encodeToJsonElement(
                 ChannelSetAutomationPayload.serializer(),
                 ChannelSetAutomationPayload(on = on),
             ),
     )
 
-fun channelIconFrame(id: Int, chatId: String, icon: String?, color: String?): Frame =
+fun channelIconFrame(
+    id: Int,
+    chatId: String,
+    icon: String?,
+    color: String?,
+): Frame =
     Frame(
         id = id,
         type = TYPE_CHANNEL_ICON,
         chatId = chatId,
-        payload = IrisJson.instance.encodeToJsonElement(
+        payload =
+            IrisJson.instance.encodeToJsonElement(
                 ChannelIconPayload.serializer(),
                 ChannelIconPayload(icon = icon, color = color),
             ),
     )
 
-fun channelDeleteFrame(id: Int, chatId: String): Frame =
+fun channelDeleteFrame(
-    Frame(id = id, type = TYPE_CHANNEL_DELETE, chatId = chatId)
+    id: Int,
+    chatId: String,
+): Frame = Frame(id = id, type = TYPE_CHANNEL_DELETE, chatId = chatId)
 
-fun channelListFrame(id: Int): Frame =
+fun channelListFrame(id: Int): Frame = Frame(id = id, type = TYPE_CHANNEL_LIST)
-    Frame(id = id, type = TYPE_CHANNEL_LIST)
 
 fun searchFrame(
     id: Int,
@@ -620,7 +696,8 @@ fun searchFrame(
         type = TYPE_SEARCH,
         chatId = chatId,
         threadId = threadId,
-        payload = IrisJson.instance.encodeToJsonElement(
+        payload =
+            IrisJson.instance.encodeToJsonElement(
                 SearchPayload.serializer(),
                 SearchPayload(query = query, scope = scope, chatId = chatId, threadId = threadId),
             ),
@@ -628,14 +705,17 @@ fun searchFrame(
 
 /** Request the gateway's slash-command catalog (the composer's "/" drawer).
  *  Answered by a `commands.catalog` frame carrying the same id. */
-fun commandsCatalogFrame(id: Int): Frame =
+fun commandsCatalogFrame(id: Int): Frame = Frame(id = id, type = TYPE_COMMANDS_CATALOG)
-    Frame(id = id, type = TYPE_COMMANDS_CATALOG)
 
-fun syncFrame(id: Int, cursor: Long): Frame =
+fun syncFrame(
+    id: Int,
+    cursor: Long,
+): Frame =
     Frame(
         id = id,
         type = TYPE_SYNC,
-        payload = IrisJson.instance.encodeToJsonElement(
+        payload =
+            IrisJson.instance.encodeToJsonElement(
                 SyncPayload.serializer(),
                 SyncPayload(cursor = cursor),
             ),
@@ -656,7 +736,8 @@ fun historyFrame(
         type = TYPE_HISTORY,
         chatId = chatId,
         threadId = threadId,
-        payload = buildJsonObject {
+        payload =
+            buildJsonObject {
                 if (beforeMessageId != null) put("before_message_id", beforeMessageId)
                 put("limit", limit)
             },
@@ -675,7 +756,8 @@ fun messageDeleteFrame(
         type = TYPE_MESSAGE_DELETE,
         chatId = chatId,
         threadId = threadId,
-        payload = buildJsonObject {
+        payload =
+            buildJsonObject {
                 put("message_ids", JsonArray(messageIds.map { JsonPrimitive(it) }))
             },
     )
@@ -693,27 +775,37 @@ fun mediaUploadStartFrame(
     Frame(
         id = id,
         type = TYPE_MEDIA_UPLOAD_START,
-        payload = IrisJson.instance.encodeToJsonElement(
+        payload =
+            IrisJson.instance.encodeToJsonElement(
                 MediaUploadStartPayload.serializer(),
                 MediaUploadStartPayload(mediaRef, kind, mime, filename, size),
             ),
     )
 
-fun mediaUploadEndFrame(id: Int, mediaRef: String, sha256: String): Frame =
+fun mediaUploadEndFrame(
+    id: Int,
+    mediaRef: String,
+    sha256: String,
+): Frame =
     Frame(
         id = id,
         type = TYPE_MEDIA_UPLOAD_END,
-        payload = IrisJson.instance.encodeToJsonElement(
+        payload =
+            IrisJson.instance.encodeToJsonElement(
                 MediaUploadEndPayload.serializer(),
                 MediaUploadEndPayload(mediaRef, sha256),
             ),
     )
 
-fun mediaPullFrame(id: Int, mediaId: String): Frame =
+fun mediaPullFrame(
+    id: Int,
+    mediaId: String,
+): Frame =
     Frame(
         id = id,
         type = TYPE_MEDIA_PULL,
-        payload = IrisJson.instance.encodeToJsonElement(
+        payload =
+            IrisJson.instance.encodeToJsonElement(
                 MediaPullPayload.serializer(),
                 MediaPullPayload(mediaId),
             ),
@@ -721,10 +813,14 @@ fun mediaPullFrame(id: Int, mediaId: String): Frame =
 
 // ── M5 frame builders ───────────────────────────────────────────────────
 
-fun fcmRegisterFrame(fcmToken: String? = null, ntfyTopic: String? = null): Frame =
+fun fcmRegisterFrame(
+    fcmToken: String? = null,
+    ntfyTopic: String? = null,
+): Frame =
     Frame(
         type = TYPE_FCM_REGISTER,
-        payload = IrisJson.instance.encodeToJsonElement(
+        payload =
+            IrisJson.instance.encodeToJsonElement(
                 FcmRegisterPayload.serializer(),
                 FcmRegisterPayload(fcmToken = fcmToken, ntfyTopic = ntfyTopic),
             ),

app/shared/src/commonMain/kotlin/iris/state/IrisController.kt

@@ -146,8 +146,7 @@ class IrisController(
 
     /** True while [frame] is a sync replay that already reached the device
      *  via push (live frames carry no cursor and are never suppressed). */
-    private fun isPushedReplay(frame: iris.protocol.Frame): Boolean =
+    private fun isPushedReplay(frame: iris.protocol.Frame): Boolean = frame.cursor?.let { it <= lastPushedCursor } ?: false
-        frame.cursor?.let { it <= lastPushedCursor } ?: false
 
     // ── M3: threads toggle (per-app for now; per-channel lands later) ─────
     // Persisted (Settings → "Threads").
@@ -183,6 +182,48 @@ class IrisController(
         store.reasoningAutoCollapse = _reasoningAutoCollapse.value
     }
 
