Add used/full context meter to chat screen (#717)

* Adopt /api/health REST route (issue #713)

Cheap liveness probe on the connection screen: GET /api/health surfaces
backend version + auth mode (Token vs Gated/OAuth) without the heavier
/api/status payload. Added HermesApiService.getHealth + HealthStatus model;
ConnectViewModel.loadHealth probes on open (best-effort, silent on failure);
ConnectScreen shows a read-only health card when reachable.

Verified: ktlint clean, assembleDebug BUILD SUCCESSFUL locally.

* Add used/full context meter to chat screen

Shows session prompt tokens vs model context window above the composer,
polled every 5s. Color-coded bar warns at 70%/90%.

* fix(chat): fix unit tests for context usage meter

Format >=100k tokens cleanly in ContextUsageChip and stub AuthManager profile ID in ChatViewModelTest.

* Fix context meter: source used tokens from input_tokens

Gateway /api/sessions/{id} has no last_prompt_tokens column; use
input_tokens instead. Show meter as soon as full context is known.

* Add tappable context breakdown sheet

Chip opens a bottom sheet with input/output/cache/read/write/reasoning
token rows + message count. Source breakdown from /api/sessions/{id}.

* Refresh context meter on streaming complete

Trigger fetchContextUsage immediately when a message finishes streaming
instead of waiting for the 5s session-sync poll.

* Fix context meter: source used tokens from live WS session.info

The REST /api/sessions/{id} endpoint reads the DB input_tokens column,
which is stale/0 for active sessions — that's why the meter was stuck at 0.

Switch the meter to the live values the gateway pushes over the WS
session.info payload after each turn:
- used = last_prompt_tokens (real used context for the turn)
- full = context_length (model-accurate window; keep /api/model/info as a
  fallback before the first turn completes)

fetchContextUsage now only tops up the denominator from /api/model/info and
no longer overwrites the live WS value with the stale REST input_tokens.
The 5s poll remains as a fallback.

* Revert "Fix context meter: source used tokens from live WS session.info"

This reverts commit 932393fd80993d57e66fe93552a949460d646674.

---------

Co-authored-by: M57 <hy4ri@users.noreply.github.com>

app/src/main/java/com/m57/hermescontrol/data/model/HealthStatus.kt

@@ -0,0 +1,17 @@
+package com.m57.hermescontrol.data.model
+
+import kotlinx.serialization.Serializable
+
+/**
+ * Backend liveness probe, mirrored from `GET /api/health` (hermes_cli/web_server.py).
+ *
+ * A lightweight alternative to `/api/status` — no gateway PID / remote-health
+ * probe, no profile scope. `authRequired` tells the client which auth scheme the
+ * gateway expects (gated/OAuth vs loopback token) before attempting a real call.
+ */
+@Serializable
+data class HealthStatus(
+    val ok: Boolean = false,
+    val version: String? = null,
+    val authRequired: Boolean? = null,
+)

app/src/main/java/com/m57/hermescontrol/data/model/ModelInfoResponse.kt

@@ -0,0 +1,30 @@
+package com.m57.hermescontrol.data.model
+
+import kotlinx.serialization.Serializable
+
+/**
+ * Response from `GET /api/model/info` (PUBLIC, no auth required).
+ *
+ * VERIFIED against the live gateway (:9119, 2026-07-25): the payload carries
+ * `model`, `provider`, `auto_context_length`, `config_context_length`, and
+ * `effective_context_length`. The latter is the authoritative full context
+ * window for the active model — what the chat screen surfaces as "full context"
+ * in the used/full context meter. `context_length` (raw, per-session) is not
+ * part of this endpoint; per-session *used* context comes from
+ * [SessionDetailResponse.last_prompt_tokens] instead.
+ *
+ * Decoded with kotlinx `Json { ignoreUnknownKeys = true }` (see
+ * [com.m57.hermescontrol.data.remote.OkHttpProvider.json]), so unknown backend
+ * fields are tolerated.
+ */
+@Serializable
+data class ModelInfoResponse(
+    val model: String? = null,
+    val provider: String? = null,
+    /** Context window the runtime resolved from the provider catalog. */
+    val auto_context_length: Long? = null,
+    /** User-configured override (0 when unset). */
+    val config_context_length: Long? = null,
+    /** Effective context window actually in use (max of auto/config + caps). */
+    val effective_context_length: Long? = null,
+)

