feat: massive overhaul to manage costs (#103)

### Explanation

_No response_

### Issue

Closes #102

### Attestations

- [ ] I have read and agree to the [Code of Conduct](https://docs.nhcarrigan.com/community/coc/)
- [ ] I have read and agree to the [Community Guidelines](https://docs.nhcarrigan.com/community/guide/).
- [ ] My contribution complies with the [Contributor Covenant](https://docs.nhcarrigan.com/dev/covenant/).

### Dependencies

- [ ] I have pinned the dependencies to a specific patch version.

### Style

- [ ] I have run the linter and resolved any errors.
- [ ] My pull request uses an appropriate title, matching the conventional commit standards.
- [ ] My scope of feat/fix/chore/etc. correctly matches the nature of changes in my pull request.

### Tests

- [ ] My contribution adds new code, and I have added tests to cover it.
- [ ] My contribution modifies existing code, and I have updated the tests to reflect these changes.
- [ ] All new and existing tests pass locally with my changes.
- [ ] Code coverage remains at or above the configured threshold.

### Documentation

_No response_

### Versioning

_No response_

Reviewed-on: #103
Co-authored-by: Naomi Carrigan <commits@nhcarrigan.com>
Co-committed-by: Naomi Carrigan <commits@nhcarrigan.com>

src/lib/stores/config.test.ts

@@ -187,6 +187,11 @@ describe("config store", () => {
           text_secondary: null,
           border_color: null,
         },
+        budget_enabled: false,
+        session_token_budget: null,
+        session_cost_budget: null,
+        budget_action: "warn",
+        budget_warning_threshold: 0.8,
       };
 
       expect(config.model).toBe("claude-sonnet-4");
@@ -227,6 +232,11 @@ describe("config store", () => {
           text_secondary: null,
           border_color: null,
         },
+        budget_enabled: false,
+        session_token_budget: null,
+        session_cost_budget: null,
+        budget_action: "warn",
+        budget_warning_threshold: 0.8,
       };
 
       expect(config.model).toBeNull();

src/lib/stores/config.ts

@@ -2,6 +2,7 @@ import { writable, derived } from "svelte/store";
 import { invoke } from "@tauri-apps/api/core";
 
 export type Theme = "dark" | "light" | "high-contrast" | "custom";
+export type BudgetAction = "warn" | "block";
 
 export interface CustomThemeColors {
   bg_primary: string | null;
@@ -37,6 +38,12 @@ export interface HikariConfig {
   profile_avatar_path: string | null;
   profile_bio: string | null;
   custom_theme_colors: CustomThemeColors;
+  // Budget settings
+  budget_enabled: boolean;
+  session_token_budget: number | null;
+  session_cost_budget: number | null;
+  budget_action: BudgetAction;
+  budget_warning_threshold: number;
 }
 
 const defaultConfig: HikariConfig = {
@@ -71,6 +78,11 @@ const defaultConfig: HikariConfig = {
     text_secondary: null,
     border_color: null,
   },
+  budget_enabled: false,
+  session_token_budget: null,
+  session_cost_budget: null,
+  budget_action: "warn",
+  budget_warning_threshold: 0.8,
 };
 
 function createConfigStore() {

src/lib/stores/conversations.ts

@@ -11,6 +11,13 @@ import { cleanupConversationTracking } from "$lib/tauri";
 import { characterState } from "$lib/stores/character";
 import { sessionsStore } from "$lib/stores/sessions";
 
+export interface ConversationSummary {
+  generatedAt: Date;
+  content: string;
+  messageCount: number;
+  tokenEstimate: number;
+}
+
 export interface Conversation {
   id: string;
   name: string;
@@ -27,6 +34,7 @@ export interface Conversation {
   createdAt: Date;
   lastActivityAt: Date;
   attachments: Attachment[];
+  summary: ConversationSummary | null;
 }
 
 function createConversationsStore() {
@@ -63,6 +71,7 @@ function createConversationsStore() {
       createdAt: new Date(),
       lastActivityAt: new Date(),
       attachments: [],
+      summary: null,
     };
   }
 
@@ -420,7 +429,12 @@ function createConversationsStore() {
       });
     },
 
-    addLine: (type: TerminalLine["type"], content: string, toolName?: string) => {
+    addLine: (
+      type: TerminalLine["type"],
+      content: string,
+      toolName?: string,
+      cost?: TerminalLine["cost"]
+    ) => {
       ensureInitialized();
       const activeId = get(activeConversationId);
       if (!activeId) return "";
@@ -431,6 +445,7 @@ function createConversationsStore() {
         content,
         timestamp: new Date(),
         toolName,
+        cost,
       };
 
       conversations.update((convs) => {
@@ -451,7 +466,8 @@ function createConversationsStore() {
       conversationId: string,
       type: TerminalLine["type"],
       content: string,
-      toolName?: string
+      toolName?: string,
+      cost?: TerminalLine["cost"]
     ) => {
       ensureInitialized();
 
