feat(usage): track cloud quotas and expose :limits

Acceptance Criteria:\n- header shows hourly/weekly usage with colored thresholds\n- :limits command prints persisted usage data and quotas\n- token usage survives restarts and emits 80%/95% toasts

Test Notes:\n- cargo test -p owlen-core usage

CHANGELOG.md

@@ -28,6 +28,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Secure credential vault integration for Ollama Cloud API keys when `privacy.encrypt_local_data = true`.
 - Input panel respects a new `ui.input_max_rows` setting so long prompts expand predictably before scrolling kicks in.
 - Command palette offers fuzzy `:model` filtering and `:provider` completions for fast switching.
+- Cloud usage tracker persists hourly/weekly token totals, adds a `:limits` command, shows live header badges, and raises toast warnings at 80 %/95 % of the configured quotas.
 - Message rendering caches wrapped lines and throttles streaming redraws to keep the TUI responsive on long sessions.
 - Model picker badges now inspect provider capabilities so vision/audio/thinking models surface the correct icons even when descriptions are sparse.
 - Chat history honors `ui.scrollback_lines`, trimming older rows to keep the TUI responsive and surfacing a "↓ New messages" badge whenever updates land off-screen.

README.md

@@ -119,6 +119,7 @@ Model discovery commands worth remembering:
 
 - `:models --local` or `:models --cloud` jump directly to the corresponding section in the picker.
 - `:cloud setup [--force-cloud-base-url]` stores your cloud API key without clobbering an existing local base URL (unless you opt in with the flag).
+- `:limits` prints the locally tracked hourly/weekly token totals for each provider and mirrors the values shown in the chat header.
 When a catalogue is unreachable, Owlen now tags the picker with `Local unavailable` / `Cloud unavailable` so you can recover without guessing.
 
 ## Documentation

config.toml

@@ -15,6 +15,8 @@ enabled = false
 provider_type = "ollama_cloud"
 base_url = "https://ollama.com"
 api_key_env = "OLLAMA_API_KEY"
+hourly_quota_tokens = 50000
+weekly_quota_tokens = 250000
 
 [providers.openai]
 enabled = false

crates/owlen-core/src/config.rs

@@ -32,6 +32,10 @@ pub const OLLAMA_API_KEY_ENV: &str = "OLLAMA_API_KEY";
 pub const LEGACY_OLLAMA_CLOUD_API_KEY_ENV: &str = "OLLAMA_CLOUD_API_KEY";
 /// Legacy environment variable used by earlier Owlen releases.
 pub const LEGACY_OWLEN_OLLAMA_CLOUD_API_KEY_ENV: &str = "OWLEN_OLLAMA_CLOUD_API_KEY";
+/// Default hourly soft quota for Ollama Cloud usage visualization (tokens).
+pub const DEFAULT_OLLAMA_CLOUD_HOURLY_QUOTA: u64 = 50_000;
+/// Default weekly soft quota for Ollama Cloud usage visualization (tokens).
+pub const DEFAULT_OLLAMA_CLOUD_WEEKLY_QUOTA: u64 = 250_000;
 /// Default base URL for local Ollama daemons.
 pub const OLLAMA_LOCAL_BASE_URL: &str = "http://localhost:11434";
 /// Default OpenAI API base URL.
@@ -880,6 +884,14 @@ fn default_ollama_cloud_config() -> ProviderConfig {
         OLLAMA_CLOUD_ENDPOINT_KEY.to_string(),
         serde_json::Value::String(OLLAMA_CLOUD_BASE_URL.to_string()),
     );
+    extra.insert(
+        "hourly_quota_tokens".to_string(),
+        serde_json::Value::Number(serde_json::Number::from(DEFAULT_OLLAMA_CLOUD_HOURLY_QUOTA)),
+    );
+    extra.insert(
+        "weekly_quota_tokens".to_string(),
+        serde_json::Value::Number(serde_json::Number::from(DEFAULT_OLLAMA_CLOUD_WEEKLY_QUOTA)),
+    );
 
     ProviderConfig {
         enabled: false,

crates/owlen-core/src/lib.rs

@@ -30,6 +30,7 @@ pub mod theme;
 pub mod tools;
 pub mod types;
 pub mod ui;
+pub mod usage;
 pub mod validation;
 pub mod wrap_cursor;
 
@@ -61,6 +62,7 @@ pub use session::*;
 pub use state::*;
 pub use theme::*;
 pub use tools::*;
+pub use usage::*;
 pub use validation::*;
 
 /// Result type used throughout the OWLEN ecosystem

crates/owlen-core/src/providers/ollama.rs

@@ -41,8 +41,9 @@ use uuid::Uuid;
 use crate::{
     Error, Result,
     config::{
-        GeneralSettings, LEGACY_OLLAMA_CLOUD_API_KEY_ENV, LEGACY_OWLEN_OLLAMA_CLOUD_API_KEY_ENV,
-        OLLAMA_API_KEY_ENV, OLLAMA_CLOUD_BASE_URL, OLLAMA_CLOUD_ENDPOINT_KEY, OLLAMA_MODE_KEY,
+        DEFAULT_OLLAMA_CLOUD_HOURLY_QUOTA, DEFAULT_OLLAMA_CLOUD_WEEKLY_QUOTA, GeneralSettings,
+        LEGACY_OLLAMA_CLOUD_API_KEY_ENV, LEGACY_OWLEN_OLLAMA_CLOUD_API_KEY_ENV, OLLAMA_API_KEY_ENV,
+        OLLAMA_CLOUD_BASE_URL, OLLAMA_CLOUD_ENDPOINT_KEY, OLLAMA_MODE_KEY,
     },
     llm::{LlmProvider, ProviderConfig},
     mcp::McpToolDescriptor,
@@ -1505,6 +1506,18 @@ impl LlmProvider for OllamaProvider {
                     "description": "Seconds to cache model listings",
                     "minimum": 5,
                     "default": DEFAULT_MODEL_CACHE_TTL_SECS
+                },
+                "hourly_quota_tokens": {
+                    "type": "integer",
+                    "description": "Soft hourly token quota used for UI alerts",
+                    "minimum": 0,
+                    "default": DEFAULT_OLLAMA_CLOUD_HOURLY_QUOTA
+                },
+                "weekly_quota_tokens": {
+                    "type": "integer",
+                    "description": "Soft weekly token quota used for UI alerts",
+                    "minimum": 0,
+                    "default": DEFAULT_OLLAMA_CLOUD_WEEKLY_QUOTA
                 }
             }
         })