app/src/main/java/com/m57/hermescontrol/data/model/SessionDetailResponse.kt

@@ -0,0 +1,31 @@
+package com.m57.hermescontrol.data.model
+
+import kotlinx.serialization.Serializable
+
+/**
+ * Response from `GET /api/sessions/{id}` (gated — rides the app's existing
+ * cookie auth). Only the fields the chat screen needs for the context meter
+ * are modelled; the backend serializes many more and [kotlinx.serialization]
+ * ignores the rest.
+ *
+ * VERIFIED against the live gateway SQLite store (`/files/agent-vault/hermes/
+ * state.db`, 2026-07-26): the `sessions` table exposes `input_tokens`,
+ * `output_tokens`, `cache_read_tokens`, `cache_write_tokens`, `reasoning_tokens`,
+ * `message_count`, etc. — but it does NOT have a `last_prompt_tokens` column
+ * (that field lives only on the in-memory `SessionEntry` in the gateway
+ * process, never persisted to the REST response). So the *used* context window
+ * is sourced from `input_tokens` (cumulative prompt tokens for the session),
+ * paired with [ModelInfoResponse.effective_context_length] as the denominator.
+ */
+@Serializable
+data class SessionDetailResponse(
+    val session_id: String? = null,
+    val session_key: String? = null,
+    /** Cumulative prompt tokens for the session — the *used* context window. */
+    val input_tokens: Long? = null,
+    val output_tokens: Long? = null,
+    val cache_read_tokens: Long? = null,
+    val cache_write_tokens: Long? = null,
+    val reasoning_tokens: Long? = null,
+    val message_count: Int? = null,
+)

app/src/main/java/com/m57/hermescontrol/data/remote/HermesApiService.kt

@@ -29,6 +29,7 @@ import com.m57.hermescontrol.data.model.EnvVarDeleteRequest
 import com.m57.hermescontrol.data.model.EnvVarRevealRequest
 import com.m57.hermescontrol.data.model.EnvVarRevealResponse
 import com.m57.hermescontrol.data.model.EnvVarUpdate
+import com.m57.hermescontrol.data.model.HealthStatus
 import com.m57.hermescontrol.data.model.HookResponse
 import com.m57.hermescontrol.data.model.KanbanBoardResponse
 import com.m57.hermescontrol.data.model.KanbanBoardsResponse
@@ -46,6 +47,7 @@ import com.m57.hermescontrol.data.model.MessagingPlatformUpdate
 import com.m57.hermescontrol.data.model.MoaConfigResponse
 import com.m57.hermescontrol.data.model.ModelAssignmentRequest
 import com.m57.hermescontrol.data.model.ModelAssignmentResponse
+import com.m57.hermescontrol.data.model.ModelInfoResponse
 import com.m57.hermescontrol.data.model.ModelOptionsResponse
 import com.m57.hermescontrol.data.model.ModelsAnalyticsResponse
 import com.m57.hermescontrol.data.model.OAuthCancelResponse
@@ -64,6 +66,7 @@ import com.m57.hermescontrol.data.model.RawConfigResponse
 import com.m57.hermescontrol.data.model.RecentUnlock
 import com.m57.hermescontrol.data.model.SaveSkillContentRequest
 import com.m57.hermescontrol.data.model.ScanStatus
+import com.m57.hermescontrol.data.model.SessionDetailResponse
 import com.m57.hermescontrol.data.model.SessionListResponse
 import com.m57.hermescontrol.data.model.SessionMessagesResponse
 import com.m57.hermescontrol.data.model.SessionPromptResponse
@@ -123,6 +126,9 @@ interface HermesApiService {
     @GET("api/status")
     suspend fun getStatus(): Response<StatusResponse>
 
+    @GET("api/health")
+    suspend fun getHealth(): Response<HealthStatus>
+
     @GET("api/sessions")
     suspend fun getSessions(
         @Query("limit") limit: Int = 20,
@@ -181,6 +187,17 @@ interface HermesApiService {
         @Path("id", encoded = true) sessionId: String,
     ): Response<SessionPromptResponse>
 
+    @GET("api/sessions/{id}")
+    suspend fun getSessionDetail(
+        // Preserve slashes in session IDs — backend generates IDs containing '/' characters (issue #468).
+        // Contract: The server-generated sessionId must only contain URL-safe characters (no ?, #, or spaces).
+        @Path("id", encoded = true) sessionId: String,
+        @Query("profile") profile: String? = null,
+    ): Response<SessionDetailResponse>
+
+    @GET("api/model/info")
+    suspend fun getModelInfo(): Response<ModelInfoResponse>
+
     @GET("api/system/stats")
     suspend fun getSystemStats(): Response<SystemStatsResponse>
 