@@ -461,6 +477,7 @@ function createConversationsStore() {
         content,
         timestamp: new Date(),
         toolName,
+        cost,
       };
 
       conversations.update((convs) => {
@@ -636,6 +653,130 @@ function createConversationsStore() {
       return conv?.attachments || [];
     },
 
+    // Summary/compaction functions
+    setSummary: (conversationId: string, summary: ConversationSummary) => {
+      conversations.update((convs) => {
+        const conv = convs.get(conversationId);
+        if (conv) {
+          conv.summary = summary;
+          conv.lastActivityAt = new Date();
+        }
+        return convs;
+      });
+    },
+
+    clearSummary: (conversationId: string) => {
+      conversations.update((convs) => {
+        const conv = convs.get(conversationId);
+        if (conv) {
+          conv.summary = null;
+          conv.lastActivityAt = new Date();
+        }
+        return convs;
+      });
+    },
+
+    getSummary: (conversationId: string): ConversationSummary | null => {
+      const convs = get(conversations);
+      const conv = convs.get(conversationId);
+      return conv?.summary || null;
+    },
+
+    // Estimate token count for a conversation (rough approximation: ~4 chars per token)
+    estimateTokenCount: (conversationId: string): number => {
+      const convs = get(conversations);
+      const conv = convs.get(conversationId);
+      if (!conv) return 0;
+
+      const relevantLines = conv.terminalLines.filter(
+        (line) => line.type === "user" || line.type === "assistant"
+      );
+
+      const totalChars = relevantLines.reduce((sum, line) => sum + line.content.length, 0);
+      return Math.ceil(totalChars / 4);
+    },
+
+    // Get conversation content suitable for summarisation
+    getConversationForSummary: (conversationId: string): string => {
+      const convs = get(conversations);
+      const conv = convs.get(conversationId);
+      if (!conv) return "";
+
+      const relevantLines = conv.terminalLines.filter(
+        (line) => line.type === "user" || line.type === "assistant"
+      );
+
+      return relevantLines
+        .map((line) => {
+          const role = line.type === "user" ? "User" : "Assistant";
+          return `${role}: ${line.content}`;
+        })
+        .join("\n\n");
+    },
+
+    // Compact conversation by keeping only recent messages
+    compactConversation: (conversationId: string, keepRecentCount: number = 10) => {
+      conversations.update((convs) => {
+        const conv = convs.get(conversationId);
+        if (conv && conv.terminalLines.length > keepRecentCount) {
+          // Keep system messages and the most recent user/assistant messages
+          const systemLines = conv.terminalLines.filter(
+            (line) => line.type !== "user" && line.type !== "assistant"
+          );
+          const chatLines = conv.terminalLines.filter(
+            (line) => line.type === "user" || line.type === "assistant"
+          );
+
+          // Keep only the most recent chat messages
+          const recentChatLines = chatLines.slice(-keepRecentCount);
+
+          // Combine: system lines at original positions + recent chat lines
+          conv.terminalLines = [...systemLines.slice(-5), ...recentChatLines];
+          conv.lastActivityAt = new Date();
+        }
+        return convs;
+      });
+    },
+
+    // Compact conversation with a summary - clears old messages and injects summary context
+    compactWithSummary: (
+      conversationId: string,
+      summaryContent: string,
+      messageCount: number,
+      tokenEstimate: number
+    ) => {
+      conversations.update((convs) => {
+        const conv = convs.get(conversationId);
+        if (conv) {
+          // Store the summary
+          conv.summary = {
+            generatedAt: new Date(),
+            content: summaryContent,
+            messageCount,
+            tokenEstimate,
+          };
+
+          // Clear all messages and add a context injection message
+          conv.terminalLines = [
+            {
+              id: generateLineId(),
+              type: "system",
+              content: `[Conversation compacted] Previous session had ${messageCount} messages (~${tokenEstimate.toLocaleString()} tokens). Context preserved below.`,
+              timestamp: new Date(),
+            },
+            {
+              id: generateLineId(),
+              type: "system",
+              content: `Previous Session Context:\n${summaryContent}`,
+              timestamp: new Date(),
+            },
+          ];
+          conv.lastActivityAt = new Date();
+        }
+        return convs;
+      });
+    },
+
     // Add initialization helper
     initialize: () => {
       ensureInitialized();