crates/owlen-core/src/session.rs

@@ -19,6 +19,7 @@ use crate::types::{
     ChatParameters, ChatRequest, ChatResponse, Conversation, Message, ModelInfo, ToolCall,
 };
 use crate::ui::{RoleLabelDisplay, UiController};
+use crate::usage::{UsageLedger, UsageQuota, UsageSnapshot};
 use crate::validation::{SchemaValidator, get_builtin_schemas};
 use crate::{ChatStream, Provider};
 use crate::{
@@ -29,10 +30,12 @@ use crate::{Error, Result};
 use chrono::Utc;
 use log::warn;
 use serde_json::{Value, json};
-use std::collections::HashMap;
+use std::collections::{HashMap, HashSet};
 use std::env;
 use std::path::PathBuf;
 use std::sync::{Arc, Mutex};
+use std::time::SystemTime;
+use tokio::fs;
 use tokio::sync::Mutex as TokioMutex;
 use tokio::sync::mpsc::UnboundedSender;
 use uuid::Uuid;
@@ -240,6 +243,7 @@ pub struct SessionController {
     event_tx: Option<UnboundedSender<ControllerEvent>>,
     pending_tool_requests: HashMap<Uuid, PendingToolRequest>,
     stream_states: HashMap<Uuid, StreamingMessageState>,
+    usage_ledger: Arc<TokioMutex<UsageLedger>>,
 }
 
 async fn build_tools(
@@ -545,6 +549,25 @@ impl SessionController {
         )
         .await?;
 
+        let usage_ledger_path = storage
+            .database_path()
+            .parent()
+            .map(|dir| dir.join("usage-ledger.json"))
+            .unwrap_or_else(|| PathBuf::from("usage-ledger.json"));
+
+        let usage_ledger_instance =
+            match UsageLedger::load_or_default(usage_ledger_path.clone()).await {
+                Ok(ledger) => ledger,
+                Err(err) => {
+                    warn!(
+                        "Failed to load usage ledger at {}: {err}. Starting with an empty ledger.",
+                        usage_ledger_path.display()
+                    );
+                    UsageLedger::empty(usage_ledger_path)
+                }
+            };
+        let usage_ledger = Arc::new(TokioMutex::new(usage_ledger_instance));
+
         Ok(Self {
             provider,
             conversation,
@@ -568,6 +591,7 @@ impl SessionController {
             event_tx,
             pending_tool_requests: HashMap::new(),
             stream_states: HashMap::new(),
+            usage_ledger,
         })
     }
 
@@ -656,6 +680,134 @@ impl SessionController {
         Some((server.to_string(), uri.to_string()))
     }
 
+    async fn persist_usage_serialized(path: PathBuf, serialized: String) {
+        if let Some(parent) = path.parent() {
+            if let Err(err) = fs::create_dir_all(parent).await {
+                warn!(
+                    "Failed to create usage ledger directory {}: {}",
+                    parent.display(),
+                    err
+                );
+                return;
+            }
+        }
+
+        if let Err(err) = fs::write(&path, serialized).await {
+            warn!("Failed to write usage ledger {}: {}", path.display(), err);
+        }
+    }
+
+    fn parse_quota_value(value: &Value) -> Option<u64> {
+        match value {
+            Value::Number(num) => num.as_u64(),
+            Value::String(text) => text.trim().parse::<u64>().ok(),
+            _ => None,
+        }
+    }
+
+    fn quota_from_config(config: &Config, provider: &str) -> UsageQuota {
+        let mut quota = UsageQuota::default();
+
+        if let Some(entry) = config.providers.get(provider) {
+            if let Some(value) = entry.extra.get("hourly_quota_tokens") {
+                quota.hourly_quota_tokens = Self::parse_quota_value(value);
+            }
+            if let Some(value) = entry.extra.get("weekly_quota_tokens") {
+                quota.weekly_quota_tokens = Self::parse_quota_value(value);
+            }
+        }
+
+        quota
+    }
+
+    pub async fn record_usage_sample(
+        &self,
+        usage: &crate::types::TokenUsage,
+    ) -> Option<UsageSnapshot> {
+        if usage.total_tokens == 0 {
+            return None;
+        }
+
+        let provider_name = self.provider.name().to_string();
+        if provider_name.trim().is_empty() {
+            return None;
+        }
+
+        let quotas = {
+            let guard = self.config.lock().await;
+            Self::quota_from_config(&guard, &provider_name)
+        };
+
+        let timestamp = SystemTime::now();
+        let mut serialized_payload: Option<(PathBuf, String)> = None;
+
+        let snapshot = {
+            let mut ledger = self.usage_ledger.lock().await;
+            ledger.record(&provider_name, usage, timestamp);
+            let snapshot = ledger.snapshot(&provider_name, quotas, timestamp);
+            match ledger.serialize() {
+                Ok(payload) => {
+                    serialized_payload = Some((ledger.path().to_path_buf(), payload));
+                }
+                Err(err) => warn!("Failed to serialize usage ledger: {}", err),
+            }
+            snapshot
+        };
+
+        if let Some((path, payload)) = serialized_payload {
+            Self::persist_usage_serialized(path, payload).await;
+        }
+
+        Some(snapshot)
+    }
+
+    pub async fn current_usage_snapshot(&self) -> Option<UsageSnapshot> {
+        let provider_name = self.provider.name().to_string();
+        if provider_name.trim().is_empty() {
+            return None;
+        }
+
+        let quotas = {
+            let guard = self.config.lock().await;
+            Self::quota_from_config(&guard, &provider_name)
+        };
+
+        let now = SystemTime::now();
+        let ledger = self.usage_ledger.lock().await;
+        Some(ledger.snapshot(&provider_name, quotas, now))
+    }
+
+    pub async fn usage_overview(&self) -> Vec<UsageSnapshot> {
+        let quota_map = {
+            let guard = self.config.lock().await;
+            guard
+                .providers
+                .iter()
+                .map(|(name, _)| (name.clone(), Self::quota_from_config(&guard, name)))
+                .collect::<HashMap<_, _>>()
+        };
+
+        let now = SystemTime::now();
+        let mut provider_names: HashSet<String> = quota_map.keys().cloned().collect();
+
+        let snapshots = {
+            let ledger = self.usage_ledger.lock().await;
+            for key in ledger.provider_keys() {
+                provider_names.insert(key.clone());
+            }
+
+            provider_names
+                .into_iter()
+                .map(|provider| {
+                    let quota = quota_map.get(&provider).cloned().unwrap_or_default();
+                    ledger.snapshot(&provider, quota, now)
+                })
+                .collect::<Vec<_>>()
+        };
+
+        snapshots
+    }
+
     // Asynchronous access to the configuration (used internally).
     pub async fn config_async(&self) -> tokio::sync::MutexGuard<'_, Config> {
         self.config.lock().await
@@ -1305,6 +1457,9 @@ impl SessionController {
                             request.messages = self.conversation.active().messages.clone();
                             continue;
                         } else {
+                            if let Some(usage) = response.usage.as_ref() {
+                                let _ = self.record_usage_sample(usage).await;
+                            }
                             self.conversation.push_message(response.message.clone());
                             return Ok(SessionOutcome::Complete(response));
                         }