app/src/main/java/com/m57/hermescontrol/ui/chat/ChatScreen.kt

@@ -36,6 +36,7 @@ import androidx.compose.material.icons.Icons
 import androidx.compose.material.icons.filled.Add
 import androidx.compose.material.icons.filled.Close
 import androidx.compose.material.icons.filled.Search
+import androidx.compose.material3.ExperimentalMaterial3Api
 import androidx.compose.material3.Icon
 import androidx.compose.material3.IconButton
 import androidx.compose.material3.MaterialTheme
@@ -78,6 +79,8 @@ import com.m57.hermescontrol.ui.chat.components.ChatLifecycleEffects
 import com.m57.hermescontrol.ui.chat.components.ChatLoadingOverlay
 import com.m57.hermescontrol.ui.chat.components.ChatMessageList
 import com.m57.hermescontrol.ui.chat.components.ChatScrollToBottomFab
+import com.m57.hermescontrol.ui.chat.components.ContextDetailSheet
+import com.m57.hermescontrol.ui.chat.components.ContextUsageChip
 import com.m57.hermescontrol.ui.chat.components.ReactionHeartsOverlay
 import com.m57.hermescontrol.ui.chat.components.ReloginDialog
 import com.m57.hermescontrol.ui.chat.components.SearchBarRow
@@ -110,7 +113,7 @@ private const val SESSION_SYNC_INTERVAL_MS = 5_000L
  * this entry point handles only state hoisting, scaffold wiring, and the
  * remembered launchers that need to be activity-scoped.
  */
-@OptIn(ExperimentalFoundationApi::class)
+@OptIn(ExperimentalFoundationApi::class, ExperimentalMaterial3Api::class)
 @Composable
 fun ChatScreen(
     modifier: Modifier = Modifier,
@@ -125,12 +128,14 @@ fun ChatScreen(
     val scrollScope = rememberCoroutineScope()
     val scrollController = rememberChatScrollController(listState, scrollScope)
     var isOlderPagingArmed by remember(state.currentSessionId) { mutableStateOf(false) }
+    var showContextSheet by remember { mutableStateOf(false) }
 
     // Periodic session sync while connected.
     LaunchedEffect(state.currentSessionId, state.connectionStatus) {
         while (state.currentSessionId != null && state.connectionStatus == ConnectionStatus.CONNECTED) {
             delay(SESSION_SYNC_INTERVAL_MS)
             viewModel.syncCurrentSession()
+            viewModel.fetchContextUsage()
         }
     }
 
@@ -512,6 +517,20 @@ fun ChatScreen(
                 }
             }
 
+            // Context-window meter (used / full) — sits above the composer so it
+            // stays visible while the session is active, grouped with the model
+            // it belongs to without crowding the title or the control row.
+            ContextUsageChip(
+                usedTokens = state.usedContextTokens,
+                fullTokens = state.fullContextTokens,
+                onClick =
+                    if (state.contextBreakdown != null) {
+                        { showContextSheet = true }
+                    } else {
+                        null
+                    },
+            )
+
             ChatInputBar(
                 inputFieldValue = inputFieldValue,
                 onInputChange = { inputFieldValue = it },
@@ -610,5 +629,13 @@ fun ChatScreen(
                 onDismiss = { viewModel.closeModelPicker() },
             )
         }
+
+        if (showContextSheet && state.contextBreakdown != null) {
+            ContextDetailSheet(
+                breakdown = state.contextBreakdown!!,
+                fullTokens = state.fullContextTokens,
+                onDismiss = { showContextSheet = false },
+            )
+        }
     }
 }

app/src/main/java/com/m57/hermescontrol/ui/chat/ChatViewModel.kt

@@ -92,6 +92,14 @@ data class ChatUiState(
     val currentSessionModel: String? = null,
     // Reasoning effort level for the current session
     val reasoningLevel: String? = null,
+    // Context-window meter (issue #XXX): tokens currently used by the session
+    // prompt (numerator) and the active model's full context window (denominator).
+    // Both null until the first successful fetch.
+    val usedContextTokens: Long? = null,
+    val fullContextTokens: Long? = null,
+    // Detailed token breakdown for the context meter's detail sheet (null until
+    // the first successful session-detail fetch).
+    val contextBreakdown: ContextBreakdown? = null,
     // Attachment state
     val pendingAttachments: List<Attachment> = emptyList(),
     // Reaction animation — set when a reaction WS event arrives, auto-clears
@@ -142,6 +150,21 @@ data class SecretPromptUi(
     val sessionId: String?,
 )
 
+/**
+ * Token breakdown backing the context meter's detail sheet. All values are
+ * token counts sourced from `GET /api/sessions/{id}` (`input_tokens`,
+ * `output_tokens`, `cache_read_tokens`, `cache_write_tokens`, `reasoning_tokens`,
+ * `message_count`) — verified present on the live gateway's `sessions` table.
+ */
+data class ContextBreakdown(
+    val inputTokens: Long,
+    val outputTokens: Long,
+    val cacheReadTokens: Long,
+    val cacheWriteTokens: Long,
+    val reasoningTokens: Long,
+    val messageCount: Int,
+)
+
 class ChatViewModel(
     application: Application,
     private val startCleanup: Boolean,
@@ -386,6 +409,12 @@ class ChatViewModel(
                 is ReducerEffect.RefreshSessions -> {
                     loadSessions()
                 }
+
+                is ReducerEffect.RefreshContextUsage -> {
+                    // Streaming finished — refresh the context meter now rather
+                    // than waiting up to 5s for the next session-sync poll.
+                    viewModelScope.launch { fetchContextUsage() }
+                }
             }
         }
 