src/lib/stores/costTracking.ts

@@ -0,0 +1,182 @@
+import { writable, derived } from "svelte/store";
+import { invoke } from "@tauri-apps/api/core";
+import { notificationManager } from "$lib/notifications/notificationManager";
+
+// Types matching Rust backend
+export interface DailyCost {
+  date: string;
+  input_tokens: number;
+  output_tokens: number;
+  cost_usd: number;
+  messages_sent: number;
+  sessions_count: number;
+}
+
+export interface CostSummary {
+  period_days: number;
+  total_input_tokens: number;
+  total_output_tokens: number;
+  total_cost: number;
+  total_messages: number;
+  total_sessions: number;
+  average_daily_cost: number;
+  daily_breakdown: DailyCost[];
+}
+
+export type AlertType = "Daily" | "Weekly" | "Monthly";
+
+export interface CostAlert {
+  alert_type: AlertType;
+  threshold: number;
+  current_cost: number;
+}
+
+export interface CostAlertThresholds {
+  daily: number | null;
+  weekly: number | null;
+  monthly: number | null;
+}
+
+// Store state
+interface CostTrackingState {
+  todayCost: number;
+  weekCost: number;
+  monthCost: number;
+  summary: CostSummary | null;
+  alerts: CostAlert[];
+  thresholds: CostAlertThresholds;
+  isLoading: boolean;
+  lastUpdated: Date | null;
+}
+
+const defaultState: CostTrackingState = {
+  todayCost: 0,
+  weekCost: 0,
+  monthCost: 0,
+  summary: null,
+  alerts: [],
+  thresholds: { daily: null, weekly: null, monthly: null },
+  isLoading: false,
+  lastUpdated: null,
+};
+
+function createCostTrackingStore() {
+  const { subscribe, set, update } = writable<CostTrackingState>(defaultState);
+
+  return {
+    subscribe,
+
+    async refresh() {
+      update((s) => ({ ...s, isLoading: true }));
+
+      try {
+        const [todayCost, weekCost, monthCost, alerts] = await Promise.all([
+          invoke<number>("get_today_cost"),
+          invoke<number>("get_week_cost"),
+          invoke<number>("get_month_cost"),
+          invoke<CostAlert[]>("get_cost_alerts"),
+        ]);
+
+        update((s) => ({
+          ...s,
+          todayCost,
+          weekCost,
+          monthCost,
+          alerts,
+          isLoading: false,
+          lastUpdated: new Date(),
+        }));
+
+        // Trigger notifications for any new alerts
+        if (alerts.length > 0) {
+          for (const alert of alerts) {
+            const message = getAlertMessage(alert);
+            notificationManager.notifyCostAlert(message);
+          }
+        }
+
+        return alerts;
+      } catch (error) {
+        console.error("Failed to refresh cost tracking:", error);
+        update((s) => ({ ...s, isLoading: false }));
+        return [];
+      }
+    },
+
+    async getSummary(days: number): Promise<CostSummary | null> {
+      try {
+        const summary = await invoke<CostSummary>("get_cost_summary", { days });
+        update((s) => ({ ...s, summary }));
+        return summary;
+      } catch (error) {
+        console.error("Failed to get cost summary:", error);
+        return null;
+      }
+    },
+
+    async setAlertThresholds(thresholds: CostAlertThresholds) {
+      try {
+        await invoke("set_cost_alert_thresholds", {
+          daily: thresholds.daily,
+          weekly: thresholds.weekly,
+          monthly: thresholds.monthly,
+        });
+        update((s) => ({ ...s, thresholds }));
+      } catch (error) {
+        console.error("Failed to set alert thresholds:", error);
+      }
+    },
+
+    async exportCsv(days: number): Promise<string | null> {
+      try {
+        return await invoke<string>("export_cost_csv", { days });
+      } catch (error) {
+        console.error("Failed to export CSV:", error);
+        return null;
+      }
+    },
+
+    reset() {
+      set(defaultState);
+    },
+  };
+}
+
+export const costTrackingStore = createCostTrackingStore();
+
+// Derived stores for formatted values
+export const formattedCosts = derived(costTrackingStore, ($store) => ({
+  today: formatCost($store.todayCost),
+  week: formatCost($store.weekCost),
+  month: formatCost($store.monthCost),
+  todayRaw: $store.todayCost,
+  weekRaw: $store.weekCost,
+  monthRaw: $store.monthCost,
+}));
+
+// Helper functions
+export function formatCost(cost: number): string {
+  if (cost < 0.01) {
+    return `$${cost.toFixed(4)}`;
+  }
+  if (cost < 1) {
+    return `$${cost.toFixed(3)}`;
+  }
+  return `$${cost.toFixed(2)}`;
+}
+
+export function formatAlertType(type: AlertType): string {
+  switch (type) {
+    case "Daily":
+      return "Today";
+    case "Weekly":
+      return "This Week";
+    case "Monthly":
+      return "This Month";
+  }
+}
+
+export function getAlertMessage(alert: CostAlert): string {
+  const period = formatAlertType(alert.alert_type);
+  return `${period}'s spending (${formatCost(alert.current_cost)}) has exceeded your ${formatCost(alert.threshold)} threshold`;
+}