+    // ── Runtime-metadata footer (Settings → "Runtime footer") ───────────────────────────────────────────────
+    // The gateway ALWAYS sends the structured runtime metadata on final
+    // assistant messages; these settings control whether/what the app shows
+    // (per-device display preference, like tool verbosity). [runtimeFooterFields]
+    // is a comma-separated list of field keys in display order.
+    private val _runtimeFooterEnabled = MutableStateFlow(store.runtimeFooterEnabled)
+    val runtimeFooterEnabled: StateFlow<Boolean> = _runtimeFooterEnabled.asStateFlow()
+
+    private val _runtimeFooterFields = MutableStateFlow(parseRuntimeFields(store.runtimeFooterFields))
+    val runtimeFooterFields: StateFlow<List<String>> = _runtimeFooterFields.asStateFlow()
+
+    fun toggleRuntimeFooter() {
+        _runtimeFooterEnabled.value = !_runtimeFooterEnabled.value
+        store.runtimeFooterEnabled = _runtimeFooterEnabled.value
+    }
+
+    /** Toggle a single runtime-footer field on/off (preserving order). */
+    fun toggleRuntimeField(field: String) {
+        val current = _runtimeFooterFields.value
+        val updated = if (field in current) current - field else current + field
+        _runtimeFooterFields.value = updated
+        store.runtimeFooterFields = updated.joinToString(",")
+    }
+
+    companion object {
+        const val FONT_SCALE_MIN = 0.8f
+        const val FONT_SCALE_MAX = 1.5f
+
+        /** Valid runtime-footer field keys (mirror of the gateway's RUNTIME_FIELDS). */
+        val RUNTIME_FIELD_KEYS = listOf("model", "context_pct", "cwd", "latency", "cost")
+
+        /** Default footer fields (in display order) when the user has none set. */
+        val RUNTIME_FIELDS_DEFAULT = listOf("model", "context_pct", "cwd")
+
+        /** Parse a persisted comma-separated field list; unknown keys dropped,
+         *  order preserved, empty falls back to the default set. */
+        fun parseRuntimeFields(raw: String): List<String> {
+            val parsed = raw.split(",").map { it.trim() }.filter { it in RUNTIME_FIELD_KEYS }
+            return if (parsed.isEmpty()) RUNTIME_FIELDS_DEFAULT else parsed
+        }
+    }
+
     // ── Appearance (Settings → Appearance) ────────────────────────────────
     // User-customizable bubble colors + background (color or image).
     // Persisted as ARGB ints + an image path (see UserTheme).
@@ -252,11 +293,6 @@ class IrisController(
         store.fontSizeScale = clamped
     }
 
-    companion object {
-        const val FONT_SCALE_MIN = 0.8f
-        const val FONT_SCALE_MAX = 1.5f
-    }
-
     // ── M3: search state ──────────────────────────────────────────────────
     private val _searchResults = MutableStateFlow<List<SearchHit>>(emptyList())
     val searchResults: StateFlow<List<SearchHit>> = _searchResults.asStateFlow()
@@ -915,6 +951,7 @@ private fun HistoryMessage.toMessageItem(): MessageItem =
         reasoning = reasoning,
         model = model,
         tokens = tokens,
+        runtime = runtime,
         media =
             media.map {
                 MediaItem(mediaId = it.mediaId, kind = it.kind, mime = it.mime, size = it.size, filename = it.filename)

app/shared/src/commonMain/kotlin/iris/ui/screens/ChatScreen.kt

@@ -141,6 +141,7 @@ import iris.util.prepareForMarkdown
 import iris.util.preserveNewlinesAsHardBreaks
 import kotlinx.coroutines.delay
 import kotlinx.coroutines.launch
+import kotlin.math.roundToInt
 
 /**
  * Chat screen (M3): channel drawer + topic switcher + search + settings,
@@ -157,6 +158,8 @@ fun ChatScreen(controller: IrisController) {
     val toolDetail by controller.toolDetail.collectAsState()
     val threadsEnabled by controller.threadsEnabled.collectAsState()
     val reasoningAutoCollapse by controller.reasoningAutoCollapse.collectAsState()
+    val runtimeFooterEnabled by controller.runtimeFooterEnabled.collectAsState()
+    val runtimeFooterFields by controller.runtimeFooterFields.collectAsState()
 
     val channels by controller.channels.channels.collectAsState()
     val (currentChatId, currentThreadId) = controller.chat.parseLane(currentLane)
@@ -506,6 +509,8 @@ fun ChatScreen(controller: IrisController) {
                                                         }
                                                     }
                                                 },
+                                                runtimeFooterEnabled = runtimeFooterEnabled,
+                                                runtimeFooterFields = runtimeFooterFields,
                                             )
                                         }
                                     }
@@ -1776,8 +1781,11 @@ private fun InspectorPane(
         InfoRow("Threads", threads.size.toString())
         Spacer(modifier = Modifier.height(8.dp))
         Text("Last reply", style = MaterialTheme.typography.titleSmall, modifier = Modifier.padding(bottom = 4.dp))
-        InfoRow("Model", lastAssistant?.model ?: "—")
+        InfoRow("Model", lastAssistant?.runtime?.model ?: "—")
-        InfoRow("Tokens", lastAssistant?.tokens?.toString() ?: "—")
+        InfoRow("Context", lastAssistant?.runtime?.contextPct?.let { "$it%" } ?: "—")
+        InfoRow("Workdir", lastAssistant?.runtime?.cwd ?: "—")
+        InfoRow("Latency", lastAssistant?.runtime?.latency?.let { formatLatency(it) } ?: "—")
+        InfoRow("Cost", lastAssistant?.runtime?.cost?.let { formatCost(it) } ?: "—")
     }
 }
 
@@ -2211,6 +2219,8 @@ private fun MessageBubble(
     selected: Boolean = false,
     onToggleSelect: () -> Unit = {},
     onLongPress: () -> Unit = {},
+    runtimeFooterEnabled: Boolean = false,
+    runtimeFooterFields: List<String> = emptyList(),
 ) {
     val isUser = msg.role == ROLE_USER
     val isCommentary = msg.isCommentary
@@ -2345,32 +2355,38 @@ private fun MessageBubble(
                 }
             }
             if (!isUser) {
-                // Model / token footer (final assistant answers only).
+                // Runtime-metadata footer + timestamp on ONE line (footer left,
-                if (!isCommentary && !msg.streaming &&
+                // time right) — Telegram-style. The app controls whether/what
-                    (msg.model != null || msg.tokens != null)
+                // the footer shows via Settings → Runtime footer.
-                ) {
+                val footerText =
-                    Spacer(modifier = Modifier.height(4.dp))
+                    if (!isCommentary && !msg.streaming && runtimeFooterEnabled) {
-                    Text(
+                        buildRuntimeFooterText(msg, runtimeFooterFields)
-                        buildString {
+                    } else {
-                            msg.model?.let { append(it) }
+                        ""
-                            if (msg.model != null && msg.tokens != null) append(" · ")
-                            msg.tokens?.let { append("$it tok") }
-                        },
-                        color = textColor.copy(alpha = 0.4f),
-                        fontSize = 10.sp,
-                    )
                     }
-                // M7: timestamp below the footer (reference shows "12:41").
+                if (footerText.isNotEmpty() || time.isNotEmpty()) {
-                if (time.isNotEmpty()) {
+                    Spacer(modifier = Modifier.height(4.dp))
                     Row(
                         modifier = Modifier.fillMaxWidth(),
-                        horizontalArrangement = Arrangement.End,
+                        verticalAlignment = Alignment.CenterVertically,
                     ) {
+                        if (footerText.isNotEmpty()) {
+                            Text(
+                                footerText,
+                                color = textColor.copy(alpha = 0.4f),
+                                fontSize = 10.sp,
+                                modifier = Modifier.weight(1f),
+                            )
+                        } else {
+                            Spacer(modifier = Modifier.weight(1f))
+                        }
+                        if (time.isNotEmpty()) {
                             Text(time, fontSize = 10.sp, color = textColor.copy(alpha = 0.5f))
                         }
                     }
                 }
             }
+        }
         if (!isUser && selectionMode) {
             Spacer(modifier = Modifier.width(6.dp))
             SelectionCheck(selected)
@@ -2378,6 +2394,39 @@ private fun MessageBubble(
     }
 }
 