@@ -1349,6 +1378,8 @@ class ChatViewModel(
                 currentSessionModel = "$provider/$model",
             )
         }
+        // Model switch changes the context-window denominator — refetch it.
+        fetchContextUsage()
         handleSlashCommand("/model $model --provider $provider --session")
     }
 
@@ -1549,6 +1580,65 @@ class ChatViewModel(
         }
     }
 
+    /**
+     * Refresh the context-window meter for the current session.
+     *
+     * Numerator (`usedContextTokens`) comes from the session record's
+     * `last_prompt_tokens` (`/api/sessions/{id}`, backend
+     * `gateway/session.py`). Denominator (`fullContextTokens`) comes from the
+     * active model's `effective_context_length` (`/api/model/info`, PUBLIC).
+     *
+     * Both calls are independent and best-effort: a failure on one must not
+     * wipe the other's already-shown value, and neither blocks the chat. The
+     * two fetches are launched separately so a slow/erroring one can't starve
+     * the other. Polled from [syncCurrentSession] via the 5s loop and re-fired
+     * on model switch (the denominator changes).
+     */
+    fun fetchContextUsage() {
+        val sessionId = _uiState.value.currentSessionId ?: return
+        val profile = AuthManager.getSelectedProfileId()
+        viewModelScope.launch(Dispatchers.IO) {
+            // Denominator: full context window (cheap, public, rarely changes).
+            val fullResult =
+                safeApiCall { ApiClient.hermesApi.getModelInfo() }
+            if (fullResult is NetworkResult.Success) {
+                val full =
+                    fullResult.data.effective_context_length
+                        ?: fullResult.data.auto_context_length
+                        ?: fullResult.data.config_context_length
+                if (full != null && full > 0L) {
+                    _uiState.update { it.copy(fullContextTokens = full) }
+                }
+            }
+            // Numerator: used context for THIS session. The gateway's sessions
+            // table persists `input_tokens` (cumulative prompt tokens) — there is
+            // NO `last_prompt_tokens` column on the REST response, so we use
+            // `input_tokens` as the used-context numerator.
+            val usedResult =
+                safeApiCall { ApiClient.hermesApi.getSessionDetail(sessionId, profile) }
+            if (usedResult is NetworkResult.Success) {
+                val d = usedResult.data
+                val used = d.input_tokens
+                if (used != null) {
+                    _uiState.update {
+                        it.copy(
+                            usedContextTokens = used,
+                            contextBreakdown =
+                                ContextBreakdown(
+                                    inputTokens = used,
+                                    outputTokens = d.output_tokens ?: 0L,
+                                    cacheReadTokens = d.cache_read_tokens ?: 0L,
+                                    cacheWriteTokens = d.cache_write_tokens ?: 0L,
+                                    reasoningTokens = d.reasoning_tokens ?: 0L,
+                                    messageCount = d.message_count ?: 0,
+                                ),
+                        )
+                    }
+                }
+            }
+        }
+    }
+
     private suspend fun fetchServerMessageCount(sessionId: String): Int {
         val known =
             _uiState.value.sessions

app/src/main/java/com/m57/hermescontrol/ui/chat/ChatWsEventReducer.kt

@@ -252,6 +252,9 @@ object ChatWsEventReducer {
             effects.add(ReducerEffect.PersistMessage(msg, sid))
         }
         effects.add(ReducerEffect.RefreshSessions)
+        // Refresh the context meter immediately on completion instead of
+        // waiting for the next 5s session-sync poll.
+        effects.add(ReducerEffect.RefreshContextUsage)
         return ReducerResult(
             state =
                 state.copy(
@@ -664,4 +667,6 @@ sealed class ReducerEffect {
     data object LoadSessions : ReducerEffect()
 
     data object RefreshSessions : ReducerEffect()
+
+    data object RefreshContextUsage : ReducerEffect()
 }