crates/owlen-core/src/usage.rs

@@ -0,0 +1,332 @@
+use crate::{Error, Result};
+use serde::{Deserialize, Serialize};
+use std::collections::{HashMap, VecDeque};
+use std::path::{Path, PathBuf};
+use std::time::{Duration, SystemTime, UNIX_EPOCH};
+use tokio::fs;
+
+const LEDGER_VERSION: u32 = 1;
+const SECONDS_PER_HOUR: i64 = 60 * 60;
+const SECONDS_PER_WEEK: i64 = 7 * 24 * 60 * 60;
+
+#[derive(Clone, Debug, Serialize, Deserialize)]
+struct UsageRecord {
+    timestamp: i64,
+    prompt_tokens: u32,
+    completion_tokens: u32,
+}
+
+#[derive(Serialize, Deserialize)]
+struct LedgerFile {
+    version: u32,
+    providers: HashMap<String, VecDeque<UsageRecord>>,
+}
+
+impl Default for LedgerFile {
+    fn default() -> Self {
+        Self {
+            version: LEDGER_VERSION,
+            providers: HashMap::new(),
+        }
+    }
+}
+
+#[derive(Clone, Debug, Default)]
+pub struct UsageLedger {
+    path: PathBuf,
+    providers: HashMap<String, VecDeque<UsageRecord>>,
+}
+
+#[derive(Clone, Debug, Default)]
+pub struct UsageQuota {
+    pub hourly_quota_tokens: Option<u64>,
+    pub weekly_quota_tokens: Option<u64>,
+}
+
+#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash)]
+pub enum UsageWindow {
+    Hour,
+    Week,
+}
+
+#[derive(Clone, Copy, Debug, PartialEq, Eq, PartialOrd, Ord)]
+pub enum UsageBand {
+    Normal = 0,
+    Warning = 1,
+    Critical = 2,
+}
+
+#[derive(Clone, Debug, Default)]
+pub struct WindowMetrics {
+    pub prompt_tokens: u64,
+    pub completion_tokens: u64,
+    pub total_tokens: u64,
+    pub quota_tokens: Option<u64>,
+}
+
+impl WindowMetrics {
+    pub fn percent_of_quota(&self) -> Option<f64> {
+        let quota = self.quota_tokens?;
+        if quota == 0 {
+            return None;
+        }
+        Some(self.total_tokens as f64 / quota as f64)
+    }
+
+    pub fn band(&self) -> UsageBand {
+        match self.percent_of_quota() {
+            Some(p) if p >= 0.95_f64 => UsageBand::Critical,
+            Some(p) if p >= 0.80_f64 => UsageBand::Warning,
+            _ => UsageBand::Normal,
+        }
+    }
+}
+
+#[derive(Clone, Debug, Default)]
+pub struct UsageSnapshot {
+    pub provider: String,
+    pub hourly: WindowMetrics,
+    pub weekly: WindowMetrics,
+    pub last_updated: Option<SystemTime>,
+}
+
+impl UsageSnapshot {
+    pub fn window(&self, window: UsageWindow) -> &WindowMetrics {
+        match window {
+            UsageWindow::Hour => &self.hourly,
+            UsageWindow::Week => &self.weekly,
+        }
+    }
+}
+
+impl UsageLedger {
+    pub fn empty(path: PathBuf) -> Self {
+        Self {
+            path,
+            providers: HashMap::new(),
+        }
+    }
+
+    pub async fn load_or_default(path: PathBuf) -> Result<Self> {
+        if !path.exists() {
+            return Ok(Self {
+                path,
+                providers: HashMap::new(),
+            });
+        }
+
+        let contents = fs::read_to_string(&path)
+            .await
+            .map_err(|err| Error::Storage(format!("Failed to read usage ledger: {err}")))?;
+
+        let file: LedgerFile = match serde_json::from_str(&contents) {
+            Ok(file) => file,
+            Err(err) => {
+                return Err(Error::Storage(format!(
+                    "Failed to parse usage ledger at {}: {err}",
+                    path.display()
+                )));
+            }
+        };
+
+        Ok(Self {
+            path,
+            providers: file.providers,
+        })
+    }
+
+    pub async fn persist(&self) -> Result<()> {
+        if let Some(parent) = self.path.parent() {
+            fs::create_dir_all(parent)
+                .await
+                .map_err(|err| Error::Storage(format!("Failed to create data directory: {err}")))?;
+        }
+
+        let serialized = self.serialize()?;
+
+        fs::write(&self.path, serialized)
+            .await
+            .map_err(|err| Error::Storage(format!("Failed to write usage ledger: {err}")))?;
+
+        Ok(())
+    }
+
+    pub fn record(
+        &mut self,
+        provider: &str,
+        usage: &crate::types::TokenUsage,
+        timestamp: SystemTime,
+    ) {
+        let total_tokens = usage.total_tokens;
+        if total_tokens == 0 {
+            return;
+        }
+
+        let ts = match timestamp.duration_since(UNIX_EPOCH) {
+            Ok(duration) => duration.as_secs() as i64,
+            Err(_) => 0,
+        };
+
+        let entry = self
+            .providers
+            .entry(provider.to_string())
+            .or_insert_with(VecDeque::new);
+
+        entry.push_back(UsageRecord {
+            timestamp: ts,
+            prompt_tokens: usage.prompt_tokens,
+            completion_tokens: usage.completion_tokens,
+        });
+
+        self.prune_old(provider, ts);
+    }
+
+    pub fn provider_keys(&self) -> impl Iterator<Item = &String> {
+        self.providers.keys()
+    }
+
+    pub fn serialize(&self) -> Result<String> {
+        let file = LedgerFile {
+            version: LEDGER_VERSION,
+            providers: self.providers.clone(),
+        };
+
+        serde_json::to_string_pretty(&file)
+            .map_err(|err| Error::Storage(format!("Failed to serialize usage ledger: {err}")))
+    }
+
+    pub fn path(&self) -> &Path {
+        &self.path
+    }
+
+    pub fn snapshot(&self, provider: &str, quotas: UsageQuota, now: SystemTime) -> UsageSnapshot {
+        let now_secs = now
+            .duration_since(UNIX_EPOCH)
+            .unwrap_or_else(|_| Duration::from_secs(0))
+            .as_secs() as i64;
+
+        let mut snapshot = UsageSnapshot {
+            provider: provider.to_string(),
+            hourly: WindowMetrics {
+                quota_tokens: quotas.hourly_quota_tokens,
+                ..Default::default()
+            },
+            weekly: WindowMetrics {
+                quota_tokens: quotas.weekly_quota_tokens,
+                ..Default::default()
+            },
+            last_updated: None,
+        };
+
+        if let Some(records) = self.providers.get(provider) {
+            for record in records {
+                if now_secs - record.timestamp <= SECONDS_PER_HOUR {
+                    snapshot.hourly.prompt_tokens += record.prompt_tokens as u64;
+                    snapshot.hourly.completion_tokens += record.completion_tokens as u64;
+                }
+
+                if now_secs - record.timestamp <= SECONDS_PER_WEEK {
+                    snapshot.weekly.prompt_tokens += record.prompt_tokens as u64;
+                    snapshot.weekly.completion_tokens += record.completion_tokens as u64;
+                }
+            }
+
+            snapshot.hourly.total_tokens =
+                snapshot.hourly.prompt_tokens + snapshot.hourly.completion_tokens;
+            snapshot.weekly.total_tokens =
+                snapshot.weekly.prompt_tokens + snapshot.weekly.completion_tokens;
+
+            snapshot.last_updated = records.back().and_then(|record| {
+                UNIX_EPOCH.checked_add(Duration::from_secs(record.timestamp as u64))
+            });
+        }
+
+        snapshot
+    }
+
+    pub fn prune_old(&mut self, provider: &str, now_secs: i64) {
+        if let Some(records) = self.providers.get_mut(provider) {
+            while let Some(front) = records.front() {
+                if now_secs - front.timestamp > SECONDS_PER_WEEK {
+                    records.pop_front();
+                } else {
+                    break;
+                }
+            }
+        }
+    }
+
+    pub fn prune_all(&mut self, now: SystemTime) {
+        let now_secs = now
+            .duration_since(UNIX_EPOCH)
+            .unwrap_or_else(|_| Duration::from_secs(0))
+            .as_secs() as i64;
+        let provider_keys: Vec<String> = self.providers.keys().cloned().collect();
+        for provider in provider_keys {
+            self.prune_old(&provider, now_secs);
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::types::TokenUsage;
+    use std::time::{Duration, UNIX_EPOCH};
+    use tempfile::tempdir;
+
+    fn make_usage(prompt: u32, completion: u32) -> TokenUsage {
+        TokenUsage {
+            prompt_tokens: prompt,
+            completion_tokens: completion,
+            total_tokens: prompt.saturating_add(completion),
+        }
+    }
+
+    #[test]
+    fn records_and_summarizes_usage() {
+        let temp = tempdir().expect("tempdir");
+        let path = temp.path().join("ledger.json");
+        let mut ledger = UsageLedger::empty(path);
+
+        let usage = make_usage(40, 10);
+        let timestamp = UNIX_EPOCH + Duration::from_secs(1);
+        ledger.record("ollama_cloud", &usage, timestamp);
+
+        let quotas = UsageQuota {
+            hourly_quota_tokens: Some(100),
+            weekly_quota_tokens: Some(1000),
+        };
+
+        let snapshot = ledger.snapshot("ollama_cloud", quotas, UNIX_EPOCH + Duration::from_secs(2));
+
+        assert_eq!(snapshot.hourly.total_tokens, 50);
+        assert_eq!(snapshot.weekly.total_tokens, 50);
+        assert_eq!(snapshot.hourly.quota_tokens, Some(100));
+        assert_eq!(snapshot.weekly.quota_tokens, Some(1000));
+        assert_eq!(snapshot.hourly.band(), UsageBand::Normal);
+    }
+
+    #[test]
+    fn prunes_records_outside_week() {
+        let temp = tempdir().expect("tempdir");
+        let path = temp.path().join("ledger.json");
+        let mut ledger = UsageLedger::empty(path);
+
+        let old_usage = make_usage(30, 5);
+        let recent_usage = make_usage(20, 5);
+
+        let base = UNIX_EPOCH;
+        ledger.record("ollama_cloud", &old_usage, base);
+
+        // Advance beyond a week for the second record.
+        let later = UNIX_EPOCH + Duration::from_secs(SECONDS_PER_WEEK as u64 + 120);
+        ledger.record("ollama_cloud", &recent_usage, later);
+
+        let quotas = UsageQuota::default();
+        let snapshot = ledger.snapshot("ollama_cloud", quotas, later);
+
+        assert_eq!(snapshot.hourly.total_tokens, 25);
+        assert_eq!(snapshot.weekly.total_tokens, 25);
+    }
+}

