From 4fe2fecf54e376641b7bc157cd4b4613debfa08f Mon Sep 17 00:00:00 2001
From: CharlesMcquade <6466275+CharlesMcquade@users.noreply.github.com>
Date: Tue, 14 Jul 2026 22:47:05 -0500
Subject: [PATCH] feat(gateway): add /context command for a detailed
 context-window view
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

A dedicated /context (alias /ctx) gateway slash command that gives a full
context-window view with:

- Usage gauge: visual bar + fraction + percentage + headroom
- Auto-compression threshold and how far away it is
- Compression count and how much the last one freed
- Cumulative session throughput (explicitly labelled as throughput,
  NOT context size — each call re-sends the window)
- Cascading fallback: running agent → cached agent → SessionStore metadata
  → rough transcript estimate

Not included (per current-main design):
- Cache reporting removed: commit 446b8e239 intentionally removed cache
  reporting from user-facing surfaces because providers that omit cached-token
  details produce misleading values
- Sync DB calls replaced with async_session_store (current main requires
  AsyncSessionStore with await)

Also rewords the /status tokens line from 'Cumulative API tokens (re-sent
each call)' to 'Lifetime tokens billed: ... (not your current context size;
use /context)' to reduce the recurring confusion that the cumulative figure
is the current context window.

Fixes salvation of PR #52184 (salvage commit replaces a 12K-commit-behind
fork branch with a fresh implementation against current main, incorporating
reviewer feedback from @whoislikemiha and the hermes-sweeper).
---
 gateway/run.py                       |   6 +
 gateway/slash_commands.py            | 187 +++++++++++++++++++++++++++
 hermes_cli/commands.py               |   2 +
 locales/af.yaml                      |  20 ++-
 locales/de.yaml                      |  20 ++-
 locales/en.yaml                      |  21 ++-
 locales/es.yaml                      |  20 ++-
 locales/fr.yaml                      |  20 ++-
 locales/ga.yaml                      |  20 ++-
 locales/hu.yaml                      |  20 ++-
 locales/it.yaml                      |  20 ++-
 locales/ja.yaml                      |  20 ++-
 locales/ko.yaml                      |  20 ++-
 locales/pt.yaml                      |  20 ++-
 locales/ru.yaml                      |  20 ++-
 locales/tr.yaml                      |  20 ++-
 locales/uk.yaml                      |  20 ++-
 locales/zh-hant.yaml                 |  20 ++-
 locales/zh.yaml                      |  20 ++-
 tests/gateway/test_status_command.py | 160 ++++++++++++++++++++++-
 20 files changed, 654 insertions(+), 22 deletions(-)

diff --git a/gateway/run.py b/gateway/run.py
index dbf5eee34d96..a8dcca69554d 100644
--- a/gateway/run.py
+++ b/gateway/run.py
@@ -11168,6 +11168,9 @@ class GatewayRunner(GatewayAuthorizationMixin, GatewayKanbanWatchersMixin, Gatew
             if event.get_command() == "status":
                 return await self._handle_status_command(event)
 
+            if event.get_command() in {"context", "ctx"}:
+                return await self._handle_context_command(event)
+
             # Resolve the command once for all early-intercept checks below.
             from hermes_cli.commands import (
                 ACTIVE_SESSION_BYPASS_COMMANDS as _DEDICATED_HANDLERS,
@@ -11743,6 +11746,9 @@ class GatewayRunner(GatewayAuthorizationMixin, GatewayKanbanWatchersMixin, Gatew
 
             return format_status_text()
 
+        if canonical == "context":
+            return await self._handle_context_command(event)
+
         if canonical == "agents":
             return await self._handle_agents_command(event)
 
diff --git a/gateway/slash_commands.py b/gateway/slash_commands.py
index 7c7035abf2d1..047e1003563e 100644
--- a/gateway/slash_commands.py
+++ b/gateway/slash_commands.py
@@ -56,6 +56,19 @@ logger = logging.getLogger("gateway.run")
 _RESET_CLEANUP_TIMEOUT_S = 30.0
 
 
+def _clean_str(value: Any) -> str:
+    """Strip and return a non-empty string value, or empty string."""
+    return value.strip() if isinstance(value, str) and value.strip() else ""
+
+
+def _int_value(value: Any) -> int:
+    """Safely coerce to int."""
+    try:
+        return int(value)
+    except (TypeError, ValueError):
+        return 0
+
+
 def _model_switch_skew_guard() -> Optional[str]:
     """Refuse a model switch when the gateway is running stale code.
 
@@ -701,6 +714,180 @@ class GatewaySlashCommandsMixin:
         digest = hashlib.sha256(text.encode("utf-8")).hexdigest()[:12]
         return f"sha256:{digest}"
 