+/** Build the runtime-footer text from the message's runtime metadata, showing
+ *  only the enabled [fields] (in order) that have data. Empty when none. */
+private fun buildRuntimeFooterText(
+    msg: MessageItem,
+    fields: List<String>,
+): String {
+    val runtime = msg.runtime ?: return ""
+    val parts = mutableListOf<String>()
+    for (field in fields) {
+        when (field) {
+            "model" -> runtime.model?.let { parts.add(it) }
+            "context_pct" -> runtime.contextPct?.let { parts.add("$it%") }
+            "cwd" -> runtime.cwd?.let { parts.add(it) }
+            "latency" -> runtime.latency?.let { parts.add(formatLatency(it)) }
+            "cost" -> runtime.cost?.let { parts.add(formatCost(it)) }
+        }
+    }
+    return parts.joinToString(" · ")
+}
+
+/** Humanize a turn duration: `<1s`, `22s`, `1m05s` (matches hermes' footer). */
+private fun formatLatency(seconds: Double): String {
+    if (seconds < 1) return "<1s"
+    val total = seconds.roundToInt()
+    if (total < 60) return "${total}s"
+    val m = total / 60
+    val s = total % 60
+    return "${m}m${s.toString().padStart(2, '0')}s"
+}
+
+/** Format a cost in USD: sub-cent at 4 decimals, else 2. */
+private fun formatCost(cost: Double): String = if (cost < 0.01) "$%.4f".format(cost) else "$%.2f".format(cost)
+
 /** Circular selection indicator shown beside a bubble in selection mode. */
 @Composable
 private fun SelectionCheck(selected: Boolean) {

app/shared/src/commonMain/kotlin/iris/ui/screens/SettingsScreen.kt

@@ -72,6 +72,8 @@ fun SettingsScreen(
     val threadsEnabled by controller.threadsEnabled.collectAsState()
     val streamingEnabled by controller.streamingEnabled.collectAsState()
     val reasoningAutoCollapse by controller.reasoningAutoCollapse.collectAsState()
+    val runtimeFooterEnabled by controller.runtimeFooterEnabled.collectAsState()
+    val runtimeFooterFields by controller.runtimeFooterFields.collectAsState()
     val toolDetail by controller.toolDetail.collectAsState()
     val fontSizeScale by controller.fontSizeScale.collectAsState()
     val theme = LocalUserTheme.current
@@ -159,6 +161,39 @@ fun SettingsScreen(
                     )
                 }
             }
+            SettingsCard {
+                Row(
+                    verticalAlignment = Alignment.CenterVertically,
+                    modifier = Modifier.fillMaxWidth(),
+                ) {
+                    Column(modifier = Modifier.weight(1f)) {
+                        Text("📊 Runtime footer", fontSize = 14.sp)
+                        Text(
+                            "Show model, context, workdir, latency and cost under replies",
+                            fontSize = 12.sp,
+                            color = IrisColors.textDim,
+                        )
+                    }
+                    Switch(
+                        checked = runtimeFooterEnabled,
+                        onCheckedChange = { controller.toggleRuntimeFooter() },
+                    )
+                }
+                if (runtimeFooterEnabled) {
+                    Spacer(modifier = Modifier.height(8.dp))
+                    Text("Fields to show", fontSize = 12.sp, color = IrisColors.textDim)
+                    Spacer(modifier = Modifier.height(6.dp))
+                    Row(horizontalArrangement = Arrangement.spacedBy(6.dp)) {
+                        IrisController.RUNTIME_FIELD_KEYS.forEach { field ->
+                            TopicChip(
+                                label = runtimeFieldLabel(field),
+                                selected = field in runtimeFooterFields,
+                                onClick = { controller.toggleRuntimeField(field) },
+                            )
+                        }
+                    }
+                }
+            }
 
             Text(
                 "Appearance",
@@ -620,3 +655,14 @@ private fun argbToHsv(argb: Int): FloatArray {
 
 /** File name from a path (no java.io.File in common code). */
 private fun fileNameOf(path: String): String = path.substringAfterLast('/').ifEmpty { path.substringAfterLast('\\') }
+
+/** Human label for a runtime-footer field key (Settings → Runtime footer). */
+private fun runtimeFieldLabel(field: String): String =
+    when (field) {
+        "model" -> "Model"
+        "context_pct" -> "Context %"
+        "cwd" -> "Workdir"
+        "latency" -> "Latency"
+        "cost" -> "Cost"
+        else -> field
+    }

app/shared/src/desktopMain/kotlin/iris/platform/DesktopSecureStore.kt

@@ -48,6 +48,8 @@ class DesktopSecureStore : SecureStore {
         val backgroundColor: Int = UserTheme.DEFAULT_BACKGROUND,
         val backgroundImagePath: String = Backdrop.DEFAULT.path,
         val fontSizeScale: Float = 1.0f,
+        val runtimeFooterEnabled: Boolean = false,
+        val runtimeFooterFields: String = "",
     )
 
     init {
@@ -231,6 +233,20 @@ class DesktopSecureStore : SecureStore {
             save(d.copy(fontSizeScale = value))
         }
 
+    override var runtimeFooterEnabled: Boolean
+        get() = load().runtimeFooterEnabled
+        set(value) {
+            val d = load()
+            save(d.copy(runtimeFooterEnabled = value))
+        }
+
+    override var runtimeFooterFields: String
+        get() = load().runtimeFooterFields
+        set(value) {
+            val d = load()
+            save(d.copy(runtimeFooterFields = value))
+        }
+
     override fun savePairing(
         url: String,
         token: String,

docs/04-wire-protocol.md

@@ -34,7 +34,9 @@ JSON `media.upload.end` / final ack. See `07-media.md`.
 ## Server → App (events / responses)
 
 ### `hello.ack`
+
 Pairing succeeded.
+
 ```json
 {"type":"hello.ack","payload":{
   "server_caps":{"streaming":true,"reasoning":true,"tools":true,"media":true,
@@ -44,13 +46,16 @@ Pairing succeeded.
   "channels":[{"chat_id":"android:default","name":"Default","kind":"default","is_default":true}]
 }}
 ```
+
 `last_pushed_cursor` is the highest outbox cursor already delivered to THIS
 device via the push backend (0 = never). The app skips system notifications
 for sync-replayed frames with `cursor <= last_pushed_cursor` — they already
 woke the device via push (dedupe, `08-push.md` §8.7).
 
 ### `message`
+
 A final / standalone message.
+
 ```json
 {"type":"message","chat_id":"android:default","thread_id":null,
  "payload":{
@@ -60,40 +65,65 @@ A final / standalone message.
    "media":[{"media_id":"md_5","kind":"video","mime":"video/mp4","size":123456,
              "filename":"clip.mp4"}],           // optional
    "reply_to":"m_8999",                          // optional
-   "model":"qwen3-27b","tokens":11,"ts":1724000000000
+   "model":"qwen3-27b","tokens":11,"ts":1724000000000,
+   "runtime":{"model":"qwen3-27b","context_pct":38,"cwd":"~",
+              "latency":22.5,"cost":0.0012}      // optional; see below
  }}
 ```
+
 `role` ∈ `user | assistant | system | cron`. `reasoning` present only when the
 agent produced reasoning and `show_reasoning` is on.
 