app/src/main/java/com/m57/hermescontrol/ui/chat/components/ContextDetailSheet.kt

@@ -0,0 +1,162 @@
+package com.m57.hermescontrol.ui.chat.components
+
+import androidx.compose.foundation.background
+import androidx.compose.foundation.layout.Arrangement
+import androidx.compose.foundation.layout.Box
+import androidx.compose.foundation.layout.Column
+import androidx.compose.foundation.layout.Row
+import androidx.compose.foundation.layout.fillMaxWidth
+import androidx.compose.foundation.layout.height
+import androidx.compose.foundation.layout.padding
+import androidx.compose.foundation.rememberScrollState
+import androidx.compose.foundation.shape.RoundedCornerShape
+import androidx.compose.foundation.verticalScroll
+import androidx.compose.material3.ExperimentalMaterial3Api
+import androidx.compose.material3.MaterialTheme
+import androidx.compose.material3.ModalBottomSheet
+import androidx.compose.material3.SheetState
+import androidx.compose.material3.Text
+import androidx.compose.material3.rememberModalBottomSheetState
+import androidx.compose.runtime.Composable
+import androidx.compose.ui.Alignment
+import androidx.compose.ui.Modifier
+import androidx.compose.ui.text.font.FontWeight
+import androidx.compose.ui.unit.dp
+import com.m57.hermescontrol.theme.LocalHermesStatusColors
+import com.m57.hermescontrol.ui.chat.ContextBreakdown
+import kotlin.math.min
+
+/**
+ * Detail sheet shown when the context meter chip is tapped. Lists the session's
+ * token breakdown (input / output / cache read / cache write / reasoning) and
+ * the message count, plus a prominent used / full bar.
+ *
+ * `fullTokens` is the model context window (denominator); `breakdown` carries
+ * the per-category counts from `GET /api/sessions/{id}`.
+ */
+@OptIn(ExperimentalMaterial3Api::class)
+@Composable
+fun ContextDetailSheet(
+    breakdown: ContextBreakdown,
+    fullTokens: Long?,
+    onDismiss: () -> Unit,
+    sheetState: SheetState = rememberModalBottomSheetState(),
+) {
+    ModalBottomSheet(
+        onDismissRequest = onDismiss,
+        sheetState = sheetState,
+    ) {
+        val full = fullTokens ?: 0L
+        val used = breakdown.inputTokens
+        val fraction = if (full > 0L) min(1f, used.toFloat() / full.toFloat()) else 0f
+        val pct = (fraction * 100).toInt()
+        val statusColors = LocalHermesStatusColors.current
+        val barColor =
+            when {
+                pct >= 90 -> statusColors.error
+                pct >= 70 -> statusColors.warning
+                else -> MaterialTheme.colorScheme.primary
+            }
+
+        Column(
+            modifier =
+                Modifier
+                    .fillMaxWidth()
+                    .padding(horizontal = 16.dp, vertical = 8.dp)
+                    .verticalScroll(rememberScrollState()),
+            verticalArrangement = Arrangement.spacedBy(12.dp),
+        ) {
+            Text(
+                text = "Context window",
+                style = MaterialTheme.typography.titleMedium,
+                fontWeight = FontWeight.SemiBold,
+            )
+
+            // Big used / full header
+            Row(
+                modifier = Modifier.fillMaxWidth(),
+                horizontalArrangement = Arrangement.SpaceBetween,
+                verticalAlignment = Alignment.Bottom,
+            ) {
+                Text(
+                    text = formatTokens(used),
+                    style = MaterialTheme.typography.headlineSmall,
+                    fontWeight = FontWeight.Bold,
+                    color = barColor,
+                )
+                Text(
+                    text = "of ${formatTokens(full)}",
+                    style = MaterialTheme.typography.bodyMedium,
+                    color = MaterialTheme.colorScheme.onSurfaceVariant,
+                )
+            }
+
+            // Progress bar
+            Box(
+                modifier =
+                    Modifier
+                        .fillMaxWidth()
+                        .height(8.dp)
+                        .background(
+                            color = MaterialTheme.colorScheme.surfaceVariant,
+                            shape = RoundedCornerShape(4.dp),
+                        ),
+            ) {
+                Box(
+                    modifier =
+                        Modifier
+                            .fillMaxWidth(fraction)
+                            .height(8.dp)
+                            .background(color = barColor, shape = RoundedCornerShape(4.dp)),
+                )
+            }
+            Text(
+                text = "$pct% of context window used",
+                style = MaterialTheme.typography.labelMedium,
+                color = MaterialTheme.colorScheme.onSurfaceVariant,
+            )
+
+            // Breakdown rows
+            ContextRow(label = "Prompt (input)", value = breakdown.inputTokens)
+            ContextRow(label = "Completion (output)", value = breakdown.outputTokens)
+            ContextRow(label = "Cache read", value = breakdown.cacheReadTokens)
+            ContextRow(label = "Cache write", value = breakdown.cacheWriteTokens)
+            ContextRow(label = "Reasoning", value = breakdown.reasoningTokens)
+            ContextRow(label = "Messages", value = breakdown.messageCount.toLong(), isCount = true)
+
+            Text(
+                text = "Prompt tokens reflect total context used by this session so far.",
+                style = MaterialTheme.typography.labelSmall,
+                color = MaterialTheme.colorScheme.onSurfaceVariant,
+                modifier = Modifier.padding(bottom = 16.dp),
+            )
+        }
+    }
+}
+
+@Composable
+private fun ContextRow(
+    label: String,
+    value: Long,
+    isCount: Boolean = false,
+) {
+    Row(
+        modifier =
+            Modifier
+                .fillMaxWidth()
+                .padding(vertical = 4.dp),
+        horizontalArrangement = Arrangement.SpaceBetween,
+    ) {
+        Text(
+            text = label,
+            style = MaterialTheme.typography.bodyMedium,
+            color = MaterialTheme.colorScheme.onSurface,
+        )
+        Text(
+            text = if (isCount) value.toString() else formatTokens(value),
+            style = MaterialTheme.typography.bodyMedium,
+            fontWeight = FontWeight.Medium,
+            color = MaterialTheme.colorScheme.onSurfaceVariant,
+        )
+    }
+}