+    async def _handle_context_command(self, event: MessageEvent) -> str:
+        """Handle /context — the dedicated context-window view.
+
+        /status shows a one-line ``used / total`` summary; this command is the
+        deep view: a usage gauge, auto-compression threshold and headroom,
+        compression count and last savings, and cumulative throughput — the last
+        clearly labelled as throughput, NOT context size.
+
+        Resolves from the running agent (mid-turn), then the cached agent
+        (between turns), then the SessionStore/SessionDB metadata for a gauge
+        even when no agent is resident. Falls back to a transcript estimate only
+        as a last resort.
+        """
+        from gateway.run import _AGENT_PENDING_SENTINEL
+
+        source = event.source
+        session_key = self._session_key_for_source(source)
+        session_entry = await self.async_session_store.get_or_create_session(source)
+
+        # Try running agent first (mid-turn), then cached agent (between turns).
+        agent = self._running_agents.get(session_key)
+        if not agent or agent is _AGENT_PENDING_SENTINEL:
+            cache_lock = getattr(self, "_agent_cache_lock", None)
+            cache = getattr(self, "_agent_cache", None)
+            if cache_lock is not None and cache is not None:
+                try:
+                    with cache_lock:
+                        cached = cache.get(session_key)
+                    if cached:
+                        agent = cached[0]
+                except Exception:
+                    agent = None
+        has_agent = bool(agent) and agent is not _AGENT_PENDING_SENTINEL
+
+        ctx = getattr(agent, "context_compressor", None) if has_agent else None
+
+        # Resolve current-context size + window with cascading fallbacks.
+        #   used  : compressor.last_prompt_tokens → SessionStore.last_prompt_tokens
+        #   model : agent.model → SessionDB row model
+        #   window: compressor.context_length → get_model_context_length(model)
+        used = 0
+        context_length = 0
+        if ctx is not None:
+            used = getattr(ctx, "last_prompt_tokens", 0) or 0
+            context_length = getattr(ctx, "context_length", 0) or 0
+
+        model_name = _clean_str(getattr(agent, "model", "")) if has_agent else ""
+
+        if not used:
+            used = _int_value(getattr(session_entry, "last_prompt_tokens", 0))
+
+        if not model_name and self._session_db:
+            try:
+                row = await self._session_db.get_session(session_entry.session_id) or {}
+                if isinstance(row, dict):
+                    model_name = _clean_str(row.get("model", ""))
+            except Exception:
+                model_name = ""
+
+        if not context_length and model_name:
+            try:
+                from agent.model_metadata import get_model_context_length
+
+                context_length = _int_value(
+                    await asyncio.to_thread(get_model_context_length, model_name)
+                )
+            except Exception:
+                context_length = 0
+
+        # Gauge path: real current-context figure
+        if used > 0 and context_length > 0:
+            pct = min(100.0, used / context_length * 100)
+            headroom = max(0, context_length - used)
+            BAR_WIDTH = 24
+            filled = int(round(pct / 100 * BAR_WIDTH))
+            bar = "█" * max(0, filled) + "░" * max(0, BAR_WIDTH - filled)
+
+            lines = [
+                t("gateway.context.header"),
+                "",
+                t("gateway.context.model", model=model_name or "?"),
+                t("gateway.context.window", total=f"{context_length:,}"),
+                t(
+                    "gateway.context.in_use",
+                    used=f"{used:,}",
+                    total=f"{context_length:,}",
+                    pct=f"{pct:.0f}",
+                ),
+                t("gateway.context.bar", bar=bar),
+                t("gateway.context.headroom", headroom=f"{headroom:,}"),
+            ]
+
+            # Full view — compression / throughput need the live agent.
+            if ctx is not None:
+                threshold = getattr(ctx, "threshold_tokens", 0) or 0
+                threshold_pct = (getattr(ctx, "threshold_percent", 0) or 0) * 100
+                lines.append("")
+                if threshold > 0:
+                    if used >= threshold:
+                        lines.append(
+                            t(
+                                "gateway.context.over_threshold",
+                                threshold=f"{threshold:,}",
+                                threshold_pct=f"{threshold_pct:.0f}",
+                            )
+                        )
+                    else:
+                        lines.append(
+                            t(
+                                "gateway.context.threshold",
+                                threshold=f"{threshold:,}",
+                                threshold_pct=f"{threshold_pct:.0f}",
+                                to_go=f"{threshold - used:,}",
+                            )
+                        )
+                compressions = getattr(ctx, "compression_count", 0) or 0
+                lines.append(t("gateway.context.compressions", count=compressions))
+                if compressions:
+                    savings = getattr(ctx, "_last_compression_savings_pct", None)
+                    if savings is not None:
+                        lines.append(
+                            t("gateway.context.last_savings", savings=f"{savings:.0f}")
+                        )
+
+                api_calls = getattr(agent, "session_api_calls", 0) or 0
+                input_tokens = getattr(agent, "session_input_tokens", 0) or 0
+                output_tokens = getattr(agent, "session_output_tokens", 0) or 0
+                reasoning_tokens = getattr(agent, "session_reasoning_tokens", 0) or 0
+                total_tokens = getattr(agent, "session_total_tokens", 0) or 0
+                lines.append("")
+                lines.append(
+                    t("gateway.context.totals_header", calls=api_calls)
+                )
+                lines.append(
+                    t(
+                        "gateway.context.totals_line",
+                        input=f"{input_tokens:,}",
+                        output=f"{output_tokens:,}",
+                        reasoning=f"{reasoning_tokens:,}",
+                    )
+                )
+                lines.append(t("gateway.context.total_billed", total=f"{total_tokens:,}"))
+                lines.append(t("gateway.context.throughput_note"))
+            else:
+                lines.append("")
+                lines.append(t("gateway.context.detail_after_first"))
+
+            return "\n".join(lines)
+
+        # Last resort: rough estimate from transcript
+        history = await self.async_session_store.load_transcript(session_entry.session_id)
+        if history:
+            from agent.model_metadata import estimate_messages_tokens_rough
+
+            msgs = [
+                m
+                for m in history
+                if m.get("role") in {"user", "assistant"} and m.get("content")
+            ]
+            approx = estimate_messages_tokens_rough(msgs)
+            return "\n".join(
+                [
+                    t("gateway.context.header"),
+                    "",
+                    t(
+                        "gateway.context.estimated",
+                        count=f"{approx:,}",
+                        messages=len(msgs),
+                    ),
+                    t("gateway.context.detail_after_first"),
+                ]
+            )
+        return t("gateway.context.no_data")
+
     def _gateway_session_origin_for_id(self, session_id: str) -> Optional[SessionSource]:
         """Best-effort origin lookup for gateway session IDs."""
         lookup = getattr(type(self.session_store), "lookup_by_session_id", None)
diff --git a/hermes_cli/commands.py b/hermes_cli/commands.py
index 87c20087013a..c1c6f37b3893 100644
--- a/hermes_cli/commands.py
+++ b/hermes_cli/commands.py
@@ -120,6 +120,8 @@ COMMAND_REGISTRY: list[CommandDef] = [
     CommandDef("status", "Show session, model, token, and context info", "Session"),
     CommandDef("egress", "Show Docker egress proxy status", "Session",
                args_hint="[status]", subcommands=("status",)),
+    CommandDef("context", "Show detailed context window view with usage gauge, compression stats, and throughput", "Session",
+               aliases=("ctx",)),
     CommandDef("whoami", "Show your slash command access (admin / user)", "Info"),
     CommandDef("profile", "Show active profile name and home directory", "Info"),
     CommandDef("sethome", "Set this chat as the home channel", "Session",
diff --git a/locales/af.yaml b/locales/af.yaml
index 4069477de2aa..525d30db5b74 100644
--- a/locales/af.yaml
+++ b/locales/af.yaml
@@ -282,6 +282,24 @@ Future messages in this room will use that transcript until `/reset` or another
   set_home:
     save_failed:           "Kon nie tuiste-kanaal stoor nie: {error}"
     success:               "✅ Tuiste-kanaal gestel op **{name}** (ID: {chat_id}).\nKron-take en kruisplatform-boodskappe sal hier afgelewer word."
+  context:
+    header:                "🧠 **Context Window**"
+    model:                 "Model: `{model}`"
+    window:                "Window: {total} tokens"
+    in_use:                "In use: {used} / {total} ({pct}%)"
+    bar:                   "{bar}"
+    headroom:              "Headroom to limit: {headroom} tokens"
+    threshold:             "Auto-compresses at: {threshold} ({threshold_pct}%) — {to_go} to go"
+    over_threshold:        "⚠️ **Over auto-compression threshold ({threshold}, {threshold_pct}%)**"
+    compressions:          "Compressions this session: {count}"
+    last_savings:          "Last compression freed: {savings}% of context"
+    totals_header:         "Session totals (cumulative across {calls} API calls)"
+    totals_line:           "Input {input} · Output {output} · Reasoning {reasoning}"
+    total_billed:          "Total billed: {total}"
+    throughput_note:       "_Throughput, not context size — each call re-sends the window above._"
+    estimated:             "Estimated context: ~{count} tokens across {messages} messages"
+    detail_after_first:    "_(Full compression and throughput stats available after the first agent response)_"
+    no_data:               "No context data available yet. Send a message to start a session."
 