src/lib/stores/stats.test.ts

@@ -1,7 +1,25 @@
 import { describe, it, expect, beforeEach, vi } from "vitest";
 import { get } from "svelte/store";
-import { stats, formattedStats, resetSessionStats } from "./stats";
-import type { UsageStats } from "./stats";
+import {
+  stats,
+  formattedStats,
+  resetSessionStats,
+  contextWarning,
+  getContextWarningMessage,
+  estimateMessageCost,
+  formatTokenCount,
+  MODEL_PRICING,
+} from "./stats";
+import type { UsageStats, ToolTokenStats } from "./stats";
+
+// Helper function to create ToolTokenStats for tests
+function toolStats(callCount: number, inputTokens = 0, outputTokens = 0): ToolTokenStats {
+  return {
+    call_count: callCount,
+    estimated_input_tokens: inputTokens,
+    estimated_output_tokens: outputTokens,
+  };
+}
 
 // Mock Tauri APIs
 vi.mock("@tauri-apps/api/event", () => ({
@@ -34,6 +52,11 @@ describe("stats store", () => {
       tools_usage: {},
       session_tools_usage: {},
       session_duration_seconds: 0,
+      context_tokens_used: 0,
+      context_window_limit: 200000,
+      context_utilisation_percent: 0,
+      potential_cache_hits: 0,
+      potential_cache_savings_tokens: 0,
     });
   });
 
@@ -63,9 +86,14 @@ describe("stats store", () => {
         session_files_edited: 2,
         files_created: 1,
         session_files_created: 1,
-        tools_usage: { Read: 5, Edit: 3 },
-        session_tools_usage: { Read: 2, Edit: 1 },
+        tools_usage: { Read: toolStats(5), Edit: toolStats(3) },
+        session_tools_usage: { Read: toolStats(2), Edit: toolStats(1) },
         session_duration_seconds: 300,
+        context_tokens_used: 500,
+        context_window_limit: 200000,
+        context_utilisation_percent: 0.25,
+        potential_cache_hits: 0,
+        potential_cache_savings_tokens: 0,
       };
 
       stats.set(newStats);
@@ -74,7 +102,8 @@ describe("stats store", () => {
       expect(currentStats.total_input_tokens).toBe(1000);
       expect(currentStats.total_output_tokens).toBe(2000);
       expect(currentStats.model).toBe("claude-sonnet-4");
-      expect(currentStats.tools_usage).toEqual({ Read: 5, Edit: 3 });
+      expect(currentStats.tools_usage.Read?.call_count).toBe(5);
+      expect(currentStats.tools_usage.Edit?.call_count).toBe(3);
     });
 
     it("can be updated with update function", () => {
@@ -109,9 +138,14 @@ describe("stats store", () => {
         session_files_edited: 2,
         files_created: 1,
         session_files_created: 1,
-        tools_usage: { Read: 5, Edit: 3 },
-        session_tools_usage: { Read: 2, Edit: 1 },
+        tools_usage: { Read: toolStats(5), Edit: toolStats(3) },
+        session_tools_usage: { Read: toolStats(2), Edit: toolStats(1) },
         session_duration_seconds: 300,
+        context_tokens_used: 500,
+        context_window_limit: 200000,
+        context_utilisation_percent: 0.25,
+        potential_cache_hits: 0,
+        potential_cache_savings_tokens: 0,
       });
 
       // Reset session stats
@@ -127,7 +161,8 @@ describe("stats store", () => {
       expect(currentStats.code_blocks_generated).toBe(3);
       expect(currentStats.files_edited).toBe(5);
       expect(currentStats.files_created).toBe(1);
-      expect(currentStats.tools_usage).toEqual({ Read: 5, Edit: 3 });
+      expect(currentStats.tools_usage.Read?.call_count).toBe(5);
+      expect(currentStats.tools_usage.Edit?.call_count).toBe(3);
       expect(currentStats.model).toBe("claude-sonnet-4");
 
       // Session stats should be reset
@@ -277,8 +312,8 @@ describe("stats store", () => {
     });
 
     it("exposes tools usage directly", () => {
-      const toolsUsage = { Read: 10, Edit: 5, Write: 3 };
-      const sessionToolsUsage = { Read: 2, Edit: 1 };
+      const toolsUsage = { Read: toolStats(10), Edit: toolStats(5), Write: toolStats(3) };
+      const sessionToolsUsage = { Read: toolStats(2), Edit: toolStats(1) };
 
       stats.update((current) => ({
         ...current,
@@ -331,9 +366,14 @@ describe("stats store", () => {
         session_files_edited: 1,
         files_created: 1,
         session_files_created: 0,
-        tools_usage: { Read: 3 },
-        session_tools_usage: { Read: 1 },
+        tools_usage: { Read: toolStats(3) },
+        session_tools_usage: { Read: toolStats(1) },
         session_duration_seconds: 60,
+        context_tokens_used: 50,
+        context_window_limit: 200000,
+        context_utilisation_percent: 0.025,
+        potential_cache_hits: 0,
+        potential_cache_savings_tokens: 0,
       };
 