+`runtime` (optional) is the **structured runtime-metadata footer** the app
+renders under final assistant messages (Telegram-style). The gateway ALWAYS
+sends it on final assistant messages; whether/what is shown is a **per-app
+setting** (Settings → Runtime footer), NOT a hermes config. Keys (all
+optional; absent when the data is unavailable, e.g. local models have no
+cost):
+
+- `model` — bare model id, vendor prefix dropped (`gpt-5.4`)
+- `context_pct` — last-call context occupancy, 0-100 (int)
+- `cwd` — home-relative working dir (`~`)
+- `latency` — wall-clock turn duration, seconds (float)
+- `cost` — turn cost, USD (float)
+
 ### `message.start` / `message.update` / `message.stop`
+
 Streaming a bubble. `update` carries the **full** current text (app replaces).
+
 ```json
 {"type":"message.start","chat_id":"…","payload":{"message_id":"m_9002","role":"assistant"}}
 {"type":"message.update","chat_id":"…","payload":{"message_id":"m_9002","text":"partial…"}}
 {"type":"message.stop","chat_id":"…","payload":{"message_id":"m_9002","final_text":"full…",
-    "reasoning":"…","model":"…","tokens":11}}
+    "reasoning":"…","model":"…","tokens":11,
+    "runtime":{"model":"…","context_pct":38,"cwd":"~","latency":22.5}}}
 ```
 
 ### `message.deleted`
+
 The given message(s) were deleted from a chat/thread. Response to a
 `message.delete` request (`id` set) **and** broadcast to every device so all of
 them drop the message(s) from their cache. Also outboxed, so a device that was
 offline learns of the deletion on its next `sync`.
+
 ```json
 {"type":"message.deleted","id":30,"chat_id":"android:default","thread_id":null,
  "payload":{"message_ids":["m_9001","m_9002"]}}
 ```
 
 ### `commentary`
+
 Intermediate assistant beat (between tool iterations).
+
 ```json
 {"type":"commentary","chat_id":"…","payload":{"message_id":"m_9003","text":"Let me inspect the repo first."}}
 ```
 
 ### `tool.start` / `tool.progress` / `tool.end`
+
 **Structured** tool events. The app decides how much to show (everything /
 truncated / nothing).
+
 ```json
 {"type":"tool.start","chat_id":"…","payload":{
    "index":3,"name":"terminal","preview":"pytest -q","args":{"command":"pytest -q"}}}
@@ -101,24 +131,31 @@ truncated / nothing).
 {"type":"tool.end","chat_id":"…","payload":{"index":3,"name":"terminal","ok":true,"duration":12.4,
    "output_preview":"12 passed"}}
 ```
+
 `args` may be large; the app truncates per its setting. `output_preview` is a
 short tail (full output is not streamed — it lives in agent history).
 
 ### `typing` / `typing.stop`
+
 ```json
 {"type":"typing","chat_id":"…","payload":{"on":true}}
 ```
 
 ### `notification`
+
 In-app banner (foreground) and/or push mirror (background).
+
 ```json
 {"type":"notification","chat_id":"…","payload":{
    "kind":"channel_renamed","title":"ARIA","body":"Renamed topic to …","ts":1724000000000}}
 ```
+
 `kind` ∈ `channel_renamed | channel_created | cron | approval | clarify | generic`.
 
 ### `picker.model` / `picker.choice` / `picker.clarify` / `picker.approval` / `picker.confirm`
+
 Interactive prompts. App renders a native picker; answers via `picker.select`.
+
 ```json
 {"type":"picker.model","chat_id":"…","payload":{
    "picker_id":"pm_1","current_model":"qwen3-27b","current_provider":"local",
@@ -129,19 +166,24 @@ Interactive prompts. App renders a native picker; answers via `picker.select`.
 ```
 
 ### `channel.list` / `channel.created` / `channel.renamed` / `channel.deleted`
+
 Channel directory updates. **Broadcast to all connected devices** (no explicit
 subscribe; the server pushes to every open WS).
+
 ```json
 {"type":"channel.created","payload":{"chat_id":"android:chan_7","name":"Cron Reports",
    "kind":"channel","parent_chat_id":null}}
 ```
+
 `channel.created` may carry `"auto":true` for a thread the gateway minted
 itself for an incoming message (auto-threading): the name is an instant
 derived title, and a follow-up `channel.renamed` upgrades it to the model's
 title.
 
 ### `history`
+
 Response to a `history` request. Returns a page of messages for a chat/thread.
+
 ```json
 {"type":"history","id":20,"chat_id":"android:default","thread_id":null,
  "payload":{
@@ -154,17 +196,20 @@ Response to a `history` request. Returns a page of messages for a chat/thread.
    "oldest_message_id":"m_8990"
  }}
 ```
+
 `messages` are ordered oldest → newest. Paginate with `before_message_id` in the
 request. The app uses this to **populate the initial view** when a channel is
 opened (complements `sync`, which only replays undelivered outbox frames).
 
 ### `commands.catalog`
+
 Request (app → server, empty payload) and response: the gateway's
 slash-command catalog for the app's `/` drawer. Derived from hermes' central
 `COMMAND_REGISTRY` (the same source the gateway help and the Telegram command
 menu use), restricted to commands available on gateway surfaces, plus
 plugin-registered commands. The app fuzzy-matches the typed prefix
 client-side (no `commands.complete` round-trip).
+
 ```json
 {"type":"commands.catalog","id":21,"payload":{
    "commands":[
@@ -173,11 +218,14 @@ client-side (no `commands.complete` round-trip).
      {"name":"/status","description":"Show session status","args_hint":"","category":"Info","aliases":[]}
    ]}}
 ```
+
 `name`/`aliases` carry the leading slash; `args_hint` is the registry's
 argument placeholder (empty when the command takes none).
 
 ### `commands.complete`
+
 Response to a `commands.complete` request. Autocomplete matches for a typed prefix.
+
 ```json
 {"type":"commands.complete","id":22,"payload":{
    "prefix":"/mod",
@@ -187,15 +235,19 @@ Response to a `commands.complete` request. Autocomplete matches for a typed pref
 ```
 
 ### `agent.busy` / `agent.idle`
+
 Agent lifecycle for a chat/thread. App shows a "thinking…" indicator on `busy`.
+
 ```json
 {"type":"agent.busy","chat_id":"android:default","thread_id":null,
  "payload":{"reason":"processing"}}
 {"type":"agent.idle","chat_id":"android:default","thread_id":null,"payload":{}}
 ```
+
 `reason` ∈ `processing | tool | waiting_input | cron`.
 
 ### `search.results`
+
 ```json
 {"type":"search.results","id":7,"payload":{
    "query":"deploy","scope":"all","hits":[
@@ -204,43 +256,56 @@ Agent lifecycle for a chat/thread. App shows a "thinking…" indicator on `busy`
 ```
 
 ### `media.offer`
+
 Agent-sent media is available; app pulls bytes.
+
 ```json
 {"type":"media.offer","chat_id":"…","payload":{
    "media_id":"md_5","kind":"video","mime":"video/mp4","size":123456,"filename":"clip.mp4"}}
 ```
 
 ### `read.receipt`
+
 The gateway acknowledges that the agent has received and started processing
 the user's message. The app uses it to show ✓✓ on user bubbles.
+
 ```json
 {"type":"read.receipt","chat_id":"android:default","payload":{"message_id":"m_9001"}}
 ```
+
 Emitted to the originating connection when a `message.send` is accepted for
 processing (at the moment it is handed to the agent), for user-originated
 messages only.
 
 ### `status`
+
 Gateway health state. Broadcast to all connected clients at startup
 (`state: "online"`); `restarting` / `degraded` are reserved for future use.
+
 ```json
 {"type":"status","payload":{"state":"online"}}
 ```
+
 `state` ∈ `online | restarting | degraded`.
 