   status:
     header:                "📊 **Hermes Gateway Status**"
@@ -299,7 +317,7 @@ Future messages in this room will use that transcript until `/reset` or another
     model_provider:        "**Model:** `{model}` ({provider})"
     context:               "**Context:** {used} / {total} ({pct}%)"
     context_used:          "**Context:** ~{used} tokens"
-    tokens:                "**Kumulatiewe API-tokens (elke oproep weer gestuur):** {tokens}"
+    tokens:                "**Lifetime tokens billed:** {tokens} _(not your current context size; use `/context`)_"
     agent_running:         "**Agent loop:** {state}"
     state_yes:             "Ja ⚡"
     state_no:              "Nee"
diff --git a/locales/de.yaml b/locales/de.yaml
index 06a6f1931cbb..693b72d5c109 100644
--- a/locales/de.yaml
+++ b/locales/de.yaml
@@ -282,6 +282,24 @@ Future messages in this room will use that transcript until `/reset` or another
   set_home:
     save_failed:           "Home-Kanal konnte nicht gespeichert werden: {error}"
     success:               "✅ Home-Kanal auf **{name}** (ID: {chat_id}) gesetzt.\nCron-Jobs und plattformübergreifende Nachrichten werden hierher geliefert."
+  context:
+    header:                "🧠 **Context Window**"
+    model:                 "Model: `{model}`"
+    window:                "Window: {total} tokens"
+    in_use:                "In use: {used} / {total} ({pct}%)"
+    bar:                   "{bar}"
+    headroom:              "Headroom to limit: {headroom} tokens"
+    threshold:             "Auto-compresses at: {threshold} ({threshold_pct}%) — {to_go} to go"
+    over_threshold:        "⚠️ **Over auto-compression threshold ({threshold}, {threshold_pct}%)**"
+    compressions:          "Compressions this session: {count}"
+    last_savings:          "Last compression freed: {savings}% of context"
+    totals_header:         "Session totals (cumulative across {calls} API calls)"
+    totals_line:           "Input {input} · Output {output} · Reasoning {reasoning}"
+    total_billed:          "Total billed: {total}"
+    throughput_note:       "_Throughput, not context size — each call re-sends the window above._"
+    estimated:             "Estimated context: ~{count} tokens across {messages} messages"
+    detail_after_first:    "_(Full compression and throughput stats available after the first agent response)_"
+    no_data:               "No context data available yet. Send a message to start a session."
 
   status:
     header:                "📊 **Hermes-Gateway-Status**"
@@ -299,7 +317,7 @@ Future messages in this room will use that transcript until `/reset` or another
     model_provider:        "**Modell:** `{model}` ({provider})"
     context:               "**Kontext:** {used} / {total} ({pct}%)"
     context_used:          "**Kontext:** ~{used} Tokens"
-    tokens:                "**Kumulierte API-Tokens (bei jedem Aufruf erneut gesendet):** {tokens}"
+    tokens:                "**Lifetime tokens billed:** {tokens} _(not your current context size; use `/context`)_"
     agent_running:         "**Agent läuft:** {state}"
     state_yes:             "Ja ⚡"
     state_no:              "Nein"
diff --git a/locales/en.yaml b/locales/en.yaml
index d027257c0bdb..0ab28192c341 100644
--- a/locales/en.yaml
+++ b/locales/en.yaml
@@ -295,6 +295,25 @@ gateway:
     save_failed:           "Failed to save home channel: {error}"
     success:               "✅ Home channel set to **{name}** (ID: {chat_id}).\nCron jobs and cross-platform messages will be delivered here."
 
+  context:
+    header:                "🧠 **Context Window**"
+    model:                 "Model: `{model}`"
+    window:                "Window: {total} tokens"
+    in_use:                "In use: {used} / {total} ({pct}%)"
+    bar:                   "{bar}"
+    headroom:              "Headroom to limit: {headroom} tokens"
+    threshold:             "Auto-compresses at: {threshold} ({threshold_pct}%) — {to_go} to go"
+    over_threshold:        "⚠️ **Over auto-compression threshold ({threshold}, {threshold_pct}%)**"
+    compressions:          "Compressions this session: {count}"
+    last_savings:          "Last compression freed: {savings}% of context"
+    totals_header:         "Session totals (cumulative across {calls} API calls)"
+    totals_line:           "Input {input} · Output {output} · Reasoning {reasoning}"
+    total_billed:          "Total billed: {total}"
+    throughput_note:       "_Throughput, not context size — each call re-sends the window above._"
+    estimated:             "Estimated context: ~{count} tokens across {messages} messages"
+    detail_after_first:    "_(Full compression and throughput stats available after the first agent response)_"
+    no_data:               "No context data available yet. Send a message to start a session."
+
   status:
     header:                "📊 **Hermes Gateway Status**"
     matrix_scope_header:   "**Matrix scope:**"
@@ -311,7 +330,7 @@ gateway:
     model_provider:        "**Model:** `{model}` ({provider})"
     context:               "**Context:** {used} / {total} ({pct}%)"
     context_used:          "**Context:** ~{used} tokens"
-    tokens:                "**Cumulative API tokens (re-sent each call):** {tokens}"
+    tokens:                "**Lifetime tokens billed:** {tokens} _(not your current context size; use `/context`)_"
     agent_running:         "**Agent Running:** {state}"
     state_yes:             "Yes ⚡"
     state_no:              "No"
diff --git a/locales/es.yaml b/locales/es.yaml
index 074cfa9f278c..f69e94cdfaca 100644
--- a/locales/es.yaml
+++ b/locales/es.yaml
@@ -279,6 +279,24 @@ gateway:
   set_home:
     save_failed:           "No se pudo guardar el canal principal: {error}"
     success:               "✅ Canal principal establecido en **{name}** (ID: {chat_id}).\nLas tareas cron y los mensajes entre plataformas se entregarán aquí."
+  context:
+    header:                "🧠 **Context Window**"
+    model:                 "Model: `{model}`"
+    window:                "Window: {total} tokens"
+    in_use:                "In use: {used} / {total} ({pct}%)"
+    bar:                   "{bar}"
+    headroom:              "Headroom to limit: {headroom} tokens"
+    threshold:             "Auto-compresses at: {threshold} ({threshold_pct}%) — {to_go} to go"
+    over_threshold:        "⚠️ **Over auto-compression threshold ({threshold}, {threshold_pct}%)**"
+    compressions:          "Compressions this session: {count}"
+    last_savings:          "Last compression freed: {savings}% of context"
+    totals_header:         "Session totals (cumulative across {calls} API calls)"
+    totals_line:           "Input {input} · Output {output} · Reasoning {reasoning}"
+    total_billed:          "Total billed: {total}"
+    throughput_note:       "_Throughput, not context size — each call re-sends the window above._"
+    estimated:             "Estimated context: ~{count} tokens across {messages} messages"
+    detail_after_first:    "_(Full compression and throughput stats available after the first agent response)_"
+    no_data:               "No context data available yet. Send a message to start a session."
 