       stats.set(fullStats);
@@ -343,4 +383,221 @@ describe("stats store", () => {
       expect(currentStats).toEqual(fullStats);
     });
   });
+
+  describe("context window tracking", () => {
+    it("tracks context tokens used", () => {
+      stats.update((current) => ({
+        ...current,
+        context_tokens_used: 100000,
+        context_window_limit: 200000,
+        context_utilisation_percent: 50.0,
+      }));
+
+      const currentStats = get(stats);
+      expect(currentStats.context_tokens_used).toBe(100000);
+      expect(currentStats.context_window_limit).toBe(200000);
+      expect(currentStats.context_utilisation_percent).toBe(50.0);
+    });
+
+    it("formats context stats correctly", () => {
+      stats.update((current) => ({
+        ...current,
+        context_tokens_used: 150000,
+        context_window_limit: 200000,
+        context_utilisation_percent: 75.5,
+      }));
+
+      const formatted = get(formattedStats);
+      expect(formatted.contextUsed).toBe("150,000");
+      expect(formatted.contextLimit).toBe("200,000");
+      expect(formatted.contextRemaining).toBe("50,000");
+      expect(formatted.contextUtilisation).toBe("75.5%");
+    });
+
+    it("calculates remaining tokens correctly at limit", () => {
+      stats.update((current) => ({
+        ...current,
+        context_tokens_used: 200000,
+        context_window_limit: 200000,
+        context_utilisation_percent: 100.0,
+      }));
+
+      const formatted = get(formattedStats);
+      expect(formatted.contextRemaining).toBe("0");
+    });
+
+    it("handles over-limit gracefully", () => {
+      stats.update((current) => ({
+        ...current,
+        context_tokens_used: 250000,
+        context_window_limit: 200000,
+        context_utilisation_percent: 125.0,
+      }));
+
+      const formatted = get(formattedStats);
+      expect(formatted.contextRemaining).toBe("0");
+    });
+  });
+
+  describe("contextWarning derived store", () => {
+    it("returns null when under 50%", () => {
+      stats.update((current) => ({
+        ...current,
+        context_utilisation_percent: 40.0,
+      }));
+
+      const warning = get(contextWarning);
+      expect(warning).toBeNull();
+    });
+
+    it("returns moderate when between 50-74%", () => {
+      stats.update((current) => ({
+        ...current,
+        context_utilisation_percent: 60.0,
+      }));
+
+      const warning = get(contextWarning);
+      expect(warning).toBe("moderate");
+    });
+
+    it("returns high when between 75-89%", () => {
+      stats.update((current) => ({
+        ...current,
+        context_utilisation_percent: 80.0,
+      }));
+
+      const warning = get(contextWarning);
+      expect(warning).toBe("high");
+    });
+
+    it("returns critical when 90%+", () => {
+      stats.update((current) => ({
+        ...current,
+        context_utilisation_percent: 95.0,
+      }));
+
+      const warning = get(contextWarning);
+      expect(warning).toBe("critical");
+    });
+  });
+
+  describe("getContextWarningMessage", () => {
+    it("returns correct message for moderate warning", () => {
+      const message = getContextWarningMessage("moderate");
+      expect(message).toContain("50%+");
+      expect(message).toContain("Consider starting a new conversation");
+    });
+
+    it("returns correct message for high warning", () => {
+      const message = getContextWarningMessage("high");
+      expect(message).toContain("75%+");
+      expect(message).toContain("Responses may degrade");
+    });
+
+    it("returns correct message for critical warning", () => {
+      const message = getContextWarningMessage("critical");
+      expect(message).toContain("90%+");
+      expect(message).toContain("Start a new conversation");
+    });
+  });
+
+  describe("formatTokenCount", () => {
+    it("formats small numbers directly", () => {
+      expect(formatTokenCount(0)).toBe("0");
+      expect(formatTokenCount(100)).toBe("100");
+      expect(formatTokenCount(999)).toBe("999");
+    });
+
+    it("formats thousands with K suffix", () => {
+      expect(formatTokenCount(1000)).toBe("1.0K");