crates/owlen-tui/src/chat_app.rs

@@ -22,8 +22,9 @@ use owlen_core::{
     session::{ControllerEvent, SessionController, SessionOutcome, ToolConsentResolution},
     storage::SessionMeta,
     theme::Theme,
-    types::{ChatParameters, ChatResponse, Conversation, ModelInfo, Role},
+    types::{ChatParameters, ChatResponse, Conversation, ModelInfo, Role, TokenUsage},
     ui::{AppState, AutoScroll, FocusedPanel, InputMode, RoleLabelDisplay},
+    usage::{UsageBand, UsageSnapshot, UsageWindow, WindowMetrics},
 };
 use owlen_markdown::from_str;
 use pathdiff::diff_paths;
@@ -59,7 +60,7 @@ use crate::state::{
     spawn_repo_search_task, spawn_symbol_search_task,
 };
 use crate::toast::{Toast, ToastLevel, ToastManager};
-use crate::ui::format_tool_output;
+use crate::ui::{format_token_short, format_tool_output};
 use crate::widgets::model_picker::FilterMode;
 use crate::{commands, highlight};
 use owlen_core::config::{
@@ -101,6 +102,14 @@ const RESIZE_SNAP_VALUES: [f32; 3] = [0.5, 0.75, 0.25];
 const DOUBLE_CTRL_C_WINDOW: Duration = Duration::from_millis(1500);
 pub(crate) const MIN_MESSAGE_CARD_WIDTH: usize = 14;
 const MOUSE_SCROLL_STEP: isize = 3;
+const DEFAULT_CONTEXT_WINDOW_TOKENS: u32 = 8_192;
+
+#[derive(Clone, Copy, Debug, Default)]
+pub struct ContextUsage {
+    pub prompt_tokens: u32,
+    pub completion_tokens: u32,
+    pub context_window: u32,
+}
 