   status:
     header:                "📊 **Estado de Hermes Gateway**"
@@ -296,7 +314,7 @@ gateway:
     model_provider:        "**Model:** `{model}` ({provider})"
     context:               "**Context:** {used} / {total} ({pct}%)"
     context_used:          "**Context:** ~{used} tokens"
-    tokens:                "**Tokens de API acumulados (reenviados en cada llamada):** {tokens}"
+    tokens:                "**Lifetime tokens billed:** {tokens} _(not your current context size; use `/context`)_"
     agent_running:         "**Agente activo:** {state}"
     state_yes:             "Sí ⚡"
     state_no:              "No"
diff --git a/locales/fr.yaml b/locales/fr.yaml
index 166f1fba8741..a4248af6f7ce 100644
--- a/locales/fr.yaml
+++ b/locales/fr.yaml
@@ -282,6 +282,24 @@ Future messages in this room will use that transcript until `/reset` or another
   set_home:
     save_failed:           "Impossible d'enregistrer le canal principal : {error}"
     success:               "✅ Canal principal défini sur **{name}** (ID : {chat_id}).\nLes tâches cron et les messages multi-plateformes seront livrés ici."
+  context:
+    header:                "🧠 **Context Window**"
+    model:                 "Model: `{model}`"
+    window:                "Window: {total} tokens"
+    in_use:                "In use: {used} / {total} ({pct}%)"
+    bar:                   "{bar}"
+    headroom:              "Headroom to limit: {headroom} tokens"
+    threshold:             "Auto-compresses at: {threshold} ({threshold_pct}%) — {to_go} to go"
+    over_threshold:        "⚠️ **Over auto-compression threshold ({threshold}, {threshold_pct}%)**"
+    compressions:          "Compressions this session: {count}"
+    last_savings:          "Last compression freed: {savings}% of context"
+    totals_header:         "Session totals (cumulative across {calls} API calls)"
+    totals_line:           "Input {input} · Output {output} · Reasoning {reasoning}"
+    total_billed:          "Total billed: {total}"
+    throughput_note:       "_Throughput, not context size — each call re-sends the window above._"
+    estimated:             "Estimated context: ~{count} tokens across {messages} messages"
+    detail_after_first:    "_(Full compression and throughput stats available after the first agent response)_"
+    no_data:               "No context data available yet. Send a message to start a session."
 
   status:
     header:                "📊 **État de Hermes Gateway**"
@@ -299,7 +317,7 @@ Future messages in this room will use that transcript until `/reset` or another
     model_provider:        "**Model:** `{model}` ({provider})"
     context:               "**Context:** {used} / {total} ({pct}%)"
     context_used:          "**Context:** ~{used} tokens"
-    tokens:                "**Jetons :** {tokens}"
+    tokens:                "**Lifetime tokens billed:** {tokens} _(not your current context size; use `/context`)_"
     agent_running:         "**Agent en cours :** {state}"
     state_yes:             "Oui ⚡"
     state_no:              "Non"
diff --git a/locales/ga.yaml b/locales/ga.yaml
index a1522e4215c0..ccc0a6287544 100644
--- a/locales/ga.yaml
+++ b/locales/ga.yaml
@@ -286,6 +286,24 @@ Future messages in this room will use that transcript until `/reset` or another
   set_home:
     save_failed:           "Theip ar shábháil chainéil bhaile: {error}"
     success:               "✅ Cainéal baile socraithe go **{name}** (ID: {chat_id}).\nSeachadfar tascanna cron agus teachtaireachtaí trasardáin anseo."
+  context:
+    header:                "🧠 **Context Window**"
+    model:                 "Model: `{model}`"
+    window:                "Window: {total} tokens"
+    in_use:                "In use: {used} / {total} ({pct}%)"
+    bar:                   "{bar}"
+    headroom:              "Headroom to limit: {headroom} tokens"
+    threshold:             "Auto-compresses at: {threshold} ({threshold_pct}%) — {to_go} to go"
+    over_threshold:        "⚠️ **Over auto-compression threshold ({threshold}, {threshold_pct}%)**"
+    compressions:          "Compressions this session: {count}"
+    last_savings:          "Last compression freed: {savings}% of context"
+    totals_header:         "Session totals (cumulative across {calls} API calls)"
+    totals_line:           "Input {input} · Output {output} · Reasoning {reasoning}"
+    total_billed:          "Total billed: {total}"
+    throughput_note:       "_Throughput, not context size — each call re-sends the window above._"
+    estimated:             "Estimated context: ~{count} tokens across {messages} messages"
+    detail_after_first:    "_(Full compression and throughput stats available after the first agent response)_"
+    no_data:               "No context data available yet. Send a message to start a session."
 
   status:
     header:                "📊 **Stádas Hermes Gateway**"
@@ -303,7 +321,7 @@ Future messages in this room will use that transcript until `/reset` or another
     model_provider:        "**Model:** `{model}` ({provider})"
     context:               "**Context:** {used} / {total} ({pct}%)"
     context_used:          "**Context:** ~{used} tokens"
-    tokens:                "**Comharthaí:** {tokens}"
+    tokens:                "**Lifetime tokens billed:** {tokens} _(not your current context size; use `/context`)_"
     agent_running:         "**Gníomhaire ag rith:** {state}"
     state_yes:             "Tá ⚡"
     state_no:              "Níl"
diff --git a/locales/hu.yaml b/locales/hu.yaml
index 2fec63848446..8bbcafe7c2b3 100644
--- a/locales/hu.yaml
+++ b/locales/hu.yaml
@@ -282,6 +282,24 @@ Future messages in this room will use that transcript until `/reset` or another
   set_home:
     save_failed:           "Nem sikerült menteni a kezdőcsatornát: {error}"
     success:               "✅ Kezdőcsatorna beállítva: **{name}** (ID: {chat_id}).\nA cron-feladatok és a platformok közötti üzenetek ide érkeznek."
+  context:
+    header:                "🧠 **Context Window**"
+    model:                 "Model: `{model}`"
+    window:                "Window: {total} tokens"
+    in_use:                "In use: {used} / {total} ({pct}%)"
+    bar:                   "{bar}"
+    headroom:              "Headroom to limit: {headroom} tokens"
+    threshold:             "Auto-compresses at: {threshold} ({threshold_pct}%) — {to_go} to go"
+    over_threshold:        "⚠️ **Over auto-compression threshold ({threshold}, {threshold_pct}%)**"
+    compressions:          "Compressions this session: {count}"
+    last_savings:          "Last compression freed: {savings}% of context"
+    totals_header:         "Session totals (cumulative across {calls} API calls)"
+    totals_line:           "Input {input} · Output {output} · Reasoning {reasoning}"
+    total_billed:          "Total billed: {total}"
+    throughput_note:       "_Throughput, not context size — each call re-sends the window above._"
+    estimated:             "Estimated context: ~{count} tokens across {messages} messages"
+    detail_after_first:    "_(Full compression and throughput stats available after the first agent response)_"
+    no_data:               "No context data available yet. Send a message to start a session."
 