+      expect(formatTokenCount(1500)).toBe("1.5K");
+      expect(formatTokenCount(10000)).toBe("10.0K");
+      expect(formatTokenCount(999999)).toBe("1000.0K");
+    });
+
+    it("formats millions with M suffix", () => {
+      expect(formatTokenCount(1000000)).toBe("1.0M");
+      expect(formatTokenCount(1500000)).toBe("1.5M");
+      expect(formatTokenCount(10000000)).toBe("10.0M");
+    });
+  });
+
+  describe("estimateMessageCost", () => {
+    it("estimates tokens at ~4 chars per token", () => {
+      const result = estimateMessageCost("test", 0, null); // 4 chars = 1 token
+      expect(result.messageTokens).toBe(1);
+    });
+
+    it("rounds up partial tokens", () => {
+      const result = estimateMessageCost("a", 0, null); // 1 char rounds up to 1 token
+      expect(result.messageTokens).toBe(1);
+
+      const result2 = estimateMessageCost("abcde", 0, null); // 5 chars = 2 tokens
+      expect(result2.messageTokens).toBe(2);
+    });
+
+    it("returns 0 tokens for empty string", () => {
+      const result = estimateMessageCost("", 0, null);
+      expect(result.messageTokens).toBe(0);
+      expect(result.estimatedCost).toBe(0);
+    });
+
+    it("adds context tokens to total", () => {
+      const result = estimateMessageCost("test", 1000, null); // 1 token + 1000 context
+      expect(result.messageTokens).toBe(1);
+      expect(result.totalInputTokens).toBe(1001);
+    });
+
+    it("calculates cost using Sonnet pricing by default", () => {
+      // 100 chars = 25 tokens, $3 per million input tokens
+      const result = estimateMessageCost("a".repeat(100), 0, null);
+      expect(result.messageTokens).toBe(25);
+      const expectedCost = (25 / 1_000_000) * 3.0;
+      expect(result.estimatedCost).toBeCloseTo(expectedCost, 8);
+    });
+
+    it("uses Opus pricing for Opus models", () => {
+      const result = estimateMessageCost("a".repeat(100), 0, "claude-opus-4-5-20251101");
+      expect(result.messageTokens).toBe(25);
+      const expectedCost = (25 / 1_000_000) * 5.0; // Opus 4.5: $5 per million input
+      expect(result.estimatedCost).toBeCloseTo(expectedCost, 8);
+    });
+
+    it("uses Haiku pricing for Haiku models", () => {
+      const result = estimateMessageCost("a".repeat(100), 0, "claude-3-5-haiku-20241022");
+      expect(result.messageTokens).toBe(25);
+      const expectedCost = (25 / 1_000_000) * 1.0; // Haiku: $1 per million
+      expect(result.estimatedCost).toBeCloseTo(expectedCost, 8);
+    });
+
+    it("falls back to Sonnet pricing for unknown models", () => {
+      const result = estimateMessageCost("a".repeat(100), 0, "unknown-model");
+      expect(result.messageTokens).toBe(25);
+      const expectedCost = (25 / 1_000_000) * 3.0; // Default Sonnet: $3 per million
+      expect(result.estimatedCost).toBeCloseTo(expectedCost, 8);
+    });
+  });
+
+  describe("MODEL_PRICING", () => {
+    it("contains expected Opus pricing", () => {
+      // Opus 4.5 has reduced pricing
+      expect(MODEL_PRICING["claude-opus-4-5-20251101"]).toEqual({ input: 5.0, output: 25.0 });
+      // Previous Opus models have higher pricing
+      expect(MODEL_PRICING["claude-opus-4-1-20250805"]).toEqual({ input: 15.0, output: 75.0 });
+      expect(MODEL_PRICING["claude-opus-4-20250514"]).toEqual({ input: 15.0, output: 75.0 });
+    });
+
+    it("contains expected Sonnet pricing", () => {
+      expect(MODEL_PRICING["claude-sonnet-4-5-20250929"]).toEqual({ input: 3.0, output: 15.0 });
+      expect(MODEL_PRICING["claude-sonnet-4-20250514"]).toEqual({ input: 3.0, output: 15.0 });
+      expect(MODEL_PRICING["claude-3-7-sonnet-20250219"]).toEqual({ input: 3.0, output: 15.0 });
+      expect(MODEL_PRICING["claude-3-5-sonnet-20241022"]).toEqual({ input: 3.0, output: 15.0 });
+    });
+
+    it("contains expected Haiku pricing", () => {
+      expect(MODEL_PRICING["claude-haiku-4-5-20251001"]).toEqual({ input: 1.0, output: 5.0 });
+      expect(MODEL_PRICING["claude-3-5-haiku-20241022"]).toEqual({ input: 1.0, output: 5.0 });
+      expect(MODEL_PRICING["claude-3-haiku-20240307"]).toEqual({ input: 0.25, output: 1.25 });
+    });
+  });
 });