 #[derive(Clone, Copy, Debug)]
 pub(crate) struct LayoutSnapshot {
@@ -606,6 +615,9 @@ pub struct ChatApp {
     system_status: String,        // System/status messages (tool execution, status, etc)
     toasts: ToastManager,
     debug_log: DebugLogState,
+    usage_snapshot: Option<UsageSnapshot>,
+    usage_thresholds: HashMap<(String, UsageWindow), UsageBand>,
+    context_usage: Option<ContextUsage>,
     last_layout: LayoutSnapshot,
     /// Simple execution budget: maximum number of tool calls allowed per session.
     _execution_budget: usize,
@@ -880,6 +892,9 @@ impl ChatApp {
             },
             toasts: ToastManager::new(),
             debug_log: DebugLogState::new(),
+            usage_snapshot: None,
+            usage_thresholds: HashMap::new(),
+            context_usage: None,
             last_layout: LayoutSnapshot::default(),
             _execution_budget: 50,
             agent_mode: false,
@@ -919,6 +934,8 @@ impl ChatApp {
             }
         }
 
+        app.refresh_usage_summary().await?;
+
         Ok((app, session_rx))
     }
 
@@ -1039,6 +1056,45 @@ impl ChatApp {
         &self.current_provider
     }
 
+    pub fn usage_snapshot(&self) -> Option<&UsageSnapshot> {
+        self.usage_snapshot.as_ref()
+    }
+
+    pub fn context_usage_with_fallback(&self) -> Option<ContextUsage> {
+        if let Some(usage) = self.context_usage {
+            Some(usage)
+        } else {
+            self.active_context_window().map(|window| ContextUsage {
+                prompt_tokens: 0,
+                completion_tokens: 0,
+                context_window: window,
+            })
+        }
+    }
+
+    fn update_context_usage(&mut self, usage: &TokenUsage) {
+        let context_window = self
+            .active_context_window()
+            .unwrap_or(DEFAULT_CONTEXT_WINDOW_TOKENS);
+        self.context_usage = Some(ContextUsage {
+            prompt_tokens: usage.prompt_tokens,
+            completion_tokens: usage.completion_tokens,
+            context_window,
+        });
+    }
+
+    fn active_context_window(&self) -> Option<u32> {
+        let current_model = self.controller.selected_model();
+
+        self.models.iter().find_map(|model| {
+            if model.id == current_model || model.name == current_model {
+                model.context_window
+            } else {
+                None
+            }
+        })
+    }
+
     pub fn should_show_code_view(&self) -> bool {
         if !matches!(self.operating_mode, owlen_core::mode::Mode::Code) {
             return false;
@@ -6968,6 +7024,8 @@ impl ChatApp {
                                                                 "Provider switched; failed to refresh models"
                                                                     .to_string();
                                                         }
+                                                        self.context_usage = None;
+                                                        self.refresh_usage_summary().await?;
                                                     }
                                                     Err(err) => {
                                                         self.error = Some(format!(
@@ -6994,6 +7052,12 @@ impl ChatApp {
                                     self.command_palette.clear();
                                     return Ok(AppState::Running);
                                 }
+                                "limits" => {
+                                    self.show_usage_limits().await?;
+                                    self.set_input_mode(InputMode::Normal);
+                                    self.command_palette.clear();
+                                    return Ok(AppState::Running);
+                                }
                                 "models" => {
                                     if args.is_empty() {
                                         if let Err(err) = self.show_model_picker(None).await {
@@ -8219,6 +8283,20 @@ impl ChatApp {
 
                 // Auto-scroll will handle this in the render loop
                 if response.is_final {
+                    let recorded_snapshot = match response.usage.as_ref() {
+                        Some(usage) => {
+                            self.update_context_usage(usage);
+                            self.controller.record_usage_sample(usage).await
+                        }
+                        None => None,
+                    };
+                    if let Some(snapshot) = recorded_snapshot {
+                        self.usage_snapshot = Some(snapshot.clone());
+                        self.update_usage_toasts(&snapshot);
+                    } else {
+                        self.refresh_usage_summary().await?;
+                    }
+
                     self.streaming.remove(&message_id);
                     self.stream_tasks.remove(&message_id);
                     self.stop_loading_animation();
@@ -8338,6 +8416,134 @@ impl ChatApp {
         self.error = None;
     }
 