   status:
     header:                "📊 **Hermes Gateway állapot**"
@@ -299,7 +317,7 @@ Future messages in this room will use that transcript until `/reset` or another
     model_provider:        "**Model:** `{model}` ({provider})"
     context:               "**Context:** {used} / {total} ({pct}%)"
     context_used:          "**Context:** ~{used} tokens"
-    tokens:                "**Tokenek:** {tokens}"
+    tokens:                "**Lifetime tokens billed:** {tokens} _(not your current context size; use `/context`)_"
     agent_running:         "**Ügynök fut:** {state}"
     state_yes:             "Igen ⚡"
     state_no:              "Nem"
diff --git a/locales/it.yaml b/locales/it.yaml
index eafcbb1e29d8..2b0a0677accb 100644
--- a/locales/it.yaml
+++ b/locales/it.yaml
@@ -282,6 +282,24 @@ Future messages in this room will use that transcript until `/reset` or another
   set_home:
     save_failed:           "Salvataggio del canale home non riuscito: {error}"
     success:               "✅ Canale home impostato su **{name}** (ID: {chat_id}).\nI cron job e i messaggi cross-platform verranno consegnati qui."
+  context:
+    header:                "🧠 **Context Window**"
+    model:                 "Model: `{model}`"
+    window:                "Window: {total} tokens"
+    in_use:                "In use: {used} / {total} ({pct}%)"
+    bar:                   "{bar}"
+    headroom:              "Headroom to limit: {headroom} tokens"
+    threshold:             "Auto-compresses at: {threshold} ({threshold_pct}%) — {to_go} to go"
+    over_threshold:        "⚠️ **Over auto-compression threshold ({threshold}, {threshold_pct}%)**"
+    compressions:          "Compressions this session: {count}"
+    last_savings:          "Last compression freed: {savings}% of context"
+    totals_header:         "Session totals (cumulative across {calls} API calls)"
+    totals_line:           "Input {input} · Output {output} · Reasoning {reasoning}"
+    total_billed:          "Total billed: {total}"
+    throughput_note:       "_Throughput, not context size — each call re-sends the window above._"
+    estimated:             "Estimated context: ~{count} tokens across {messages} messages"
+    detail_after_first:    "_(Full compression and throughput stats available after the first agent response)_"
+    no_data:               "No context data available yet. Send a message to start a session."
 
   status:
     header:                "📊 **Stato del Gateway Hermes**"
@@ -299,7 +317,7 @@ Future messages in this room will use that transcript until `/reset` or another
     model_provider:        "**Model:** `{model}` ({provider})"
     context:               "**Context:** {used} / {total} ({pct}%)"
     context_used:          "**Context:** ~{used} tokens"
-    tokens:                "**Token:** {tokens}"
+    tokens:                "**Lifetime tokens billed:** {tokens} _(not your current context size; use `/context`)_"
     agent_running:         "**Agente in esecuzione:** {state}"
     state_yes:             "Sì ⚡"
     state_no:              "No"
diff --git a/locales/ja.yaml b/locales/ja.yaml
index ea78e7aa6f7b..0b40dc1f614a 100644
--- a/locales/ja.yaml
+++ b/locales/ja.yaml
@@ -282,6 +282,24 @@ Future messages in this room will use that transcript until `/reset` or another
   set_home:
     save_failed:           "ホームチャンネルを保存できませんでした: {error}"
     success:               "✅ ホームチャンネルを **{name}** (ID: {chat_id}) に設定しました。\nCron ジョブとプラットフォーム間メッセージはここに配信されます。"
+  context:
+    header:                "🧠 **Context Window**"
+    model:                 "Model: `{model}`"
+    window:                "Window: {total} tokens"
+    in_use:                "In use: {used} / {total} ({pct}%)"
+    bar:                   "{bar}"
+    headroom:              "Headroom to limit: {headroom} tokens"
+    threshold:             "Auto-compresses at: {threshold} ({threshold_pct}%) — {to_go} to go"
+    over_threshold:        "⚠️ **Over auto-compression threshold ({threshold}, {threshold_pct}%)**"
+    compressions:          "Compressions this session: {count}"
+    last_savings:          "Last compression freed: {savings}% of context"
+    totals_header:         "Session totals (cumulative across {calls} API calls)"
+    totals_line:           "Input {input} · Output {output} · Reasoning {reasoning}"
+    total_billed:          "Total billed: {total}"
+    throughput_note:       "_Throughput, not context size — each call re-sends the window above._"
+    estimated:             "Estimated context: ~{count} tokens across {messages} messages"
+    detail_after_first:    "_(Full compression and throughput stats available after the first agent response)_"
+    no_data:               "No context data available yet. Send a message to start a session."
 
   status:
     header:                "📊 **Hermes ゲートウェイ状態**"
@@ -299,7 +317,7 @@ Future messages in this room will use that transcript until `/reset` or another
     model_provider:        "**Model:** `{model}` ({provider})"
     context:               "**Context:** {used} / {total} ({pct}%)"
     context_used:          "**Context:** ~{used} tokens"
-    tokens:                "**トークン:** {tokens}"
+    tokens:                "**Lifetime tokens billed:** {tokens} _(not your current context size; use `/context`)_"
     agent_running:         "**エージェント実行中:** {state}"
     state_yes:             "はい ⚡"
     state_no:              "いいえ"
diff --git a/locales/ko.yaml b/locales/ko.yaml
index 2411c9f1d179..db38ba98e8d6 100644
--- a/locales/ko.yaml
+++ b/locales/ko.yaml
@@ -282,6 +282,24 @@ Future messages in this room will use that transcript until `/reset` or another
   set_home:
     save_failed:           "홈 채널 저장에 실패했습니다: {error}"
     success:               "✅ 홈 채널이 **{name}**(ID: {chat_id})(으)로 설정되었습니다.\n크론 작업과 플랫폼 간 메시지가 여기로 전달됩니다."
+  context:
+    header:                "🧠 **Context Window**"
+    model:                 "Model: `{model}`"
+    window:                "Window: {total} tokens"
+    in_use:                "In use: {used} / {total} ({pct}%)"
+    bar:                   "{bar}"
+    headroom:              "Headroom to limit: {headroom} tokens"
+    threshold:             "Auto-compresses at: {threshold} ({threshold_pct}%) — {to_go} to go"
+    over_threshold:        "⚠️ **Over auto-compression threshold ({threshold}, {threshold_pct}%)**"
+    compressions:          "Compressions this session: {count}"
+    last_savings:          "Last compression freed: {savings}% of context"
+    totals_header:         "Session totals (cumulative across {calls} API calls)"
+    totals_line:           "Input {input} · Output {output} · Reasoning {reasoning}"
+    total_billed:          "Total billed: {total}"
+    throughput_note:       "_Throughput, not context size — each call re-sends the window above._"
+    estimated:             "Estimated context: ~{count} tokens across {messages} messages"
+    detail_after_first:    "_(Full compression and throughput stats available after the first agent response)_"
+    no_data:               "No context data available yet. Send a message to start a session."
 