src/lib/stores/stats.ts

@@ -1,6 +1,66 @@
 import { writable, derived } from "svelte/store";
 import { listen } from "@tauri-apps/api/event";
 import { invoke } from "@tauri-apps/api/core";
+import { costTrackingStore } from "./costTracking";
+
+export type ContextWarning = "moderate" | "high" | "critical";
+export type BudgetType = "token" | "cost";
+
+// Model pricing (per million tokens) - keep in sync with stats.rs
+// Source: https://platform.claude.com/docs/en/about-claude/models/overview
+export const MODEL_PRICING: Record<string, { input: number; output: number }> = {
+  // Current generation (Claude 4.5)
+  "claude-opus-4-5-20251101": { input: 5.0, output: 25.0 },
+  "claude-sonnet-4-5-20250929": { input: 3.0, output: 15.0 },
+  "claude-haiku-4-5-20251001": { input: 1.0, output: 5.0 },
+  // Previous generation (Claude 4.x)
+  "claude-opus-4-1-20250805": { input: 15.0, output: 75.0 },
+  "claude-opus-4-20250514": { input: 15.0, output: 75.0 },
+  "claude-sonnet-4-20250514": { input: 3.0, output: 15.0 },
+  // Legacy (Claude 3.x)
+  "claude-3-7-sonnet-20250219": { input: 3.0, output: 15.0 },
+  "claude-3-5-sonnet-20241022": { input: 3.0, output: 15.0 },
+  "claude-3-5-sonnet-20240620": { input: 3.0, output: 15.0 },
+  "claude-3-5-haiku-20241022": { input: 1.0, output: 5.0 },
+  "claude-3-opus-20240229": { input: 15.0, output: 75.0 },
+  "claude-3-sonnet-20240229": { input: 3.0, output: 15.0 },
+  "claude-3-haiku-20240307": { input: 0.25, output: 1.25 },
+};
+
+const DEFAULT_PRICING = { input: 3.0, output: 15.0 }; // Default to Sonnet
+
+export interface CostEstimate {
+  messageTokens: number;
+  totalInputTokens: number;
+  estimatedCost: number;
+}
+
+// Estimate cost for a message before sending
+export function estimateMessageCost(
+  messageText: string,
+  contextTokensUsed: number,
+  model: string | null
+): CostEstimate {
+  // Estimate tokens using ~4 chars per token heuristic
+  const messageTokens = Math.ceil(messageText.length / 4);
+  const totalInputTokens = contextTokensUsed + messageTokens;
+
+  const pricing = model ? (MODEL_PRICING[model] ?? DEFAULT_PRICING) : DEFAULT_PRICING;
+  const estimatedCost = (totalInputTokens / 1_000_000) * pricing.input;
+
+  return { messageTokens, totalInputTokens, estimatedCost };
+}
+export type BudgetStatus =
+  | { type: "ok" }
+  | { type: "warning"; budget_type: BudgetType; percent_used: number }
+  | { type: "exceeded"; budget_type: BudgetType };
+
+// Per-tool token usage statistics
+export interface ToolTokenStats {
+  call_count: number;
+  estimated_input_tokens: number;
+  estimated_output_tokens: number;
+}
 
 export interface UsageStats {
   total_input_tokens: number;
@@ -20,9 +80,18 @@ export interface UsageStats {
   session_files_edited: number;
   files_created: number;
   session_files_created: number;
-  tools_usage: Record<string, number>;
-  session_tools_usage: Record<string, number>;
+  tools_usage: Record<string, ToolTokenStats>;
+  session_tools_usage: Record<string, ToolTokenStats>;
   session_duration_seconds: number;
+
+  // Context window tracking
+  context_tokens_used: number;
+  context_window_limit: number;
+  context_utilisation_percent: number;
+
+  // Cache analytics (tracks potential savings from repeated tool calls)
+  potential_cache_hits: number;
+  potential_cache_savings_tokens: number;
 }
 
 // Main stats store
@@ -45,8 +114,24 @@ export const stats = writable<UsageStats>({
   tools_usage: {},
   session_tools_usage: {},
   session_duration_seconds: 0,
+  context_tokens_used: 0,
+  context_window_limit: 200000,
+  context_utilisation_percent: 0,
+  potential_cache_hits: 0,
+  potential_cache_savings_tokens: 0,
 });
 
+// Format token count with K/M suffix
+export function formatTokenCount(tokens: number): string {
+  if (tokens >= 1000000) {
+    return `${(tokens / 1000000).toFixed(1)}M`;
+  }
+  if (tokens >= 1000) {
+    return `${(tokens / 1000).toFixed(1)}K`;
+  }
+  return tokens.toString();
+}
+
 // Derived store for formatted display values
 export const formattedStats = derived(stats, ($stats) => {
   const formatNumber = (num: number) => num.toLocaleString();
@@ -65,6 +150,20 @@ export const formattedStats = derived(stats, ($stats) => {
     }
   };
 
