feat(usage): add per-task token usage tracking

Extract token usage from Claude Code's stream-json output in real-time
during task execution, replacing the inaccurate global JSONL log scanner.

- New `task_usage` table: tracks (task_id, provider, model) level usage
- Agent SDK: parse `message.usage` from assistant messages, accumulate
  per-model and return in Result
- Daemon: convert agent usage to entries, send with CompleteTask
- Server: store usage on task completion, expose workspace-level
  aggregation APIs (GET /api/usage/daily, GET /api/usage/summary)

server/pkg/agent/agent.go

@@ -62,6 +62,14 @@ type Message struct {
 	Level   string         // log level (Log)
 }
 
+// TokenUsage tracks token consumption for a single model.
+type TokenUsage struct {
+	InputTokens      int64
+	OutputTokens     int64
+	CacheReadTokens  int64
+	CacheWriteTokens int64
+}
+
 // Result is the final outcome after an agent session completes.
 type Result struct {
 	Status     string // "completed", "failed", "aborted", "timeout"
@@ -69,6 +77,7 @@ type Result struct {
 	Error      string // error message if failed
 	DurationMs int64
 	SessionID  string
+	Usage      map[string]TokenUsage // keyed by model name
 }
 
 // Config configures a Backend instance.

server/pkg/agent/claude.go

@@ -91,6 +91,7 @@ func (b *claudeBackend) Execute(ctx context.Context, prompt string, opts ExecOpt
 		var sessionID string
 		finalStatus := "completed"
 		var finalError string
+		usage := make(map[string]TokenUsage)
 
 		scanner := bufio.NewScanner(stdout)
 		scanner.Buffer(make([]byte, 0, 1024*1024), 10*1024*1024)
@@ -108,7 +109,7 @@ func (b *claudeBackend) Execute(ctx context.Context, prompt string, opts ExecOpt
 
 			switch msg.Type {
 			case "assistant":
-				b.handleAssistant(msg, msgCh, &output)
+				b.handleAssistant(msg, msgCh, &output, usage)
 			case "user":
 				b.handleUser(msg, msgCh)
 			case "system":
@@ -162,18 +163,29 @@ func (b *claudeBackend) Execute(ctx context.Context, prompt string, opts ExecOpt
 			Error:      finalError,
 			DurationMs: duration.Milliseconds(),
 			SessionID:  sessionID,
+			Usage:      usage,
 		}
 	}()
 
 	return &Session{Messages: msgCh, Result: resCh}, nil
 }
 
-func (b *claudeBackend) handleAssistant(msg claudeSDKMessage, ch chan<- Message, output *strings.Builder) {
+func (b *claudeBackend) handleAssistant(msg claudeSDKMessage, ch chan<- Message, output *strings.Builder, usage map[string]TokenUsage) {
 	var content claudeMessageContent
 	if err := json.Unmarshal(msg.Message, &content); err != nil {
 		return
 	}
 
+	// Accumulate token usage per model.
+	if content.Usage != nil && content.Model != "" {
+		u := usage[content.Model]
+		u.InputTokens += content.Usage.InputTokens
+		u.OutputTokens += content.Usage.OutputTokens
+		u.CacheReadTokens += content.Usage.CacheReadInputTokens
+		u.CacheWriteTokens += content.Usage.CacheCreationInputTokens
+		usage[content.Model] = u
+	}
+
 	for _, block := range content.Content {
 		switch block.Type {
 		case "text":
@@ -287,8 +299,17 @@ type claudeLogEntry struct {
 }
 
 type claudeMessageContent struct {
-	Role    string             `json:"role"`
+	Role    string               `json:"role"`
+	Model   string               `json:"model"`
 	Content []claudeContentBlock `json:"content"`
+	Usage   *claudeUsage         `json:"usage,omitempty"`
+}
+
+type claudeUsage struct {
+	InputTokens              int64 `json:"input_tokens"`
+	OutputTokens             int64 `json:"output_tokens"`
+	CacheReadInputTokens     int64 `json:"cache_read_input_tokens"`
+	CacheCreationInputTokens int64 `json:"cache_creation_input_tokens"`
 }
 
 type claudeContentBlock struct {

server/pkg/agent/claude_test.go

@@ -25,7 +25,7 @@ func TestClaudeHandleAssistantText(t *testing.T) {
 		}),
 	}
 
-	b.handleAssistant(msg, ch, &output)
+	b.handleAssistant(msg, ch, &output, make(map[string]TokenUsage))
 
 	if output.String() != "Hello world" {
 		t.Fatalf("expected output 'Hello world', got %q", output.String())
@@ -62,7 +62,7 @@ func TestClaudeHandleAssistantToolUse(t *testing.T) {
 		}),
 	}
 
-	b.handleAssistant(msg, ch, &output)
+	b.handleAssistant(msg, ch, &output, make(map[string]TokenUsage))
 
 	if output.String() != "" {
 		t.Fatalf("tool_use should not add to output, got %q", output.String())
@@ -162,7 +162,7 @@ func TestClaudeHandleAssistantInvalidJSON(t *testing.T) {
 	}
 
 	// Should not panic
-	b.handleAssistant(msg, ch, &output)
+	b.handleAssistant(msg, ch, &output, make(map[string]TokenUsage))
 
 	if output.String() != "" {
 		t.Fatalf("expected empty output for invalid JSON, got %q", output.String())