   status:
     header:                "📊 **Hermes 게이트웨이 상태**"
@@ -299,7 +317,7 @@ Future messages in this room will use that transcript until `/reset` or another
     model_provider:        "**Model:** `{model}` ({provider})"
     context:               "**Context:** {used} / {total} ({pct}%)"
     context_used:          "**Context:** ~{used} tokens"
-    tokens:                "**토큰:** {tokens}"
+    tokens:                "**Lifetime tokens billed:** {tokens} _(not your current context size; use `/context`)_"
     agent_running:         "**에이전트 실행 중:** {state}"
     state_yes:             "예 ⚡"
     state_no:              "아니오"
diff --git a/locales/pt.yaml b/locales/pt.yaml
index ada4682708fb..0f32bf53df26 100644
--- a/locales/pt.yaml
+++ b/locales/pt.yaml
@@ -282,6 +282,24 @@ Future messages in this room will use that transcript until `/reset` or another
   set_home:
     save_failed:           "Falha ao guardar o canal principal: {error}"
     success:               "✅ Canal principal definido como **{name}** (ID: {chat_id}).\nAs tarefas cron e mensagens entre plataformas serão entregues aqui."
+  context:
+    header:                "🧠 **Context Window**"
+    model:                 "Model: `{model}`"
+    window:                "Window: {total} tokens"
+    in_use:                "In use: {used} / {total} ({pct}%)"
+    bar:                   "{bar}"
+    headroom:              "Headroom to limit: {headroom} tokens"
+    threshold:             "Auto-compresses at: {threshold} ({threshold_pct}%) — {to_go} to go"
+    over_threshold:        "⚠️ **Over auto-compression threshold ({threshold}, {threshold_pct}%)**"
+    compressions:          "Compressions this session: {count}"
+    last_savings:          "Last compression freed: {savings}% of context"
+    totals_header:         "Session totals (cumulative across {calls} API calls)"
+    totals_line:           "Input {input} · Output {output} · Reasoning {reasoning}"
+    total_billed:          "Total billed: {total}"
+    throughput_note:       "_Throughput, not context size — each call re-sends the window above._"
+    estimated:             "Estimated context: ~{count} tokens across {messages} messages"
+    detail_after_first:    "_(Full compression and throughput stats available after the first agent response)_"
+    no_data:               "No context data available yet. Send a message to start a session."
 
   status:
     header:                "📊 **Estado do Hermes Gateway**"
@@ -299,7 +317,7 @@ Future messages in this room will use that transcript until `/reset` or another
     model_provider:        "**Model:** `{model}` ({provider})"
     context:               "**Context:** {used} / {total} ({pct}%)"
     context_used:          "**Context:** ~{used} tokens"
-    tokens:                "**Tokens de API cumulativos (reenviados a cada chamada):** {tokens}"
+    tokens:                "**Lifetime tokens billed:** {tokens} _(not your current context size; use `/context`)_"
     agent_running:         "**Agente em execução:** {state}"
     state_yes:             "Sim ⚡"
     state_no:              "Não"
diff --git a/locales/ru.yaml b/locales/ru.yaml
index c83483eae799..51347514d052 100644
--- a/locales/ru.yaml
+++ b/locales/ru.yaml
@@ -282,6 +282,24 @@ Future messages in this room will use that transcript until `/reset` or another
   set_home:
     save_failed:           "Не удалось сохранить главный канал: {error}"
     success:               "✅ Главный канал установлен на **{name}** (ID: {chat_id}).\nCron-задачи и межплатформенные сообщения будут доставляться сюда."
+  context:
+    header:                "🧠 **Context Window**"
+    model:                 "Model: `{model}`"
+    window:                "Window: {total} tokens"
+    in_use:                "In use: {used} / {total} ({pct}%)"
+    bar:                   "{bar}"
+    headroom:              "Headroom to limit: {headroom} tokens"
+    threshold:             "Auto-compresses at: {threshold} ({threshold_pct}%) — {to_go} to go"
+    over_threshold:        "⚠️ **Over auto-compression threshold ({threshold}, {threshold_pct}%)**"
+    compressions:          "Compressions this session: {count}"
+    last_savings:          "Last compression freed: {savings}% of context"
+    totals_header:         "Session totals (cumulative across {calls} API calls)"
+    totals_line:           "Input {input} · Output {output} · Reasoning {reasoning}"
+    total_billed:          "Total billed: {total}"
+    throughput_note:       "_Throughput, not context size — each call re-sends the window above._"
+    estimated:             "Estimated context: ~{count} tokens across {messages} messages"
+    detail_after_first:    "_(Full compression and throughput stats available after the first agent response)_"
+    no_data:               "No context data available yet. Send a message to start a session."
 
   status:
     header:                "📊 **Состояние Hermes Gateway**"
@@ -299,7 +317,7 @@ Future messages in this room will use that transcript until `/reset` or another
     model_provider:        "**Model:** `{model}` ({provider})"
     context:               "**Context:** {used} / {total} ({pct}%)"
     context_used:          "**Context:** ~{used} tokens"
-    tokens:                "**Токены:** {tokens}"
+    tokens:                "**Lifetime tokens billed:** {tokens} _(not your current context size; use `/context`)_"
     agent_running:         "**Агент активен:** {state}"
     state_yes:             "Да ⚡"
     state_no:              "Нет"
diff --git a/locales/tr.yaml b/locales/tr.yaml
index 33d2c9383f1a..d1a9808f79c3 100644
--- a/locales/tr.yaml
+++ b/locales/tr.yaml
@@ -282,6 +282,24 @@ Future messages in this room will use that transcript until `/reset` or another
   set_home:
     save_failed:           "Ana kanal kaydedilemedi: {error}"
     success:               "✅ Ana kanal **{name}** (ID: {chat_id}) olarak ayarlandı.\nCron işleri ve platformlar arası mesajlar buraya iletilecek."
+  context:
+    header:                "🧠 **Context Window**"
+    model:                 "Model: `{model}`"
+    window:                "Window: {total} tokens"
+    in_use:                "In use: {used} / {total} ({pct}%)"
+    bar:                   "{bar}"
+    headroom:              "Headroom to limit: {headroom} tokens"
+    threshold:             "Auto-compresses at: {threshold} ({threshold_pct}%) — {to_go} to go"
+    over_threshold:        "⚠️ **Over auto-compression threshold ({threshold}, {threshold_pct}%)**"
+    compressions:          "Compressions this session: {count}"
+    last_savings:          "Last compression freed: {savings}% of context"
+    totals_header:         "Session totals (cumulative across {calls} API calls)"
+    totals_line:           "Input {input} · Output {output} · Reasoning {reasoning}"
+    total_billed:          "Total billed: {total}"
+    throughput_note:       "_Throughput, not context size — each call re-sends the window above._"
+    estimated:             "Estimated context: ~{count} tokens across {messages} messages"
+    detail_after_first:    "_(Full compression and throughput stats available after the first agent response)_"
+    no_data:               "No context data available yet. Send a message to start a session."
 
   status:
     header:                "📊 **Hermes Gateway Durumu**"
@@ -299,7 +317,7 @@ Future messages in this room will use that transcript until `/reset` or another
     model_provider:        "**Model:** `{model}` ({provider})"
     context:               "**Context:** {used} / {total} ({pct}%)"
     context_used:          "**Context:** ~{used} tokens"
-    tokens:                "**Token:** {tokens}"
+    tokens:                "**Lifetime tokens billed:** {tokens} _(not your current context size; use `/context`)_"
     agent_running:         "**Aracı çalışıyor:** {state}"
     state_yes:             "Evet ⚡"
     state_no:              "Hayır"
diff --git a/locales/uk.yaml b/locales/uk.yaml
index f9e5cfd7d450..0c67278b775d 100644
--- a/locales/uk.yaml
+++ b/locales/uk.yaml
@@ -282,6 +282,24 @@ Future messages in this room will use that transcript until `/reset` or another
   set_home:
     save_failed:           "Не вдалося зберегти головний канал: {error}"
     success:               "✅ Головний канал встановлено на **{name}** (ID: {chat_id}).\nCron-завдання та міжплатформні повідомлення доставлятимуться сюди."
+  context:
+    header:                "🧠 **Context Window**"
+    model:                 "Model: `{model}`"
+    window:                "Window: {total} tokens"
+    in_use:                "In use: {used} / {total} ({pct}%)"
+    bar:                   "{bar}"
+    headroom:              "Headroom to limit: {headroom} tokens"
+    threshold:             "Auto-compresses at: {threshold} ({threshold_pct}%) — {to_go} to go"
+    over_threshold:        "⚠️ **Over auto-compression threshold ({threshold}, {threshold_pct}%)**"
+    compressions:          "Compressions this session: {count}"
+    last_savings:          "Last compression freed: {savings}% of context"
+    totals_header:         "Session totals (cumulative across {calls} API calls)"
+    totals_line:           "Input {input} · Output {output} · Reasoning {reasoning}"
+    total_billed:          "Total billed: {total}"
+    throughput_note:       "_Throughput, not context size — each call re-sends the window above._"
+    estimated:             "Estimated context: ~{count} tokens across {messages} messages"
+    detail_after_first:    "_(Full compression and throughput stats available after the first agent response)_"
+    no_data:               "No context data available yet. Send a message to start a session."
 