+  // Format tool stats with token info
+  const formatToolStats = (toolStats: Record<string, ToolTokenStats>) => {
+    return Object.entries(toolStats).map(([name, stats]) => ({
+      name,
+      callCount: stats.call_count,
+      totalTokens: stats.estimated_input_tokens + stats.estimated_output_tokens,
+      formattedTokens: formatTokenCount(
+        stats.estimated_input_tokens + stats.estimated_output_tokens
+      ),
+      inputTokens: stats.estimated_input_tokens,
+      outputTokens: stats.estimated_output_tokens,
+    }));
+  };
+
   return {
     totalTokens: formatNumber($stats.total_input_tokens + $stats.total_output_tokens),
     totalInputTokens: formatNumber($stats.total_input_tokens),
@@ -88,9 +187,116 @@ export const formattedStats = derived(stats, ($stats) => {
     sessionDuration: formatDuration($stats.session_duration_seconds),
     toolsUsage: $stats.tools_usage,
     sessionToolsUsage: $stats.session_tools_usage,
+    // Formatted tool stats with token info
+    sessionToolsFormatted: formatToolStats($stats.session_tools_usage),
+    toolsFormatted: formatToolStats($stats.tools_usage),
+
+    // Context window tracking
+    contextUsed: formatNumber($stats.context_tokens_used),
+    contextLimit: formatNumber($stats.context_window_limit),
+    contextRemaining: formatNumber(
+      Math.max(0, $stats.context_window_limit - $stats.context_tokens_used)
+    ),
+    contextUtilisation: `${$stats.context_utilisation_percent.toFixed(1)}%`,
   };
 });
 
+// Derived store for context warning state
+export const contextWarning = derived(stats, ($stats): ContextWarning | null => {
+  if ($stats.context_utilisation_percent >= 90) {
+    return "critical";
+  } else if ($stats.context_utilisation_percent >= 75) {
+    return "high";
+  } else if ($stats.context_utilisation_percent >= 50) {
+    return "moderate";
+  }
+  return null;
+});
+
+// Get warning message for context utilisation
+export function getContextWarningMessage(warning: ContextWarning): string {
+  switch (warning) {
+    case "moderate":
+      return "Context window is 50%+ full. Consider starting a new conversation for better performance.";
+    case "high":
+      return "Context window is 75%+ full. Responses may degrade. Consider summarising or starting fresh.";
+    case "critical":
+      return "Context window is nearly full (90%+)! Start a new conversation to avoid errors.";
+  }
+}
+
+// Budget checking functions
+export function checkBudget(
+  stats: UsageStats,
+  budgetEnabled: boolean,
+  tokenBudget: number | null,
+  costBudget: number | null,
+  warningThreshold: number
+): BudgetStatus {
+  if (!budgetEnabled) {
+    return { type: "ok" };
+  }
+
+  const sessionTokens = stats.session_input_tokens + stats.session_output_tokens;
+
+  // Check token budget
+  if (tokenBudget !== null) {
+    if (sessionTokens >= tokenBudget) {
+      return { type: "exceeded", budget_type: "token" };
+    }
+    const percentUsed = sessionTokens / tokenBudget;
+    if (percentUsed >= warningThreshold) {
+      return { type: "warning", budget_type: "token", percent_used: percentUsed * 100 };
+    }
+  }
+
+  // Check cost budget
+  if (costBudget !== null) {
+    if (stats.session_cost_usd >= costBudget) {
+      return { type: "exceeded", budget_type: "cost" };
+    }
+    const percentUsed = stats.session_cost_usd / costBudget;
+    if (percentUsed >= warningThreshold) {
+      return { type: "warning", budget_type: "cost", percent_used: percentUsed * 100 };
+    }
+  }
+
+  return { type: "ok" };
+}
+
+// Get budget status message
+export function getBudgetStatusMessage(status: BudgetStatus): string | null {
+  if (status.type === "ok") {
+    return null;
+  }
+
+  const budgetTypeLabel = status.budget_type === "token" ? "token" : "cost";
+
+  if (status.type === "exceeded") {
+    return `Session ${budgetTypeLabel} budget exceeded! Consider starting a new session.`;
+  }
+
+  return `Approaching ${budgetTypeLabel} budget limit (${status.percent_used.toFixed(0)}% used).`;
+}
+
+// Get remaining budget values
+export function getRemainingTokenBudget(
+  stats: UsageStats,
+  tokenBudget: number | null
+): number | null {
+  if (tokenBudget === null) return null;
+  const used = stats.session_input_tokens + stats.session_output_tokens;
+  return Math.max(0, tokenBudget - used);
+}
+
+export function getRemainingCostBudget(
+  stats: UsageStats,
+  costBudget: number | null
+): number | null {
+  if (costBudget === null) return null;
+  return Math.max(0, costBudget - stats.session_cost_usd);
+}
+
 // Note: Cost calculation is now done in the Rust backend
 
 // Initialize stats listener
@@ -102,6 +308,9 @@ export async function initStatsListener() {
 
     // The backend already tracks all totals - just set the stats directly
     stats.set(newStats);
+
+    // Refresh cost tracking to check for alerts (debounced - won't spam)
+    costTrackingStore.refresh();
   });
 
   // Load initial persisted stats from backend (no bridge required)