app/src/main/java/com/m57/hermescontrol/ui/chat/components/ContextUsageChip.kt

@@ -0,0 +1,147 @@
+package com.m57.hermescontrol.ui.chat.components
+
+import androidx.compose.foundation.background
+import androidx.compose.foundation.clickable
+import androidx.compose.foundation.layout.Arrangement
+import androidx.compose.foundation.layout.Box
+import androidx.compose.foundation.layout.Column
+import androidx.compose.foundation.layout.Row
+import androidx.compose.foundation.layout.fillMaxWidth
+import androidx.compose.foundation.layout.height
+import androidx.compose.foundation.layout.padding
+import androidx.compose.foundation.shape.RoundedCornerShape
+import androidx.compose.material3.MaterialTheme
+import androidx.compose.material3.Text
+import androidx.compose.runtime.Composable
+import androidx.compose.ui.Alignment
+import androidx.compose.ui.Modifier
+import androidx.compose.ui.text.style.TextOverflow
+import androidx.compose.ui.unit.dp
+import com.m57.hermescontrol.theme.LocalHermesStatusColors
+import kotlin.math.min
+
+/**
+ * Compact "used / full context" meter for the chat screen.
+ *
+ * Renders `<used> / <full> context` (e.g. `12.3k / 262k`) plus a color-coded
+ * progress bar. The bar turns amber past 70% and red past 90% of the context
+ * window so the user sees compression pressure building before a turn silently
+ * compacts.
+ *
+ * The denominator (`fullTokens`) comes from `GET /api/model/info` (PUBLIC) and
+ * is always available. The numerator (`usedTokens`) comes from
+ * `GET /api/sessions/{id}` (`input_tokens`) — gated, so it may lag or be absent
+ * on first render. The chip shows as soon as `fullTokens` is known; when
+ * `usedTokens` is still null it renders `— / <full>` (unknown used) rather than
+ * hiding entirely, so a missing gated fetch never blanks the meter.
+ *
+ * @param usedTokens tokens currently in the session prompt (numerator), or null
+ * @param fullTokens the active model's full context window (denominator)
+ */
+@Composable
+fun ContextUsageChip(
+    usedTokens: Long?,
+    fullTokens: Long?,
+    onClick: (() -> Unit)? = null,
+    modifier: Modifier = Modifier,
+) {
+    val full = fullTokens
+    if (full == null || full <= 0L) return
+
+    val used = usedTokens ?: 0L
+    val fraction = min(1f, used.toFloat() / full.toFloat())
+    val pct = (fraction * 100).toInt()
+    val statusColors = LocalHermesStatusColors.current
+    val barColor =
+        when {
+            pct >= 90 -> statusColors.error
+            pct >= 70 -> statusColors.warning
+            else -> MaterialTheme.colorScheme.primary
+        }
+
+    Column(
+        modifier =
+            modifier
+                .fillMaxWidth()
+                .then(
+                    if (onClick != null) {
+                        Modifier
+                            .clickable(onClick = onClick)
+                            .padding(horizontal = 12.dp, vertical = 4.dp)
+                    } else {
+                        Modifier.padding(horizontal = 12.dp, vertical = 4.dp)
+                    },
+                ),
+        verticalArrangement = Arrangement.spacedBy(3.dp),
+    ) {
+        Row(
+            modifier = Modifier.fillMaxWidth(),
+            horizontalArrangement = Arrangement.SpaceBetween,
+            verticalAlignment = Alignment.CenterVertically,
+        ) {
+            Text(
+                text =
+                    if (usedTokens != null) {
+                        "${formatTokens(used)} / ${formatTokens(full)} context"
+                    } else {
+                        "— / ${formatTokens(full)} context"
+                    },
+                style = MaterialTheme.typography.labelSmall,
+                color = MaterialTheme.colorScheme.onSurfaceVariant,
+                maxLines = 1,
+                overflow = TextOverflow.Ellipsis,
+            )
+            Text(
+                text = "$pct%",
+                style = MaterialTheme.typography.labelSmall,
+                color = barColor,
+                maxLines = 1,
+            )
+        }
+        Box(
+            modifier =
+                Modifier
+                    .fillMaxWidth()
+                    .height(3.dp)
+                    .background(
+                        color = MaterialTheme.colorScheme.surfaceVariant,
+                        shape = RoundedCornerShape(2.dp),
+                    ),
+        ) {
+            Box(
+                modifier =
+                    Modifier
+                        .fillMaxWidth(fraction)
+                        .height(3.dp)
+                        .background(
+                            color = barColor,
+                            shape = RoundedCornerShape(2.dp),
+                        ),
+            )
+        }
+    }
+}
+
+/**
+ * Compact token formatting: 1_500 → "1.5k", 262_144 → "262k", 950 → "950".
+ * Keeps the chip narrow so it fits above the composer on small screens.
+ */
+internal fun formatTokens(tokens: Long): String =
+    when {
+        tokens >= 1_000_000 -> {
+            "${tokens / 1_000_000}M"
+        }
+
+        tokens >= 100_000 -> {
+            "${tokens / 1000}k"
+        }
+
+        tokens >= 1_000 -> {
+            val k = tokens / 1000.0
+            if (k % 1.0 == 0.0) "${k.toInt()}k" else String.format(java.util.Locale.US, "%.1fk", k)
+        }
+
+        else -> {
+            tokens.toString()
+        }
+    }