   status:
     header:                "📊 **Стан Hermes Gateway**"
@@ -299,7 +317,7 @@ Future messages in this room will use that transcript until `/reset` or another
     model_provider:        "**Model:** `{model}` ({provider})"
     context:               "**Context:** {used} / {total} ({pct}%)"
     context_used:          "**Context:** ~{used} tokens"
-    tokens:                "**Токени:** {tokens}"
+    tokens:                "**Lifetime tokens billed:** {tokens} _(not your current context size; use `/context`)_"
     agent_running:         "**Агент активний:** {state}"
     state_yes:             "Так ⚡"
     state_no:              "Ні"
diff --git a/locales/zh-hant.yaml b/locales/zh-hant.yaml
index 8df92cc23b22..e5ef7b82ddbb 100644
--- a/locales/zh-hant.yaml
+++ b/locales/zh-hant.yaml
@@ -282,6 +282,24 @@ Future messages in this room will use that transcript until `/reset` or another
   set_home:
     save_failed:           "無法儲存主頻道：{error}"
     success:               "✅ 主頻道已設定為 **{name}**（ID：{chat_id}）。\n排程任務和跨平台訊息將傳送至此處。"
+  context:
+    header:                "🧠 **Context Window**"
+    model:                 "Model: `{model}`"
+    window:                "Window: {total} tokens"
+    in_use:                "In use: {used} / {total} ({pct}%)"
+    bar:                   "{bar}"
+    headroom:              "Headroom to limit: {headroom} tokens"
+    threshold:             "Auto-compresses at: {threshold} ({threshold_pct}%) — {to_go} to go"
+    over_threshold:        "⚠️ **Over auto-compression threshold ({threshold}, {threshold_pct}%)**"
+    compressions:          "Compressions this session: {count}"
+    last_savings:          "Last compression freed: {savings}% of context"
+    totals_header:         "Session totals (cumulative across {calls} API calls)"
+    totals_line:           "Input {input} · Output {output} · Reasoning {reasoning}"
+    total_billed:          "Total billed: {total}"
+    throughput_note:       "_Throughput, not context size — each call re-sends the window above._"
+    estimated:             "Estimated context: ~{count} tokens across {messages} messages"
+    detail_after_first:    "_(Full compression and throughput stats available after the first agent response)_"
+    no_data:               "No context data available yet. Send a message to start a session."
 
   status:
     header:                "📊 **Hermes 閘道狀態**"
@@ -299,7 +317,7 @@ Future messages in this room will use that transcript until `/reset` or another
     model_provider:        "**Model:** `{model}` ({provider})"
     context:               "**Context:** {used} / {total} ({pct}%)"
     context_used:          "**Context:** ~{used} tokens"
-    tokens:                "**Token 數：** {tokens}"
+    tokens:                "**Lifetime tokens billed:** {tokens} _(not your current context size; use `/context`)_"
     agent_running:         "**代理執行中：** {state}"
     state_yes:             "是 ⚡"
     state_no:              "否"
diff --git a/locales/zh.yaml b/locales/zh.yaml
index d13d8cdfa185..f16f330028ab 100644
--- a/locales/zh.yaml
+++ b/locales/zh.yaml
@@ -282,6 +282,24 @@ Future messages in this room will use that transcript until `/reset` or another
   set_home:
     save_failed:           "无法保存主频道：{error}"
     success:               "✅ 主频道已设置为 **{name}**（ID：{chat_id}）。\n定时任务和跨平台消息将发送到此处。"
+  context:
+    header:                "🧠 **Context Window**"
+    model:                 "Model: `{model}`"
+    window:                "Window: {total} tokens"
+    in_use:                "In use: {used} / {total} ({pct}%)"
+    bar:                   "{bar}"
+    headroom:              "Headroom to limit: {headroom} tokens"
+    threshold:             "Auto-compresses at: {threshold} ({threshold_pct}%) — {to_go} to go"
+    over_threshold:        "⚠️ **Over auto-compression threshold ({threshold}, {threshold_pct}%)**"
+    compressions:          "Compressions this session: {count}"
+    last_savings:          "Last compression freed: {savings}% of context"
+    totals_header:         "Session totals (cumulative across {calls} API calls)"
+    totals_line:           "Input {input} · Output {output} · Reasoning {reasoning}"
+    total_billed:          "Total billed: {total}"
+    throughput_note:       "_Throughput, not context size — each call re-sends the window above._"
+    estimated:             "Estimated context: ~{count} tokens across {messages} messages"
+    detail_after_first:    "_(Full compression and throughput stats available after the first agent response)_"
+    no_data:               "No context data available yet. Send a message to start a session."
 
   status:
     header:                "📊 **Hermes 网关状态**"
@@ -299,7 +317,7 @@ Future messages in this room will use that transcript until `/reset` or another
     model_provider:        "**Model:** `{model}` ({provider})"
     context:               "**Context:** {used} / {total} ({pct}%)"
     context_used:          "**Context:** ~{used} tokens"
-    tokens:                "**Token 数：** {tokens}"
+    tokens:                "**Lifetime tokens billed:** {tokens} _(not your current context size; use `/context`)_"
     agent_running:         "**代理运行中：** {state}"
     state_yes:             "是 ⚡"
     state_no:              "否"
diff --git a/tests/gateway/test_status_command.py b/tests/gateway/test_status_command.py
index 2da89d16dc8f..a4e864985a18 100644
--- a/tests/gateway/test_status_command.py
+++ b/tests/gateway/test_status_command.py
@@ -100,7 +100,7 @@ async def test_status_command_reports_running_agent_without_interrupt(monkeypatc
     result = await runner._handle_message(_make_event("/status"))
 
     assert "**Session ID:** `sess-1`" in result
-    assert "**Cumulative API tokens (re-sent each call):** 321" in result
+    assert "**Lifetime tokens billed:** 321" in result
     assert "**Agent Running:** Yes ⚡" in result
     assert "**Title:**" not in result
     running_agent.interrupt.assert_not_called()
@@ -153,7 +153,7 @@ async def test_status_command_reads_token_totals_from_session_db():
     result = await runner._handle_message(_make_event("/status"))
 