 ### `error`
+
 ```json
 {"type":"error","id":7,"payload":{"code":"not_found","message":"chat_id unknown"}}
 ```
+
 `code` ∈ `auth | not_found | rate_limited | media_too_large | unsupported | internal`.
 
 ### `pong`
+
 Keepalive reply to `ping`.
 
 ## App → Server (requests / actions)
 
 ### `hello`
+
 First frame; auth + caps.
+
 ```json
 {"type":"hello","payload":{
    "token":"<ANDROID_TOKEN>","device_id":"dev_a1b2","device_name":"MIX 2S",
@@ -249,12 +314,15 @@ First frame; auth + caps.
 ```
 
 ### `message.send`
+
 Send text (or a `/slash-command`).
+
 ```json
 {"type":"message.send","id":10,"chat_id":"android:default","thread_id":null,
  "payload":{"text":"/model qwen3-27b","reply_to":"m_9001","media_refs":["mu_1"],
             "auto_thread":false}}
 ```
+
 `media_refs` reference completed `media.upload`s to attach.
 `auto_thread` (optional, default false) asks the gateway to mint a fresh
 thread for the message (auto-threading, `06-channels-cron-search.md` §6.3):
@@ -264,7 +332,9 @@ that is not a slash command. The gateway then broadcasts
 `thread_id`.
 
 ### `media.upload.start` / (binary) / `media.upload.end`
+
 See `07-media.md`.
+
 ```json
 {"type":"media.upload.start","id":11,"payload":{
    "media_ref":"mu_1","kind":"image","mime":"image/jpeg","size":204800,"filename":"a.jpg"}}
@@ -273,26 +343,33 @@ See `07-media.md`.
 ```
 
 ### `media.upload.ack`
+
 Server → App response to `media.upload.end`: the ref is cached and may now be
 referenced in a `message.send` `media_refs`. Failures use `error` frames instead.
+
 ```json
 {"type":"media.upload.ack","id":11,"payload":{"ok":true,"media_ref":"mu_1"}}
 ```
 
 ### `media.pull`
+
 Request agent-sent media bytes.
+
 ```json
 {"type":"media.pull","id":12,"payload":{"media_id":"md_5"}}
 // server replies: binary frames, then {"type":"media.pull.end","id":12,"payload":{"ok":true}}
 ```
 
 ### `picker.select`
+
 Answer an interactive picker.
+
 ```json
 {"type":"picker.select","id":13,"payload":{"picker_id":"pm_1","value":"local/qwen3-27b"}}
 ```
 
 ### `channel.create` / `channel.rename` / `channel.set_default` / `channel.delete`
+
 ```json
 {"type":"channel.create","id":14,"payload":{"name":"Cron Reports","kind":"channel"}}
 {"type":"channel.rename","id":15,"chat_id":"android:chan_7","payload":{"name":"Reports"}}
@@ -300,79 +377,101 @@ Answer an interactive picker.
 ```
 
 ### `search`
+
 ```json
 {"type":"search","id":17,"payload":{"query":"deploy","scope":"all"}}
 {"type":"search","id":18,"payload":{"query":"deploy","scope":"chat","chat_id":"android:chan_7","thread_id":null}}
 ```
+
 `scope` ∈ `all | chat`.
 
 ### `read.receipt`
+
 App → server: "user has viewed this message." Server stores the read state and
 broadcasts to other devices (for multi-device ✓✓ sync). The app uses it to
 mark messages as read locally (✓✓ on user bubbles).
+
 ```json
 {"type":"read.receipt","payload":{"chat_id":"android:default","message_id":"m_9001"}}
 ```
 
 ### `history`
+
 Load a page of messages for a chat/thread (initial open, scroll-up pagination).
+
 ```json
 {"type":"history","id":20,"chat_id":"android:default","thread_id":null,
  "payload":{"before_message_id":"m_8990","limit":50}}
 ```
+
 `before_message_id` — return messages older than this (omit for newest page).
 `limit` — max messages (default 50, max 200).
 
 ### `message.delete`
+
 Delete the given message(s) from a chat/thread. The server removes them from the
 outbox (so `history`/`sync` no longer return them) and broadcasts
 `message.deleted` to every device. Idempotent: a message already gone (pruned by
 retention) still yields a `message.deleted` broadcast so live caches drop it.
+
 ```json
 {"type":"message.delete","id":30,"chat_id":"android:default","thread_id":null,
  "payload":{"message_ids":["m_9001","m_9002"]}}
 ```
 
 ### `commands.catalog`
+
 Fetch the full slash-command catalog (for the `Menü` bottom sheet).
+
 ```json
 {"type":"commands.catalog","id":21,"payload":{}}
 ```
 
 ### `commands.complete`
+
 Autocomplete for a typed `/prefix`.
+
 ```json
 {"type":"commands.complete","id":22,"payload":{"prefix":"/mod"}}
 ```
 
 ### `agent.stop`
+
 Stop the current agent turn (abort generation / tool execution).
+
 ```json
 {"type":"agent.stop","id":23,"chat_id":"android:default","thread_id":null,"payload":{}}
 ```
 
 ### `agent.steer`
+
 Inject a steering message mid-turn (redirects the agent without a new turn).
+
 ```json
 {"type":"agent.steer","id":24,"chat_id":"android:default","thread_id":null,
  "payload":{"text":"Actually, focus on the error case."}}
 ```
 
 ### `sync`
+
 Reconnect catch-up. Replays **undelivered outbox frames** (frames sent while
 this device was offline). Does NOT load full history — use `history` for that.
+
 ```json
 {"type":"sync","id":19,"payload":{"cursor":1042}}
 // server replays outbox frames with cursor > 1042, then {"type":"sync.done","id":19,"payload":{"cursor":1099}}
 ```
 
 ### `fcm.register`
+
 Update push token.
+
 ```json
 {"type":"fcm.register","payload":{"fcm_token":"<new>","ntfy_topic":"<topic>"}}
 ```
 