+    async fn refresh_usage_summary(&mut self) -> Result<()> {
+        if let Some(snapshot) = self.controller.current_usage_snapshot().await {
+            self.usage_snapshot = Some(snapshot.clone());
+            self.update_usage_toasts(&snapshot);
+        } else {
+            self.usage_snapshot = None;
+        }
+        Ok(())
+    }
+
+    fn update_usage_toasts(&mut self, snapshot: &UsageSnapshot) {
+        for window in [UsageWindow::Hour, UsageWindow::Week] {
+            let key = (snapshot.provider.clone(), window);
+            let metrics = snapshot.window(window);
+            let quota = match metrics.quota_tokens {
+                Some(value) if value > 0 => value,
+                _ => {
+                    self.usage_thresholds.remove(&key);
+                    continue;
+                }
+            };
+
+            let previous = self
+                .usage_thresholds
+                .get(&key)
+                .copied()
+                .unwrap_or(UsageBand::Normal);
+            let current = metrics.band();
+
+            if current > previous {
+                if let Some(percent_ratio) = metrics.percent_of_quota() {
+                    let percent_value = percent_ratio * 100.0;
+                    let percent_text = Self::format_percent_value(percent_value.min(999.9));
+                    let quota_text = format_token_short(quota);
+                    let used_text = format_token_short(metrics.total_tokens);
+                    let provider_display = Self::provider_display_name(&snapshot.provider);
+                    let window_label = Self::usage_window_label(window);
+                    let message = format!(
+                        "{} {} usage at {}% ({}/{})",
+                        provider_display, window_label, percent_text, used_text, quota_text
+                    );
+                    let level = if current == UsageBand::Critical {
+                        ToastLevel::Error
+                    } else {
+                        ToastLevel::Warning
+                    };
+                    self.push_toast(level, message);
+                }
+            } else if current == UsageBand::Normal && previous != UsageBand::Normal {
+                self.usage_thresholds.insert(key.clone(), UsageBand::Normal);
+            }
+
+            self.usage_thresholds.insert(key, current);
+        }
+    }
+
+    async fn show_usage_limits(&mut self) -> Result<()> {
+        let snapshots = self.controller.usage_overview().await;
+        if snapshots.is_empty() {
+            let message = "Usage: no data recorded yet.".to_string();
+            self.status = message.clone();
+            self.error = None;
+            self.push_toast(ToastLevel::Info, message);
+            return Ok(());
+        }
+
+        let mut parts = Vec::new();
+        let mut current_snapshot: Option<UsageSnapshot> = None;
+
+        for snapshot in snapshots.iter() {
+            if snapshot.provider == self.current_provider {
+                current_snapshot = Some(snapshot.clone());
+            }
+            self.update_usage_toasts(snapshot);
+
+            let provider_display = Self::provider_display_name(&snapshot.provider);
+            let hour = Self::summarize_usage_window("hour", snapshot.window(UsageWindow::Hour));
+            let week = Self::summarize_usage_window("week", snapshot.window(UsageWindow::Week));
+            parts.push(format!("{provider_display}: {hour}; {week}"));
+        }
+
+        if let Some(snapshot) = current_snapshot {
+            self.usage_snapshot = Some(snapshot);
+        }
+
+        let message = parts.join(" | ");
+        self.status = format!("Usage • {message}");
+        self.error = None;
+        self.push_toast(ToastLevel::Info, message.clone());
+
+        Ok(())
+    }
+
+    fn summarize_usage_window(label: &str, metrics: &WindowMetrics) -> String {
+        let used = format_token_short(metrics.total_tokens);
+        if let Some(quota) = metrics.quota_tokens {
+            if quota == 0 {
+                return format!("{label} {used} tokens");
+            }
+            let quota_text = format_token_short(quota);
+            let percent = metrics
+                .percent_of_quota()
+                .map(|ratio| ratio * 100.0)
+                .unwrap_or(0.0);
+            let percent_text = Self::format_percent_value(percent.min(999.9));
+            format!("{label} {used}/{quota_text} ({percent_text}%)")
+        } else {
+            format!("{label} {used} tokens")
+        }
+    }
+
+    fn usage_window_label(window: UsageWindow) -> &'static str {
+        match window {
+            UsageWindow::Hour => "hourly",
+            UsageWindow::Week => "weekly",
+        }
+    }
+
+    fn format_percent_value(percent: f64) -> String {
+        if percent >= 100.0 || percent == 0.0 {
+            format!("{percent:.0}")
+        } else if percent >= 10.0 {
+            format!("{percent:.0}")
+        } else {
+            format!("{percent:.1}")
+        }
+    }
+
     async fn collect_models_from_all_providers(
         &self,
     ) -> (
@@ -10178,10 +10384,14 @@ impl ChatApp {
         let timeout_duration = std::time::Duration::from_secs(30);
 
         match tokio::time::timeout(timeout_duration, request_future).await {
-            Ok(Ok(SessionOutcome::Complete(_response))) => {
+            Ok(Ok(SessionOutcome::Complete(response))) => {
+                if let Some(usage) = response.usage.as_ref() {
+                    self.update_context_usage(usage);
+                }
                 self.stop_loading_animation();
                 self.status = "Ready".to_string();
                 self.error = None;
+                self.refresh_usage_summary().await?;
                 Ok(())
             }
             Ok(Ok(SessionOutcome::Streaming {

crates/owlen-tui/src/ui.rs

@@ -13,7 +13,8 @@ use unicode_segmentation::UnicodeSegmentation;
 use unicode_width::UnicodeWidthStr;
 
 use crate::chat_app::{
-    ChatApp, HELP_TAB_COUNT, LayoutSnapshot, MIN_MESSAGE_CARD_WIDTH, MessageRenderContext,
+    ChatApp, ContextUsage, HELP_TAB_COUNT, LayoutSnapshot, MIN_MESSAGE_CARD_WIDTH,
+    MessageRenderContext,
 };
 use crate::highlight;
 use crate::state::{
@@ -25,6 +26,7 @@ use crate::widgets::model_picker::render_model_picker;
 use owlen_core::theme::Theme;
 use owlen_core::types::Role;
 use owlen_core::ui::{FocusedPanel, InputMode, RoleLabelDisplay};
+use owlen_core::usage::{UsageBand, UsageSnapshot, UsageWindow, WindowMetrics};
 use textwrap::wrap;
 
 const PRIVACY_TAB_INDEX: usize = HELP_TAB_COUNT - 1;
@@ -183,6 +185,75 @@ mod focus_tests {
     }
 }
 
+#[cfg(test)]
+mod context_usage_tests {
+    use super::*;
+    use ratatui::style::{Color, Modifier};
+
+    fn theme() -> Theme {
+        Theme::default()
+    }
+
+    #[test]
+    fn context_badge_formats_label_and_highlights() {
+        let theme = theme();
+        let usage = ContextUsage {
+            prompt_tokens: 2600,
+            completion_tokens: 0,
+            total_tokens: 2600,
+            context_window: 8000,
+        };
+
+        let (label, style) = context_usage_badge(usage, &theme).expect("badge should render");
+        assert_eq!(label, "Context 2.6k / 8k (33%)");
+        assert_eq!(style.fg, Some(theme.info));
+        assert!(style.add_modifier.contains(Modifier::BOLD));
+    }
+
+    #[test]
+    fn context_badge_warns_near_limits() {
+        let theme = theme();
+        let usage = ContextUsage {
+            prompt_tokens: 7000,
+            completion_tokens: 0,
+            total_tokens: 7000,
+            context_window: 10000,
+        };
+
+        let (_, style) = context_usage_badge(usage, &theme).expect("badge should render");
+        assert_eq!(style.fg, Some(Color::Yellow));
+    }
+
+    #[test]
+    fn context_badge_flags_danger_zone() {
+        let theme = theme();
+        let usage = ContextUsage {
+            prompt_tokens: 9000,
+            completion_tokens: 0,
+            total_tokens: 9000,
+            context_window: 10000,
+        };
+
+        let (_, style) = context_usage_badge(usage, &theme).expect("badge should render");
+        assert_eq!(style.fg, Some(theme.error));
+    }
+
+    #[test]
+    fn context_badge_handles_zero_usage() {
+        let theme = theme();
+        let usage = ContextUsage {
+            prompt_tokens: 0,
+            completion_tokens: 0,
+            total_tokens: 0,
+            context_window: 32000,
+        };
+
+        let (label, style) = context_usage_badge(usage, &theme).expect("badge should render");
+        assert_eq!(label, "Context 0 / 32k (0%)");
+        assert_eq!(style.fg, Some(theme.info));
+    }
+}
+
 pub fn render_chat(frame: &mut Frame<'_>, app: &mut ChatApp) {
     // Update thinking content from last message
     app.update_thinking_from_last_message();
@@ -1501,6 +1572,24 @@ fn render_messages(frame: &mut Frame<'_>, area: Rect, app: &mut ChatApp) {
     let scroll_position = app.scroll().min(u16::MAX as usize) as u16;
 