app/src/main/java/com/m57/hermescontrol/ui/connect/ConnectScreen.kt

@@ -35,6 +35,7 @@ import androidx.compose.material3.IconButton
 import androidx.compose.material3.MaterialTheme
 import androidx.compose.material3.OutlinedButton
 import androidx.compose.material3.OutlinedTextField
+import androidx.compose.material3.Surface
 import androidx.compose.material3.Text
 import androidx.compose.material3.TextButton
 import androidx.compose.runtime.Composable
@@ -316,6 +317,43 @@ fun ConnectScreen(
                         )
                     }
                 }
+
+                // Backend health read-out (issue #713): cheap /api/health probe.
+                // Shows version + auth mode when reachable; nothing when null.
+                state.health?.let { health ->
+                    val authMode =
+                        if (health.authRequired == true) {
+                            stringResource(R.string.connect_health_auth_gated)
+                        } else {
+                            stringResource(R.string.connect_health_auth_token)
+                        }
+                    Surface(
+                        color = MaterialTheme.colorScheme.surfaceContainer,
+                        shape = MaterialTheme.shapes.medium,
+                        modifier = Modifier.fillMaxWidth(),
+                    ) {
+                        Column(
+                            modifier = Modifier.padding(12.dp),
+                            verticalArrangement = Arrangement.spacedBy(4.dp),
+                        ) {
+                            Text(
+                                text = stringResource(R.string.connect_health_title),
+                                style = MaterialTheme.typography.labelMedium,
+                                color = MaterialTheme.colorScheme.onSurfaceVariant,
+                            )
+                            Text(
+                                text =
+                                    stringResource(
+                                        R.string.connect_health_body,
+                                        health.version ?: "?",
+                                        authMode,
+                                    ),
+                                style = MaterialTheme.typography.bodyMedium,
+                                color = MaterialTheme.colorScheme.onSurface,
+                            )
+                        }
+                    }
+                }
             }
         }
     }

app/src/main/java/com/m57/hermescontrol/ui/connect/ConnectViewModel.kt

@@ -8,6 +8,7 @@ import com.m57.hermescontrol.R
 import com.m57.hermescontrol.data.config.ConnectionProfile
 import com.m57.hermescontrol.data.config.resolveBaseUrl
 import com.m57.hermescontrol.data.local.AuthManager
+import com.m57.hermescontrol.data.model.HealthStatus
 import com.m57.hermescontrol.data.remote.ApiClient
 import com.m57.hermescontrol.data.remote.CleartextPolicy
 import com.m57.hermescontrol.data.remote.NetworkError
@@ -33,6 +34,7 @@ data class ConnectUiState(
     val saveProfile: Boolean = false,
     val profiles: List<ConnectionProfile> = emptyList(),
     val selectedProfile: ConnectionProfile? = null,
+    val health: HealthStatus? = null,
 )
 