     # 1000 + 250 + 500 + 100 + 50 = 1,900
-    assert "**Cumulative API tokens (re-sent each call):** 1,900" in result
+    assert "**Lifetime tokens billed:** 1,900" in result
 
 
 @pytest.mark.asyncio
@@ -174,7 +174,7 @@ async def test_status_command_tokens_zero_when_session_db_row_missing():
 
     result = await runner._handle_message(_make_event("/status"))
 
-    assert "**Cumulative API tokens (re-sent each call):** 0" in result
+    assert "**Lifetime tokens billed:** 0" in result
 
 
 @pytest.mark.asyncio
@@ -212,8 +212,7 @@ async def test_status_command_includes_live_agent_model_and_context():
 
     assert "**Model:** `openai/gpt-test` (openai)" in result
     assert "**Context:** 12,345 / 100,000 (12%)" in result
-    assert "**Cumulative API tokens (re-sent each call):** 1,250" in result
-    assert "1,250 (cumulative)" not in result
+    assert "**Lifetime tokens billed:** 1,250" in result
 
 
 @pytest.mark.asyncio
@@ -245,7 +244,7 @@ async def test_status_command_includes_persisted_model_and_context_when_agent_no
 
     assert "**Model:** `openai/gpt-persisted` (openai-codex)" in result
     assert "**Context:** 24,000 / 272,000 (9%)" in result
-    assert "**Cumulative API tokens (re-sent each call):** 2,500" in result
+    assert "**Lifetime tokens billed:** 2,500" in result
 
 
 @pytest.mark.asyncio
@@ -821,3 +820,152 @@ async def test_post_delivery_callback_generation_snapshot_happens_after_bind():
     assert fired == []
     assert session_key in adapter._post_delivery_callbacks
     assert adapter._post_delivery_callbacks[session_key][0] == 2
+
+
+# ── /context command tests ────────────────────────────────────────────────
+
+def _stub_agent(**overrides) -> SimpleNamespace:
+    """Build a stub agent with the attributes _handle_context_command reads."""
+    props = dict(
+        model="openai/gpt-test",
+        context_compressor=SimpleNamespace(
+            last_prompt_tokens=47_231,
+            context_length=200_000,
+            threshold_tokens=100_000,
+            threshold_percent=0.5,
+            compression_count=2,
+            _last_compression_savings_pct=63.0,
+        ),
+        session_api_calls=47,
+        session_input_tokens=410_000,
+        session_output_tokens=38_000,
+        session_reasoning_tokens=12_000,
+        session_total_tokens=3_158_641,
+        session_cache_read_tokens=2_900_000,
+        session_cache_write_tokens=48_000,
+    )
+    props.update(overrides)
+    return SimpleNamespace(**props)
+
+
+@pytest.mark.asyncio
+async def test_context_command_live_agent():
+    """/context with a live running agent shows the full view: gauge,
+    compression, and throughput — but NOT cache stats."""
+    session_entry = SessionEntry(
+        session_key=build_session_key(_make_source()),
+        session_id="sess-1",
+        created_at=datetime.now(),
+        updated_at=datetime.now(),
+        platform=Platform.TELEGRAM,
+        chat_type="dm",
+    )
+    runner = _make_runner(session_entry)
+    session_key = session_entry.session_key
+    agent = _stub_agent()
+    runner._running_agents[session_key] = agent
+
+    result = await runner._handle_context_command(_make_event("/context"))
+
+    assert "🧠 **Context Window**" in result
+    assert "Model: `openai/gpt-test`" in result
+    assert "Window: 200,000 tokens" in result
+    assert "In use: 47,231 / 200,000 (24%)" in result
+    assert "Headroom to limit: 152,769 tokens" in result
+    # Compression section
+    assert "Auto-compresses at: 100,000 (50%)" in result
+    assert "Compressions this session: 2" in result
+    assert "Last compression freed: 63% of context" in result
+    # Throughput — NOT cache
+    assert "Session totals (cumulative across 47 API calls)" in result
+    assert "Input 410,000" in result
+    assert "Output 38,000" in result
+    assert "Reasoning 12,000" in result
+    assert "Total billed: 3,158,641" in result
+    assert "each call re-sends the window above" in result
+    # Cache stats must NOT appear (removed per design)
+    assert "Cache read" not in result
+    assert "Cache write" not in result
+    assert "Cache hit" not in result
+    assert "Hit rate" not in result
+
+
+@pytest.mark.asyncio
+async def test_context_command_over_threshold():
+    """When used >= threshold, the over-threshold warning is shown."""
+    session_entry = SessionEntry(
+        session_key=build_session_key(_make_source()),
+        session_id="sess-2",
+        created_at=datetime.now(),
+        updated_at=datetime.now(),
+        platform=Platform.TELEGRAM,
+        chat_type="dm",
+    )
+    runner = _make_runner(session_entry)
+    session_key = session_entry.session_key
+    agent = _stub_agent(
+        context_compressor=SimpleNamespace(
+            last_prompt_tokens=150_000,
+            context_length=200_000,
+            threshold_tokens=100_000,
+            threshold_percent=0.5,
+            compression_count=5,
+            _last_compression_savings_pct=40.0,
+        )
+    )
+    runner._running_agents[session_key] = agent
+
+    result = await runner._handle_context_command(_make_event("/context"))
+
+    assert "⚠️" in result
+    assert "Over auto-compression threshold" in result
+
+
+@pytest.mark.asyncio
+async def test_context_command_no_agent_transcript_fallback():
+    """When no agent is resident and session_entry has no last_prompt_tokens,
+    /context falls back to a transcript estimate."""
+    session_entry = SessionEntry(
+        session_key=build_session_key(_make_source()),
+        session_id="sess-3",
+        created_at=datetime.now(),
+        updated_at=datetime.now(),
+        platform=Platform.TELEGRAM,
+        chat_type="dm",
+        last_prompt_tokens=0,  # No live context data
+    )
+    runner = _make_runner(session_entry)
+    # Stub the transcript so estimate_messages_tokens_rough has something to work with
+    runner.session_store.load_transcript.return_value = [
+        {"role": "user", "content": "Hello"},
+        {"role": "assistant", "content": "Hi there!"},
+        {"role": "user", "content": "What's my balance?"},
+        {"role": "assistant", "content": "Your balance is $1,000."},
+    ]
+
+    result = await runner._handle_context_command(_make_event("/context"))
+
+    assert "🧠 **Context Window**" in result
+    assert "Estimated context:" in result
+    assert "4 messages" in result
+
+
+@pytest.mark.asyncio
+async def test_context_command_no_data():
+    """When there's no agent, no session data, and no transcript,
+    /context returns the no-data message."""
+    session_entry = SessionEntry(
+        session_key=build_session_key(_make_source()),
+        session_id="sess-4",
+        created_at=datetime.now(),
+        updated_at=datetime.now(),
+        platform=Platform.TELEGRAM,
+        chat_type="dm",
+        last_prompt_tokens=0,
+    )
+    runner = _make_runner(session_entry)
+    runner.session_store.load_transcript.return_value = []
+
+    result = await runner._handle_context_command(_make_event("/context"))
+
+    assert "No context data available yet" in result