     let mut title_spans = panel_title_spans("Chat", true, has_focus, &theme);
+
+    let active_model = app.active_model_label();
+    if !active_model.is_empty() {
+        let model_display = truncate_with_ellipsis(&active_model, 28);
+        title_spans.push(Span::raw(" · "));
+        title_spans.push(Span::styled(
+            model_display,
+            Style::default().fg(theme.pane_header_active),
+        ));
+    }
+
+    if let Some(usage) = app.context_usage_with_fallback() {
+        if let Some((label, style)) = context_usage_badge(usage, &theme) {
+            title_spans.push(Span::raw(" · "));
+            title_spans.push(Span::styled(label, style));
+        }
+    }
+
     title_spans.push(Span::raw("  "));
     title_spans.push(Span::styled(
         "PgUp/PgDn scroll · g/G jump · s save · Ctrl+2 focus",
@@ -2166,9 +2255,9 @@ fn render_status(frame: &mut Frame<'_>, area: Rect, app: &ChatApp) {
     let columns = Layout::default()
         .direction(Direction::Horizontal)
         .constraints([
-            Constraint::Percentage(30),
-            Constraint::Percentage(40),
-            Constraint::Percentage(30),
+            Constraint::Length(32),
+            Constraint::Min(24),
+            Constraint::Length(48),
         ])
         .split(inner);
 
@@ -2294,9 +2383,11 @@ fn render_status(frame: &mut Frame<'_>, area: Rect, app: &ChatApp) {
     frame.render_widget(mid_paragraph, columns[1]);
 
     let provider = app.current_provider();
+    let provider_display = truncate_with_ellipsis(provider, 16);
     let model_label = app.active_model_label();
+    let model_display = truncate_with_ellipsis(&model_label, 24);
     let mut right_spans = vec![Span::styled(
-        format!("{} ▸ {}", provider, model_label),
+        format!("{} ▸ {}", provider_display, model_display),
         Style::default().fg(theme.text).add_modifier(Modifier::BOLD),
     )];
 
@@ -2315,6 +2406,18 @@ fn render_status(frame: &mut Frame<'_>, area: Rect, app: &ChatApp) {
         ));
     }
 
+    if let Some(usage) = app.context_usage_with_fallback() {
+        if let Some((label, style)) = context_usage_badge(usage, theme) {
+            right_spans.push(Span::styled(format!(" · {}", label), style));
+        }
+    }
+
+    if let Some(snapshot) = app.usage_snapshot() {
+        for span in usage_badge_spans(snapshot, theme) {
+            right_spans.push(span);
+        }
+    }
+
     right_spans.push(Span::styled(
         " · LSP:✓",
         Style::default()
@@ -2322,12 +2425,219 @@ fn render_status(frame: &mut Frame<'_>, area: Rect, app: &ChatApp) {
             .add_modifier(Modifier::DIM),
     ));
 
-    let right_paragraph = Paragraph::new(Line::from(right_spans))
+    let right_line = spans_within_width(right_spans, columns[2].width);
+    let right_paragraph = Paragraph::new(right_line)
         .alignment(Alignment::Right)
         .style(Style::default().bg(theme.status_background).fg(theme.text));
     frame.render_widget(right_paragraph, columns[2]);
 }
 