 class ConnectViewModel(
@@ -60,6 +62,29 @@ class ConnectViewModel(
                 profileName = selectedProfile?.name ?: "",
             )
         }
+        loadHealth()
+    }
+
+    /**
+     * Cheap liveness probe (issue #713): GET /api/health against the current
+     * URL/token. Surfaces backend version + auth mode on the connection screen
+     * without the heavier /api/status payload. Best-effort: failures are silent
+     * (health stays null and the screen simply shows no read-out).
+     */
+    fun loadHealth() {
+        val state = _uiState.value
+        val endpoint = runCatching { ServerEndpoint.parseForBuild(state.baseUrl) }.getOrNull()
+        if (endpoint == null || state.token.isBlank()) return
+        viewModelScope.launch {
+            val result =
+                withContext(Dispatchers.IO) {
+                    val tempApi = ApiClient.createTempService(endpoint.baseUrl.toString(), state.token)
+                    safeApiCall { tempApi.getHealth() }
+                }
+            if (result is NetworkResult.Success) {
+                _uiState.update { it.copy(health = result.data) }
+            }
+        }
     }
 
     fun onProfileNameChange(value: String) {

app/src/main/res/values-ko/strings.xml

@@ -179,6 +179,10 @@
     <!-- Connect Screen -->
     <string name="connect_selected_profile">프로필: %1$s</string>
     <string name="connect_action_select_profile">저장된 프로필 선택</string>
+    <string name="connect_health_title">백엔드 상태</string>
+    <string name="connect_health_body">Hermes %1$s · %2$s</string>
+    <string name="connect_health_auth_gated">게이트(OAuth)</string>
+    <string name="connect_health_auth_token">토큰</string>
 
     <!-- Connect Error Messages -->
     <string name="connect_error_token_required">토큰이 필요합니다</string>

app/src/main/res/values/strings.xml

@@ -207,6 +207,10 @@
     <!-- Connect Screen -->
     <string name="connect_selected_profile">Profile: %1$s</string>
     <string name="connect_action_select_profile">Select Saved Profile</string>
+    <string name="connect_health_title">Backend health</string>
+    <string name="connect_health_body">Hermes %1$s · %2$s</string>
+    <string name="connect_health_auth_gated">Gated (OAuth)</string>
+    <string name="connect_health_auth_token">Token</string>
 
     <!-- Connect Error Messages -->
     <string name="connect_error_token_required">Token is required</string>

app/src/test/java/com/m57/hermescontrol/ui/chat/ChatViewModelTest.kt

@@ -83,6 +83,7 @@ class ChatViewModelTest {
         mockConnectionStatus.value = ConnectionStatus.DISCONNECTED
 
         every { AuthManager.getToken() } returns "test-token"
+        every { AuthManager.getSelectedProfileId() } returns null
         every { AuthManager.isTypingEffectEnabled() } returns true
         every { AuthManager.getTypingEffectDelayMs() } returns 30
         every { AuthManager.isAutoReconnect() } returns false
@@ -1742,7 +1743,10 @@ class ChatViewModelTest {
                 "hasOlderMessages must be false when initial page is empty",
                 viewModel.uiState.value.hasOlderMessages,
             )
-            assertTrue(viewModel.uiState.value.messages.isEmpty())
+            assertTrue(
+                viewModel.uiState.value.messages
+                    .isEmpty(),
+            )
             assertFalse(viewModel.uiState.value.isLoading)
         }
 

app/src/test/java/com/m57/hermescontrol/ui/chat/ContextUsageChipTest.kt

@@ -0,0 +1,38 @@
+package com.m57.hermescontrol.ui.chat.components
+
+import org.junit.Assert.assertEquals
+import org.junit.Test
+
+/**
+ * Unit tests for the token-compaction math in [ContextUsageChip] — the only
+ * non-trivial pure logic in the context meter. Formats must stay compact so
+ * the chip fits above the composer on a narrow phone screen.
+ */
+class ContextUsageChipTest {
+    @Test
+    fun formatTokens_subThousand_returnsRawCount() {
+        assertEquals("950", formatTokens(950))
+        assertEquals("0", formatTokens(0))
+    }
+
+    @Test
+    fun formatTokens_thousands_dropsTrailingPointZero() {
+        // 262_144 -> "262k" (no ".0"), not "262.1k".
+        assertEquals("262k", formatTokens(262_144))
+        // 12_300 -> "12.3k" (one decimal preserved).
+        assertEquals("12.3k", formatTokens(12_300))
+    }
+
+    @Test
+    fun formatTokens_millions_usesMSuffix() {
+        assertEquals("1M", formatTokens(1_000_000))
+        assertEquals("2M", formatTokens(2_048_000))
+    }
+
+    @Test
+    fun formatTokens_exactThousandBoundary_roundsDownCleanly() {
+        // 1_000 -> "1k" not "1.0k".
+        assertEquals("1k", formatTokens(1_000))
+        assertEquals("100k", formatTokens(100_000))
+    }
+}