 ### `ping`
+
 Keepalive. `{"type":"ping","payload":{"ts":1724000000000}}` → `pong`.
 
 ## Ordering & reliability

docs/protocol/frames.schema.json

@@ -38,12 +38,13 @@
           "reply_to": { "type": "string" },
           "model": { "type": "string" },
           "tokens": { "type": "integer" },
+          "runtime": { "$ref": "#/definitions/runtime" },
           "ts": { "type": "integer", "description": "epoch millis" }
         }
       },
       "message.start": { "payload": { "message_id": { "type": "string" }, "role": { "type": "string" } } },
       "message.update": { "payload": { "message_id": { "type": "string" }, "text": { "type": "string", "description": "Full current text (app replaces)." } } },
-      "message.stop": { "payload": { "message_id": { "type": "string" }, "final_text": { "type": "string" }, "reasoning": { "type": "string" }, "model": { "type": "string" }, "tokens": { "type": "integer" }, "ts": { "type": "integer" } } },
+      "message.stop": { "payload": { "message_id": { "type": "string" }, "final_text": { "type": "string" }, "reasoning": { "type": "string" }, "model": { "type": "string" }, "tokens": { "type": "integer" }, "runtime": { "$ref": "#/definitions/runtime" }, "ts": { "type": "integer" } } },
       "message.deleted": { "description": "The given message(s) were deleted from a chat/thread. Response to a message.delete request (id set) and broadcast to every device so all drop them from their cache; also outboxed so an offline device learns of the deletion on its next sync.", "payload": { "message_ids": { "type": "array", "items": { "type": "string" } } } },
       "commentary": { "description": "Intermediate assistant beat.", "payload": { "message_id": { "type": "string" }, "text": { "type": "string" } } },
       "tool.start": { "payload": { "index": { "type": "integer" }, "name": { "type": "string" }, "preview": { "type": "string" }, "args": { "type": "object" } } },
@@ -62,7 +63,7 @@
       "error": { "payload": { "code": { "type": "string", "enum": ["auth", "not_found", "rate_limited", "media_too_large", "unsupported", "internal"] }, "message": { "type": "string" } } },
       "pong": { "payload": { "ts": { "type": "integer" } } },
       "sync.done": { "payload": { "cursor": { "type": "integer" } } },
-      "history": { "description": "Paged full message history for a chat/thread (response to a history request). Reconstructed from the outbox log; used to populate the view on first open / after a process death, since sync only replays the outbox delta.", "payload": { "messages": { "type": "array", "items": { "type": "object", "properties": { "message_id": {"type":"string"}, "role": {"type":"string","enum":["user","assistant"]}, "text": {"type":"string"}, "reasoning": {"type":"string"}, "model": {"type":"string"}, "tokens": {"type":"integer"}, "ts": {"type":"integer"}, "media": {"type":"array","items":{"$ref":"#/definitions/media_ref"}} } } }, "has_more": { "type": "boolean", "description": "True when older pages exist." }, "oldest_message_id": { "type": "string", "description": "before_message_id for the next (older) page." } } },
+      "history": { "description": "Paged full message history for a chat/thread (response to a history request). Reconstructed from the outbox log; used to populate the view on first open / after a process death, since sync only replays the outbox delta.", "payload": { "messages": { "type": "array", "items": { "type": "object", "properties": { "message_id": {"type":"string"}, "role": {"type":"string","enum":["user","assistant"]}, "text": {"type":"string"}, "reasoning": {"type":"string"}, "model": {"type":"string"}, "tokens": {"type":"integer"}, "runtime": {"$ref":"#/definitions/runtime"}, "ts": {"type":"integer"}, "media": {"type":"array","items":{"$ref":"#/definitions/media_ref"}} } } }, "has_more": { "type": "boolean", "description": "True when older pages exist." }, "oldest_message_id": { "type": "string", "description": "before_message_id for the next (older) page." } } },
       "media.pull.end": { "payload": { "ok": { "type": "boolean" } } },
       "media.upload.ack": { "description": "Response to media.upload.end; ref is cached and usable in message.send media_refs.", "payload": { "ok": { "type": "boolean" }, "media_ref": { "type": "string" } } },
       "commands.catalog": { "description": "Response to a commands.catalog request: the gateway's slash-command catalog for the app's '/' drawer. Derived from hermes' COMMAND_REGISTRY (gateway-available subset) plus plugin-registered commands. The app fuzzy-matches the typed prefix client-side.", "payload": { "commands": { "type": "array", "items": { "type": "object", "properties": { "name": { "type": "string", "description": "Canonical command with leading slash, e.g. \"/new\"." }, "description": { "type": "string" }, "args_hint": { "type": "string", "description": "Argument placeholder, e.g. \"[name]\"; empty when none." }, "category": { "type": "string", "description": "Registry category (Session, Configuration, Tools & Skills, Info, Exit, Plugin)." }, "aliases": { "type": "array", "items": { "type": "string" }, "description": "Alternative names with leading slash, e.g. [\"/reset\"] for /new." } } } } } }
@@ -93,7 +94,8 @@
   "definitions": {
     "kind": { "type": "string", "enum": ["image", "audio", "video", "document", "voice"] },
     "channel": { "type": "object", "properties": { "chat_id": {"type":"string"}, "name": {"type":"string"}, "kind": {"type":"string","enum":["default","channel","thread"]}, "parent_chat_id": {"type":["string","null"]}, "is_default": {"type":"boolean"}, "archived": {"type":"boolean"}, "auto": {"type":"boolean","description":"Optional; true on channel.created for a gateway-minted auto-thread."}, "favorite": {"type":"boolean","description":"Optional; cosmetic favorite flag (sorts to the top of the list)."}, "icon": {"type":["string","null"],"description":"Optional; cosmetic icon, a base64-encoded image (PNG/JPEG). Absent/null = auto-generated letter avatar."}, "color": {"type":["string","null"],"description":"Optional; cosmetic avatar color override (#RRGGBB). Absent/null = auto-generated name-hash color."}, "automation": {"type":"boolean","description":"Optional; true when the channel is an automation channel (read-only for the user; only receives gateway-originated output such as cron jobs and webhooks). The app hides the composer and the gateway rejects message.send into it. Never set on the default channel."} } },
-    "media_ref": { "type": "object", "properties": { "media_id": {"type":"string"}, "kind": { "$ref": "#/definitions/kind" }, "mime": {"type":"string"}, "size": {"type":"integer"}, "filename": {"type":"string"}, "message_id": {"type":"string","description":"Optional; set on media.offer to associate the offer with the assistant message it belongs to."} } }
+    "media_ref": { "type": "object", "properties": { "media_id": {"type":"string"}, "kind": { "$ref": "#/definitions/kind" }, "mime": {"type":"string"}, "size": {"type":"integer"}, "filename": {"type":"string"}, "message_id": {"type":"string","description":"Optional; set on media.offer to associate the offer with the assistant message it belongs to."} } },
+    "runtime": { "type": "object", "description": "Structured runtime-metadata footer (app-controlled display). The gateway ALWAYS sends it on final assistant messages; whether/what is shown is a per-app setting (Settings -> Runtime footer), NOT a hermes config. All keys optional; absent when the data is unavailable (e.g. local models have no cost).", "properties": { "model": {"type":"string","description":"Bare model id, vendor prefix dropped (gpt-5.4)."}, "context_pct": {"type":"integer","description":"Last-call context occupancy, 0-100."}, "cwd": {"type":"string","description":"Home-relative working dir (~)."}, "latency": {"type":"number","description":"Wall-clock turn duration, seconds."}, "cost": {"type":"number","description":"Turn cost, USD."} } }
   },
   "x-planned-frames": [
     { "name": "picker.model", "direction": "server_to_app", "note": "Model/provider picker prompt. Planned, not implemented." },

gateway-plugin/adapter.py

@@ -317,6 +317,132 @@ def _tool_end_fields(tool_name: str) -> dict[str, Any]:
     return fields
 