+fn context_usage_badge(usage: ContextUsage, theme: &Theme) -> Option<(String, Style)> {
+    let context_window = usage.context_window.max(1);
+    let ratio = usage.prompt_tokens as f64 / context_window as f64;
+    let percent = ((ratio * 100.0).round() as u32).min(100);
+    let used = format_token_short(usage.prompt_tokens as u64);
+    let window = format_token_short(context_window as u64);
+    let label = format!("Context {} / {} ({}%)", used, window, percent);
+
+    let style = if ratio < 0.60 {
+        Style::default().fg(theme.info).add_modifier(Modifier::BOLD)
+    } else if ratio < 0.85 {
+        Style::default()
+            .fg(Color::Yellow)
+            .add_modifier(Modifier::BOLD)
+    } else {
+        Style::default()
+            .fg(theme.error)
+            .add_modifier(Modifier::BOLD)
+    };
+
+    Some((label, style))
+}
+
+fn usage_badge_spans(snapshot: &UsageSnapshot, theme: &Theme) -> Vec<Span<'static>> {
+    let mut spans = Vec::new();
+    let provider_label = ChatApp::provider_display_name(&snapshot.provider);
+
+    let hour_metrics = snapshot.window(UsageWindow::Hour);
+    let week_metrics = snapshot.window(UsageWindow::Week);
+
+    if hour_metrics.quota_tokens.is_none()
+        && hour_metrics.total_tokens == 0
+        && week_metrics.quota_tokens.is_none()
+        && week_metrics.total_tokens == 0
+    {
+        return spans;
+    }
+
+    spans.push(Span::styled(
+        format!(" · {} usage", provider_label),
+        Style::default()
+            .fg(theme.placeholder)
+            .add_modifier(Modifier::DIM),
+    ));
+
+    if let Some(span) = usage_window_span("hr", hour_metrics, theme) {
+        spans.push(span);
+    }
+    if let Some(span) = usage_window_span("wk", week_metrics, theme) {
+        spans.push(span);
+    }
+
+    spans
+}
+
+fn usage_window_span(label: &str, metrics: &WindowMetrics, theme: &Theme) -> Option<Span<'static>> {
+    if metrics.quota_tokens.is_none() && metrics.total_tokens == 0 {
+        return None;
+    }
+
+    let used = format_token_short(metrics.total_tokens);
+    let text = if let Some(quota) = metrics.quota_tokens {
+        if quota == 0 {
+            format!(" · {} {} tokens", label, used)
+        } else if let Some(percent_ratio) = metrics.percent_of_quota() {
+            let quota_text = format_token_short(quota);
+            let percent_text = format_percent_short((percent_ratio * 100.0).min(999.9));
+            format!(" · {} {}/{} ({}%)", label, used, quota_text, percent_text)
+        } else {
+            let quota_text = format_token_short(quota);
+            format!(" · {} {}/{}", label, used, quota_text)
+        }
+    } else {
+        format!(" · {} {} tokens", label, used)
+    };
+
+    let style = match metrics.band() {
+        UsageBand::Normal => Style::default().fg(theme.info).add_modifier(Modifier::BOLD),
+        UsageBand::Warning => Style::default()
+            .fg(Color::Yellow)
+            .add_modifier(Modifier::BOLD),
+        UsageBand::Critical => Style::default()
+            .fg(theme.error)
+            .add_modifier(Modifier::BOLD),
+    };
+
+    Some(Span::styled(text, style))
+}
+
+fn format_percent_short(percent: f64) -> String {
+    let clamped = percent.max(0.0);
+    if clamped >= 100.0 || clamped == 0.0 {
+        format!("{clamped:.0}")
+    } else if clamped >= 10.0 {
+        format!("{clamped:.0}")
+    } else {
+        format!("{clamped:.1}")
+    }
+}
+
+pub(crate) fn format_token_short(value: u64) -> String {
+    if value >= 1_000_000_000 {
+        format_compact(value as f64 / 1_000_000_000.0, "B")
+    } else if value >= 1_000_000 {
+        format_compact(value as f64 / 1_000_000.0, "M")
+    } else if value >= 1_000 {
+        format_compact(value as f64 / 1_000.0, "k")
+    } else {
+        value.to_string()
+    }
+}
+
+fn format_compact(value: f64, suffix: &str) -> String {
+    let formatted = if value >= 100.0 {
+        format!("{:.0}", value)
+    } else {
+        format!("{:.1}", value)
+    };
+
+    let trimmed = formatted.trim_end_matches('0').trim_end_matches('.');
+    let mut result = trimmed.to_string();
+    if result.is_empty() {
+        result.push('0');
+    }
+    result.push_str(suffix);
+    result
+}
+
+fn truncate_with_ellipsis(text: &str, max_width: usize) -> String {
+    if max_width == 0 {
+        return String::new();
+    }
+
+    let current = UnicodeWidthStr::width(text);
+    if current <= max_width {
+        return text.to_string();
+    }
+
+    let ellipsis = "…";
+    let ellipsis_width = UnicodeWidthStr::width(ellipsis);
+    if ellipsis_width >= max_width {
+        return ellipsis.to_string();
+    }
+
+    let keep_width = max_width - ellipsis_width;
+    let prefix = truncate_to_width(text, keep_width);
+    if prefix.is_empty() {
+        ellipsis.to_string()
+    } else {
+        format!("{}{}", prefix, ellipsis)
+    }
+}
+
+fn spans_within_width(spans: Vec<Span<'static>>, max_width: u16) -> Line<'static> {
+    if max_width == 0 {
+        return Line::from(Vec::<Span<'static>>::new());
+    }
+
+    let mut remaining = max_width as usize;
+    let mut output = Vec::new();
+
+    for span in spans.into_iter() {
+        if remaining == 0 {
+            break;
+        }
+
+        let text = span.content.as_ref();
+        let width = UnicodeWidthStr::width(text);
+        if width == 0 {
+            continue;
+        }
+
+        let style = span.style;
+        if width <= remaining {
+            output.push(Span::styled(text.to_string(), style));
+            remaining -= width;
+        } else {
+            let truncated = truncate_to_width(text, remaining);
+            if !truncated.is_empty() {
+                output.push(Span::styled(truncated, style));
+            }
+            break;
+        }
+    }
+
+    Line::from(output)
+}
+
+fn truncate_to_width(text: &str, max_width: usize) -> String {
+    if max_width == 0 {
+        return String::new();
+    }
+
+    let mut result = String::new();
+    let mut used = 0;
+    for grapheme in text.graphemes(true) {
+        let width = UnicodeWidthStr::width(grapheme);
+        if used + width > max_width {
+            break;
+        }
+        result.push_str(grapheme);
+        used += width;
+    }
+    result
+}
+
 fn render_code_workspace(frame: &mut Frame<'_>, area: Rect, app: &mut ChatApp) {
     let theme = app.theme().clone();
     frame.render_widget(Clear, area);

docs/configuration.md

@@ -177,10 +177,14 @@ base_url = "http://localhost:11434"
 enabled = true
 base_url = "https://ollama.com"
 api_key_env = "OLLAMA_API_KEY"
+hourly_quota_tokens = 50000
+weekly_quota_tokens = 250000
 ```
 
 Requests target the same `/api/chat` endpoint documented by Ollama and automatically include the API key using a `Bearer` authorization header. If you prefer not to store the key in the config file, either rely on `api_key_env` or export the `OLLAMA_API_KEY` environment variable manually (legacy names `OLLAMA_CLOUD_API_KEY` and `OWLEN_OLLAMA_CLOUD_API_KEY` continue to work, but Owlen will emit a warning). Owlen normalises the base URL automatically—it enforces HTTPS, trims trailing slashes, and accepts both `https://ollama.com` and `https://api.ollama.com` without rewriting the host.
 
+The quota fields are optional and purely informational—they are never sent to the provider. Owlen uses them to display hourly/weekly token usage in the chat header, emit pre-limit toasts at 80 % and 95 %, and power the `:limits` command. Adjust the numbers to reflect the soft limits on your account or remove the keys altogether if you do not want usage tracking.
+
 > **Tip:** If the official `ollama signin` flow fails on Linux v0.12.3, follow the [Linux Ollama sign-in workaround](#linux-ollama-sign-in-workaround-v0123) in the troubleshooting guide to copy keys from a working machine or register them manually.
 
 ### Managing cloud credentials via CLI