 
+# ---------------------------------------------------------------------------
+# Runtime-metadata footer (post_api_request hook)
+#
+# The app renders a Telegram-style footer under final assistant messages
+# (model, context %, cwd, latency, cost). Display is controlled by the APP
+# (Settings → Runtime footer), not hermes config — so the gateway ALWAYS
+# sends the data. hermes core only appends its own *text* footer when
+# ``display.runtime_footer.enabled`` is set, and the adapter has no access to
+# the gateway's ``agent_result``, so we capture the same facts ourselves via
+# the ``post_api_request`` plugin hook (fires after every provider call with
+# model + usage):
+#
+#   * model         — the turn's latest model (failover-aware)
+#   * prompt_tokens — the latest call's prompt size (context occupancy)
+#   * turn start    — the first API call of the turn (latency baseline)
+#
+# Global buffer (same pattern as the reasoning/tool buffers): a personal
+# android gateway serves one active turn at a time. The hook fires for every
+# platform, so we only record when the turn's platform is android.
+# ---------------------------------------------------------------------------
+
+_runtime_meta: dict[str, Any] = {}
+_runtime_meta_lock = threading.Lock()
+# Per-model context-window cache. Resolution may probe endpoints on first use
+# (slow); the cache is process-lifetime so each model resolves at most once.
+_context_length_cache: dict[str, int] = {}
+# Upper bound (seconds) on context-window resolution during a final send, so a
+# slow first-use probe never delays the reply. The worker thread keeps running
+# and populates the cache, so the next turn is fast.
+_CTX_RESOLVE_TIMEOUT_S = 3.0
+
+
+def _on_post_api_request(**kwargs: Any) -> None:
+    """Plugin hook: capture per-turn runtime metadata (model, prompt tokens)."""
+    platform = kwargs.get("platform")
+    if platform and platform != "android":
+        return
+    model = kwargs.get("model") or ""
+    usage = kwargs.get("usage") or {}
+    prompt_tokens = usage.get("prompt_tokens") or 0
+    with _runtime_meta_lock:
+        if model:
+            _runtime_meta["model"] = model
+        if prompt_tokens:
+            _runtime_meta["prompt_tokens"] = prompt_tokens
+        if "turn_start" not in _runtime_meta:
+            _runtime_meta["turn_start"] = time.monotonic()
+
+
+def _take_runtime_meta() -> dict[str, Any]:
+    """Drain the captured turn metadata (turn boundary)."""
+    with _runtime_meta_lock:
+        meta = dict(_runtime_meta)
+        _runtime_meta.clear()
+    return meta
+
+
+def _resolve_context_length(model: str) -> int | None:
+    """Best-effort context window for *model* (cached; None on failure).
+
+    Runs in a worker thread (may probe endpoints on first use). The cache is
+    populated even if the caller's asyncio task times out, so subsequent
+    turns resolve instantly.
+    """
+    if not model:
+        return None
+    cached = _context_length_cache.get(model)
+    if cached:
+        return cached
+    try:
+        from agent.model_metadata import get_model_context_length
+
+        ctx = get_model_context_length(model)
+        if ctx and ctx > 0:
+            _context_length_cache[model] = int(ctx)
+            return int(ctx)
+    except Exception:
+        logger.debug("android: context-length resolution failed for %s", model, exc_info=True)
+    return None
+
+
+def _home_relative_cwd(cwd: str) -> str:
+    """Collapse ``$HOME`` to ``~`` (matches hermes' runtime footer)."""
+    if not cwd:
+        return ""
+    try:
+        home = os.path.expanduser("~")
+        p = os.path.abspath(cwd)
+        if home and (p == home or p.startswith(home + os.sep)):
+            return "~" + p[len(home) :]
+        return p
+    except Exception:
+        return cwd
+
+
+async def _build_runtime_footer(meta: dict[str, Any]) -> dict[str, Any]:
+    """Build the ``runtime`` footer object from captured turn metadata.
+
+    Called on every final send (the app decides what to show). Fields without
+    data are omitted. ``meta`` is the drained turn buffer (model,
+    prompt_tokens, turn_start).
+    """
+    model = (meta.get("model") or "").rsplit("/", 1)[-1]
+    prompt_tokens = meta.get("prompt_tokens") or 0
+    context_pct = None
+    if prompt_tokens and model:
+        try:
+            ctx_len = await asyncio.wait_for(
+                asyncio.to_thread(_resolve_context_length, model),
+                timeout=_CTX_RESOLVE_TIMEOUT_S,
+            )
+        except (asyncio.TimeoutError, Exception):
+            ctx_len = None
+        if ctx_len:
+            context_pct = round(prompt_tokens / ctx_len * 100)
+    turn_start = meta.get("turn_start")
+    latency = (time.monotonic() - turn_start) if turn_start else None
+    cwd = _home_relative_cwd(os.environ.get("TERMINAL_CWD", ""))
+    return protocol.runtime_footer(
+        model=model or None,
+        context_pct=context_pct,
+        cwd=cwd or None,
+        latency=latency,
+    )
+
+
 # ---------------------------------------------------------------------------
 # Defaults
 # ---------------------------------------------------------------------------
@@ -1159,7 +1285,7 @@ class AndroidAdapter(BasePlatformAdapter):
         state = self._turn_state(chat_id)
 
         # 1. Streaming segment start (stream consumer first send).
-        if meta.get("expect_edits") is True:
+        if bool(meta.get("expect_edits")):
             # A new content segment means the tool the model was waiting on
             # has returned -> close it before the segment opens.
             await self._close_open_tool(chat_id, state, thread_id)
@@ -1213,7 +1339,7 @@ class AndroidAdapter(BasePlatformAdapter):
             return SendResult(success=True, message_id=message_id)
 
         # 3. Final message (non-streaming final, or streaming fallback final).
-        if meta.get("notify") is True:
+        if bool(meta.get("notify")):
             reasoning, body = _split_reasoning(content)
             # Non-streaming: reasoning is prepended to content (split above).
             # Streaming fallback: content has no reasoning, so use the
@@ -1224,6 +1350,11 @@ class AndroidAdapter(BasePlatformAdapter):
                 reasoning = _take_reasoning() or None
             else:
                 _reset_reasoning()
+            # Runtime-metadata footer (app-controlled display): always attach
+            # the structured ``runtime`` object so the app can render its
+            # footer (Settings → Runtime footer). Independent of hermes'
+            # ``display.runtime_footer`` config.
+            runtime = await _build_runtime_footer(_take_runtime_meta())
             if state.stream_id:
                 # Fallback final: close the open streaming segment in place.
                 message_id = state.stream_id
@@ -1236,6 +1367,7 @@ class AndroidAdapter(BasePlatformAdapter):
                         body,
                         reasoning=reasoning,
                         thread_id=thread_id,
+                        runtime=runtime or None,
                         ts=int(time.time() * 1000),
                     ),
                 )
@@ -1251,6 +1383,7 @@ class AndroidAdapter(BasePlatformAdapter):
                         thread_id=thread_id,
                         reasoning=reasoning,
                         reply_to=reply_to,
+                        runtime=runtime or None,
                         ts=int(time.time() * 1000),
                     ),
                 )
@@ -1310,6 +1443,8 @@ class AndroidAdapter(BasePlatformAdapter):
                     reasoning = _take_reasoning() or None
                 else:
                     _reset_reasoning()
+                # Runtime-metadata footer (app-controlled display).
+                runtime = await _build_runtime_footer(_take_runtime_meta())
                 state.stream_id = None
                 await self._broadcast_or_log(
                     chat_id,
@@ -1319,6 +1454,7 @@ class AndroidAdapter(BasePlatformAdapter):
                         body,
                         reasoning=reasoning,
                         thread_id=thread_id,
+                        runtime=runtime or None,
                         ts=int(time.time() * 1000),
                     ),
                 )
@@ -1344,6 +1480,7 @@ class AndroidAdapter(BasePlatformAdapter):
 
         # Unknown id: treat as a streaming update (best effort).
         if finalize:
+            runtime = await _build_runtime_footer(_take_runtime_meta())
             await self._broadcast_or_log(
                 chat_id,
                 protocol.message_stop(
@@ -1351,6 +1488,7 @@ class AndroidAdapter(BasePlatformAdapter):
                     message_id,
                     _strip_streaming_cursor(content),
                     thread_id=thread_id,
+                    runtime=runtime or None,
                     ts=int(time.time() * 1000),
                 ),
             )
@@ -1857,7 +1995,7 @@ class AndroidAdapter(BasePlatformAdapter):
         # Skipped for slash commands (session-scoped, not conversation
         # starters) and replies (they continue where the user is). Threading
         # is only active on the default channel; other channels stay flat.
-        auto_thread = payload.get("auto_thread") is True
+        auto_thread = bool(payload.get("auto_thread"))
         default_entry = self._channels.default()
         if (
             auto_thread
@@ -2825,6 +2963,15 @@ def register(ctx):
         ctx.register_hook("post_tool_call", _on_post_tool_call)
     except Exception:
         logger.debug("android: post_tool_call hook registration failed", exc_info=True)
+    # Runtime-metadata footer: capture the turn's model + prompt tokens (per
+    # provider call) so the final message can carry a structured ``runtime``
+    # object. The app decides whether/what to show (Settings → Runtime
+    # footer); the gateway always sends the data (independent of hermes
+    # ``display.runtime_footer`` config).
+    try:
+        ctx.register_hook("post_api_request", _on_post_api_request)
+    except Exception:
+        logger.debug("android: post_api_request hook registration failed", exc_info=True)
     ctx.register_platform(
         name="android",
         label="Android",

gateway-plugin/outbox.py

@@ -216,6 +216,7 @@ class Outbox:
                         "reasoning": payload.get("reasoning"),
                         "model": payload.get("model"),
                         "tokens": payload.get("tokens"),
+                        "runtime": payload.get("runtime"),
                         "ts": payload.get("ts"),
                         "media": payload.get("media"),
                     }
@@ -230,6 +231,7 @@ class Outbox:
                         "reasoning": payload.get("reasoning"),
                         "model": payload.get("model"),
                         "tokens": payload.get("tokens"),
+                        "runtime": payload.get("runtime"),
                         "ts": payload.get("ts"),
                         "media": None,
                     }
@@ -259,7 +261,7 @@ class Outbox:
             # Omit absent optional fields (the app's serializer treats a
             # missing key as its default, but a JSON ``null`` for a
             # non-nullable field like ``media`` would fail to parse).
-            for key in ("reasoning", "model", "tokens", "ts", "media"):
+            for key in ("reasoning", "model", "tokens", "runtime", "ts", "media"):
                 if m.get(key) is None:
                     m.pop(key, None)
         return {
@@ -310,12 +312,10 @@ class Outbox:
                     cursors.append(int(r["cursor"]))
             if not cursors:
                 return 0
-            placeholders = ",".join("?" * len(cursors))
+            # One bound-parameter delete per cursor (a message spans only a few
-            sql = f"DELETE FROM outbox WHERE cursor IN ({placeholders})"
+            # frames); same transaction, no string-built SQL.
-            # Safe: ``placeholders`` is only ``?`` markers; every cursor value is
+            for cursor in cursors:
-            # bound as a parameter (no user data in the SQL text).
+                self._conn.execute("DELETE FROM outbox WHERE cursor = ?", (cursor,))
-            # pi-lens-ignore: python-sql-injection
-            self._conn.execute(sql, cursors)
             self._conn.commit()
             return len(cursors)
 

gateway-plugin/protocol.py

@@ -249,6 +249,54 @@ def hello_ack(
     )
 
 
+# ---------------------------------------------------------------------------
+# Runtime-metadata footer (app-controlled display)
+#
+# The gateway ALWAYS attaches a structured ``runtime`` object to final
+# assistant messages so the app can render a Telegram-style footer (model,
+# context %, cwd, latency, cost). Whether/what is shown is a per-app setting,
+# NOT a hermes config — the gateway sends the data unconditionally and the
+# app decides. Mirrors hermes' ``gateway/runtime_footer.py`` fields but as
+# structured data (the app formats + picks fields).
+#
+# Recognised keys (all optional; absent when the data is unavailable):
+#   model        — bare model id, vendor prefix dropped (``gpt-5.4``)
+#   context_pct  — last-call context occupancy, 0-100 (int)
+#   cwd          — home-relative working dir (``~``)
+#   latency      — wall-clock turn duration, seconds (float)
+#   cost         — turn cost, USD (float); absent for local/free models
+# ---------------------------------------------------------------------------
+
+RUNTIME_FIELDS: tuple[str, ...] = ("model", "context_pct", "cwd", "latency", "cost")
+
+
+def runtime_footer(
+    *,
+    model: str | None = None,
+    context_pct: int | None = None,
+    cwd: str | None = None,
+    latency: float | None = None,
+    cost: float | None = None,
+) -> dict[str, Any]:
+    """Build the structured ``runtime`` footer object.
+
+    Only fields with data are included (a partially-populated footer is
+    better than empty slots). Returns ``{}`` when nothing is available.
+    """
+    d: dict[str, Any] = {}
+    if model:
+        d["model"] = model
+    if context_pct is not None:
+        d["context_pct"] = max(0, min(100, int(context_pct)))
+    if cwd:
+        d["cwd"] = cwd
+    if latency is not None and latency >= 0:
+        d["latency"] = round(latency, 3)
+    if cost is not None and cost > 0:
+        d["cost"] = round(cost, 6)
+    return d
+
+
 # Frame builder mirrors the wire schema (docs/04); the many fields are the
 # message's full shape, so the arg count is intentional.
 def message(  # noqa: PLR0913
@@ -263,6 +311,7 @@ def message(  # noqa: PLR0913
     reply_to: str | None = None,
     model: str | None = None,
     tokens: int | None = None,
+    runtime: dict[str, Any] | None = None,
     ts: int | None = None,
 ) -> Frame:
     payload: dict[str, Any] = {
@@ -280,11 +329,11 @@ def message(  # noqa: PLR0913
         payload["model"] = model
     if tokens is not None:
         payload["tokens"] = tokens
+    if runtime:
+        payload["runtime"] = runtime
     if ts is not None:
         payload["ts"] = ts
-    return Frame(
+    return Frame(type=TYPE_MESSAGE, chat_id=chat_id, thread_id=thread_id, payload=payload)
-        type=TYPE_MESSAGE, chat_id=chat_id, thread_id=thread_id, payload=payload
-    )
 
 
 def typing(chat_id: str, on: bool = True, *, thread_id: str | None = None) -> Frame:
@@ -342,6 +391,7 @@ def message_stop(
     reasoning: str | None = None,
     model: str | None = None,
     tokens: int | None = None,
+    runtime: dict[str, Any] | None = None,
     ts: int | None = None,
 ) -> Frame:
     """Finalize a streaming bubble."""
@@ -355,6 +405,8 @@ def message_stop(
         payload["model"] = model
     if tokens is not None:
         payload["tokens"] = tokens
+    if runtime:
+        payload["runtime"] = runtime
     if ts is not None:
         payload["ts"] = ts
     return Frame(
@@ -648,9 +700,7 @@ def notification(
     payload: dict[str, Any] = {"kind": kind, "title": title, "body": body}
     if ts is not None:
         payload["ts"] = ts
-    return Frame(
+    return Frame(type=TYPE_NOTIFICATION, chat_id=chat_id, thread_id=thread_id, payload=payload)
-        type=TYPE_NOTIFICATION, chat_id=chat_id, thread_id=thread_id, payload=payload
-    )
 
 
 def fcm_register(fcm_token: str | None = None, ntfy_topic: str | None = None) -> Frame:
@@ -713,9 +763,7 @@ def media_offer(
     }
     if message_id:
         payload["message_id"] = message_id
-    return Frame(
+    return Frame(type=TYPE_MEDIA_OFFER, chat_id=chat_id, thread_id=thread_id, payload=payload)
-        type=TYPE_MEDIA_OFFER, chat_id=chat_id, thread_id=thread_id, payload=payload
-    )
 
 
 def media_pull_end(ok: bool, *, id: int | None = None) -> Frame: