f4d798e Merge pull request #144 from bnema/feat/queue-image-prompts · zut

M README.md +1 -1


@@ -902,7 +902,7 @@ ZUT_COMPACT_INPUT=1 zut # compact, just this run
902	902
903	903		## Queued messages
904	904
905		-	You can keep typing while the agent is working. Pressing `enter` during a turn queues the message instead of interrupting: it shows up above the status bar as `sliding in: <text>` and is delivered as the next user turn the moment the current one finishes. Queue as many as you want; they run in order. `esc` cancels the active turn, returns the most recently queued message to the input for editing, and drops any remaining stale follow-ups; `ctrl+c` while busy arms the exit hint instead of interrupting, a second `ctrl+c` within two seconds exits zut.
	905	+	You can keep typing while the agent is working. Pressing `enter` during a turn queues the message, including pasted screenshots, instead of interrupting. It appears above the status bar as `sliding in: <text> [image]` and is delivered at the next safe model-call boundary. Queue as many messages as you want; they run in order. `alt+up` returns the most recently queued message and its image markers to the input for editing. `esc` cancels the active turn, restores the most recent queued message in the same way, and drops any remaining stale follow-ups; `ctrl+c` while busy arms the exit hint instead of interrupting, and a second `ctrl+c` within two seconds exits zut.
906	906
907	907		To recover the most recently queued message back into the editor (to tweak it before it runs), press `Option+↑`. In VS Code's integrated terminal that chord doesn't survive xterm.js's macOS key handling — use `Option+Shift+↑` there. zut's hint line under the sliding-in queue adapts automatically based on `$TERM_PROGRAM`.
908	908


@@ -902,7 +902,7 @@ ZUT_COMPACT_INPUT=1 zut # compact, just this run
902			902
903		## Queued messages	903		## Queued messages
904			904
905	-	You can keep typing while the agent is working. Pressing `enter` during a turn queues the message instead of interrupting: it shows up above the status bar as `sliding in: <text>` and is delivered as the next user turn the moment the current one finishes. Queue as many as you want; they run in order. `esc` cancels the active turn, returns the most recently queued message to the input for editing, and drops any remaining stale follow-ups; `ctrl+c` while busy arms the exit hint instead of interrupting, a second `ctrl+c` within two seconds exits zut.	905	+	You can keep typing while the agent is working. Pressing `enter` during a turn queues the message, including pasted screenshots, instead of interrupting. It appears above the status bar as `sliding in: <text> [image]` and is delivered at the next safe model-call boundary. Queue as many messages as you want; they run in order. `alt+up` returns the most recently queued message and its image markers to the input for editing. `esc` cancels the active turn, restores the most recent queued message in the same way, and drops any remaining stale follow-ups; `ctrl+c` while busy arms the exit hint instead of interrupting, and a second `ctrl+c` within two seconds exits zut.
906			906
907		To recover the most recently queued message back into the editor (to tweak it before it runs), press `Option+↑`. In VS Code's integrated terminal that chord doesn't survive xterm.js's macOS key handling — use `Option+Shift+↑` there. zut's hint line under the sliding-in queue adapts automatically based on `$TERM_PROGRAM`.	907		To recover the most recently queued message back into the editor (to tweak it before it runs), press `Option+↑`. In VS Code's integrated terminal that chord doesn't survive xterm.js's macOS key handling — use `Option+Shift+↑` there. zut's hint line under the sliding-in queue adapts automatically based on `$TERM_PROGRAM`.
908			908

M packages/agent/internal/orchestration/coordinator.go +32 -10


@@ -4,7 +4,10 @@
4	4		// events and execute the returned actions.
5	5		package orchestration
6	6
7		-	import "github.com/bnema/zut/packages/agent/subagents"
	7	+	import (
	8	+	"github.com/bnema/zut/packages/agent/subagents"
	9	+	"github.com/bnema/zut/packages/provider"
	10	+	)
8	11
9	12		// Coordinator collects one sealed worker wave per manager turn. A terminal
10	13		// worker cannot wake the manager until the owning turn is sealed. Queued user

@@ -17,7 +20,12 @@ type Coordinator struct {
17	20		goalActive bool
18	21		workers map[string]*worker
19	22		workerOrder []string
20		-	queuedUser []string
	23	+	queuedUser []queuedUserInput
	24	+	}
	25	+
	26	+	type queuedUserInput struct {
	27	+	text string
	28	+	images []provider.ImageBlock
21	29		}
22	30
23	31		type worker struct {

@@ -43,6 +51,7 @@ type Event struct {
43	51		WorkerID string
44	52		Completion subagents.Completion
45	53		Text string
	54	+	Images []provider.ImageBlock
46	55		GoalActive bool
47	56		}
48	57

@@ -66,6 +75,7 @@ type Action struct {
66	75		Kind ActionKind
67	76		Reason WakeReason
68	77		Text string
	78	+	Images []provider.ImageBlock
69	79		Completions []subagents.Completion
70	80		}
71	81

@@ -124,10 +134,10 @@ func (c *Coordinator) Apply(event Event) Result {
124	134		return c.wakeIfIdle()
125	135		}
126	136		case EventUserInput:
127		-	if event.Text == "" {
	137	+	if event.Text == "" && len(event.Images) == 0 {
128	138		return Result{}
129	139		}
130		-	c.queuedUser = append(c.queuedUser, event.Text)
	140	+	c.queuedUser = append(c.queuedUser, queuedUserInput{text: event.Text, images: cloneImages(event.Images)})
131	141		return c.wakeIfIdle()
132	142		case EventGoalChanged:
133	143		c.goalActive = event.GoalActive

@@ -145,22 +155,34 @@ func (c *Coordinator) wakeIfIdle() Result {
145	155		}
146	156		completions := c.completedWorkers()
147	157		if len(c.queuedUser) != 0 {
148		-	text := c.queuedUser[0]
	158	+	input := c.queuedUser[0]
149	159		c.queuedUser = c.queuedUser[1:]
150		-	return c.startManager(WakeUser, text, completions)
	160	+	return c.startManager(WakeUser, input.text, input.images, completions)
151	161		}
152	162		if len(completions) != 0 {
153		-	return c.startManager(WakeWorkers, "", completions)
	163	+	return c.startManager(WakeWorkers, "", nil, completions)
154	164		}
155	165		if c.goalActive {
156		-	return c.startManager(WakeGoal, "", nil)
	166	+	return c.startManager(WakeGoal, "", nil, nil)
157	167		}
158	168		return Result{}
159	169		}
160	170
161		-	func (c *Coordinator) startManager(reason WakeReason, text string, completions []subagents.Completion) Result {
	171	+	func (c *Coordinator) startManager(reason WakeReason, text string, images []provider.ImageBlock, completions []subagents.Completion) Result {
162	172		c.managerActive = true
163		-	return Result{Actions: []Action{{Kind: ActionRunManager, Reason: reason, Text: text, Completions: completions}}}
	173	+	return Result{Actions: []Action{{Kind: ActionRunManager, Reason: reason, Text: text, Images: images, Completions: completions}}}
	174	+	}
	175	+
	176	+	func cloneImages(images []provider.ImageBlock) []provider.ImageBlock {
	177	+	if len(images) == 0 {
	178	+	return nil
	179	+	}
	180	+	cloned := make([]provider.ImageBlock, len(images))
	181	+	for index, image := range images {
	182	+	cloned[index] = image
	183	+	cloned[index].Data = append([]byte(nil), image.Data...)
	184	+	}
	185	+	return cloned
164	186		}
165	187
166	188		func (c *Coordinator) hasPendingWorkers() bool {

 // events and execute the returned actions.
 package orchestration
-import "github.com/bnema/zut/packages/agent/subagents"
+import (
+	"github.com/bnema/zut/packages/agent/subagents"
+	"github.com/bnema/zut/packages/provider"
+)
 // Coordinator collects one sealed worker wave per manager turn. A terminal
 // worker cannot wake the manager until the owning turn is sealed. Queued user
 	goalActive    bool
 	workers       map[string]*worker
 	workerOrder   []string
-	queuedUser    []string
+	queuedUser    []queuedUserInput
+}
+type queuedUserInput struct {
+	text   string
+	images []provider.ImageBlock
 }
 type worker struct {
 	WorkerID   string
 	Completion subagents.Completion
 	Text       string
+	Images     []provider.ImageBlock
 	GoalActive bool
 }
 	Kind        ActionKind
 	Reason      WakeReason
 	Text        string
+	Images      []provider.ImageBlock
 	Completions []subagents.Completion
 }
 			return c.wakeIfIdle()
 		}
 	case EventUserInput:
-		if event.Text == "" {
+		if event.Text == "" && len(event.Images) == 0 {
 			return Result{}
 		}
-		c.queuedUser = append(c.queuedUser, event.Text)
+		c.queuedUser = append(c.queuedUser, queuedUserInput{text: event.Text, images: cloneImages(event.Images)})
 		return c.wakeIfIdle()
 	case EventGoalChanged:
 		c.goalActive = event.GoalActive
 	}
 	completions := c.completedWorkers()
 	if len(c.queuedUser) != 0 {
-		text := c.queuedUser[0]
+		input := c.queuedUser[0]
 		c.queuedUser = c.queuedUser[1:]
-		return c.startManager(WakeUser, text, completions)
+		return c.startManager(WakeUser, input.text, input.images, completions)
 	}
 	if len(completions) != 0 {
-		return c.startManager(WakeWorkers, "", completions)
+		return c.startManager(WakeWorkers, "", nil, completions)
 	}
 	if c.goalActive {
-		return c.startManager(WakeGoal, "", nil)
+		return c.startManager(WakeGoal, "", nil, nil)
 	}
 	return Result{}
 }
-func (c *Coordinator) startManager(reason WakeReason, text string, completions []subagents.Completion) Result {
+func (c *Coordinator) startManager(reason WakeReason, text string, images []provider.ImageBlock, completions []subagents.Completion) Result {
 	c.managerActive = true
-	return Result{Actions: []Action{{Kind: ActionRunManager, Reason: reason, Text: text, Completions: completions}}}
+	return Result{Actions: []Action{{Kind: ActionRunManager, Reason: reason, Text: text, Images: images, Completions: completions}}}
+}
+func cloneImages(images []provider.ImageBlock) []provider.ImageBlock {
+	if len(images) == 0 {
+		return nil
+	}
+	cloned := make([]provider.ImageBlock, len(images))
+	for index, image := range images {
+		cloned[index] = image
+		cloned[index].Data = append([]byte(nil), image.Data...)
+	}
+	return cloned
 }
 func (c *Coordinator) hasPendingWorkers() bool {

M packages/agent/internal/orchestration/coordinator_test.go +12 -2


@@ -4,6 +4,7 @@ import (
4	4		"testing"
5	5
6	6		"github.com/bnema/zut/packages/agent/subagents"
	7	+	"github.com/bnema/zut/packages/provider"
7	8		)
8	9
9	10		func TestCoordinatorSealsWorkerWaveAndWakesOnce(t *testing.T) {

@@ -24,9 +25,10 @@ func TestCoordinatorGivesQueuedUserInputPriority(t *testing.T) {
24	25		c.Apply(Event{Kind: EventManagerStarted})
25	26		c.Apply(Event{Kind: EventWorkerRegistered, WorkerID: "worker"})
26	27		c.Apply(Event{Kind: EventManagerFinished})
27		-	c.Apply(Event{Kind: EventUserInput, Text: "stop and explain"})
	28	+	image := provider.ImageBlock{MimeType: "image/png", Data: []byte("png-1")}
	29	+	c.Apply(Event{Kind: EventUserInput, Text: "stop and explain", Images: []provider.ImageBlock{image}})
28	30		assertActions(t, c.Apply(Event{Kind: EventWorkerFinished, WorkerID: "worker", Completion: subagents.Completion{AgentID: "worker", Status: "failed"}}), []Action{{
29		-	Kind: ActionRunManager, Reason: WakeUser, Text: "stop and explain",
	31	+	Kind: ActionRunManager, Reason: WakeUser, Text: "stop and explain", Images: []provider.ImageBlock{image},
30	32		Completions: []subagents.Completion{{AgentID: "worker", Status: "failed"}},
31	33		}})
32	34		}

@@ -55,6 +57,14 @@ func assertActions(t *testing.T, got Result, want []Action) {
55	57		if got.Actions[i].Kind != want[i].Kind \|\| got.Actions[i].Reason != want[i].Reason \|\| got.Actions[i].Text != want[i].Text {
56	58		t.Fatalf("action %d = %#v, want %#v", i, got.Actions[i], want[i])
57	59		}
	60	+	if len(got.Actions[i].Images) != len(want[i].Images) {
	61	+	t.Fatalf("action %d images = %#v, want %#v", i, got.Actions[i].Images, want[i].Images)
	62	+	}
	63	+	for j := range want[i].Images {
	64	+	if got.Actions[i].Images[j].MimeType != want[i].Images[j].MimeType \|\| string(got.Actions[i].Images[j].Data) != string(want[i].Images[j].Data) {
	65	+	t.Fatalf("action %d image %d = %#v, want %#v", i, j, got.Actions[i].Images[j], want[i].Images[j])
	66	+	}
	67	+	}
58	68		if len(got.Actions[i].Completions) != len(want[i].Completions) {
59	69		t.Fatalf("action %d completions = %#v, want %#v", i, got.Actions[i].Completions, want[i].Completions)
60	70		}

 	"testing"
 	"github.com/bnema/zut/packages/agent/subagents"
+	"github.com/bnema/zut/packages/provider"
 )
 func TestCoordinatorSealsWorkerWaveAndWakesOnce(t *testing.T) {
 	c.Apply(Event{Kind: EventManagerStarted})
 	c.Apply(Event{Kind: EventWorkerRegistered, WorkerID: "worker"})
 	c.Apply(Event{Kind: EventManagerFinished})
-	c.Apply(Event{Kind: EventUserInput, Text: "stop and explain"})
+	image := provider.ImageBlock{MimeType: "image/png", Data: []byte("png-1")}
+	c.Apply(Event{Kind: EventUserInput, Text: "stop and explain", Images: []provider.ImageBlock{image}})
 	assertActions(t, c.Apply(Event{Kind: EventWorkerFinished, WorkerID: "worker", Completion: subagents.Completion{AgentID: "worker", Status: "failed"}}), []Action{{
-		Kind: ActionRunManager, Reason: WakeUser, Text: "stop and explain",
+		Kind: ActionRunManager, Reason: WakeUser, Text: "stop and explain", Images: []provider.ImageBlock{image},
 		Completions: []subagents.Completion{{AgentID: "worker", Status: "failed"}},
 	}})
 }
 		if got.Actions[i].Kind != want[i].Kind || got.Actions[i].Reason != want[i].Reason || got.Actions[i].Text != want[i].Text {
 			t.Fatalf("action %d = %#v, want %#v", i, got.Actions[i], want[i])
 		}
+		if len(got.Actions[i].Images) != len(want[i].Images) {
+			t.Fatalf("action %d images = %#v, want %#v", i, got.Actions[i].Images, want[i].Images)
+		}
+		for j := range want[i].Images {
+			if got.Actions[i].Images[j].MimeType != want[i].Images[j].MimeType || string(got.Actions[i].Images[j].Data) != string(want[i].Images[j].Data) {
+				t.Fatalf("action %d image %d = %#v, want %#v", i, j, got.Actions[i].Images[j], want[i].Images[j])
+			}
+		}
 		if len(got.Actions[i].Completions) != len(want[i].Completions) {
 			t.Fatalf("action %d completions = %#v, want %#v", i, got.Actions[i].Completions, want[i].Completions)
 		}

M packages/agent/modes/clipboard_images.go +27 -0

@@ -1,8 +1,10 @@
1	1		package modes
2	2
3	3		import (
	4	+	"fmt"
4	5		"strings"
5	6
	7	+	"github.com/bnema/zut/packages/core"
6	8		"github.com/bnema/zut/packages/provider"
7	9		)
8	10

@@ -28,6 +30,31 @@ func preparePromptWithClipboardImages(text string, pending []clipboardImageAttac
28	30		return strings.TrimSpace(out), images
29	31		}
30	32
	33	+	func (i *Interactive) restoreQueuedMessageToEditor(message core.QueuedMessage) bool {
	34	+	if message.Text == "" && len(message.Images) == 0 {
	35	+	return false
	36	+	}
	37	+	text := message.Text
	38	+	attachments := make([]clipboardImageAttachment, 0, len(message.Images))
	39	+	for index, image := range message.Images {
	40	+	markerIndex := index + 1
	41	+	marker := fmt.Sprintf("[clipboard image #%d]", markerIndex)
	42	+	for strings.Contains(text, marker) {
	43	+	markerIndex++
	44	+	marker = fmt.Sprintf("[clipboard image #%d]", markerIndex)
	45	+	}
	46	+	if text != "" {
	47	+	text += " "
	48	+	}
	49	+	text += marker
	50	+	attachments = append(attachments, clipboardImageAttachment{Marker: marker, Image: image})
	51	+	}
	52	+	i.clipboardImages = attachments
	53	+	i.ed.SetValue(text)
	54	+	i.inputHistoryIndex = -1
	55	+	return true
	56	+	}
	57	+
31	58		func removeClipboardMarker(text, marker string) string {
32	59		for {
33	60		idx := strings.Index(text, marker)

@@ -1,8 +1,10 @@
1	package modes	1		package modes
2		2
3	import (	3		import (
		4	+	"fmt"
4	"strings"	5		"strings"
5		6
		7	+	"github.com/bnema/zut/packages/core"
6	"github.com/bnema/zut/packages/provider"	8		"github.com/bnema/zut/packages/provider"
7	)	9		)
8		10

@@ -28,6 +30,31 @@ func preparePromptWithClipboardImages(text string, pending []clipboardImageAttac
28	return strings.TrimSpace(out), images	30		return strings.TrimSpace(out), images
29	}	31		}
30		32
		33	+	func (i *Interactive) restoreQueuedMessageToEditor(message core.QueuedMessage) bool {
		34	+	if message.Text == "" && len(message.Images) == 0 {
		35	+	return false
		36	+	}
		37	+	text := message.Text
		38	+	attachments := make([]clipboardImageAttachment, 0, len(message.Images))
		39	+	for index, image := range message.Images {
		40	+	markerIndex := index + 1
		41	+	marker := fmt.Sprintf("[clipboard image #%d]", markerIndex)
		42	+	for strings.Contains(text, marker) {
		43	+	markerIndex++
		44	+	marker = fmt.Sprintf("[clipboard image #%d]", markerIndex)
		45	+	}
		46	+	if text != "" {
		47	+	text += " "
		48	+	}
		49	+	text += marker
		50	+	attachments = append(attachments, clipboardImageAttachment{Marker: marker, Image: image})
		51	+	}
		52	+	i.clipboardImages = attachments
		53	+	i.ed.SetValue(text)
		54	+	i.inputHistoryIndex = -1
		55	+	return true
		56	+	}
		57	+
31	func removeClipboardMarker(text, marker string) string {	58		func removeClipboardMarker(text, marker string) string {
32	for {	59		for {
33	idx := strings.Index(text, marker)	60		idx := strings.Index(text, marker)

M packages/agent/modes/clipboard_images_test.go +21 -0


@@ -4,6 +4,7 @@ import (
4	4		"strings"
5	5		"testing"
6	6
	7	+	"github.com/bnema/zut/packages/core"
7	8		"github.com/bnema/zut/packages/provider"
8	9		)
9	10

@@ -14,6 +15,26 @@ func testClipboardImage(marker string, data string) clipboardImageAttachment {
14	15		}
15	16		}
16	17
	18	+	func TestRestoredQueuedImageDoesNotConsumeLiteralClipboardMarker(t *testing.T) {
	19	+	i := NewInteractive(InteractiveConfig{})
	20	+	message := core.QueuedMessage{
	21	+	Text: "keep literal [clipboard image #1]",
	22	+	Images: []provider.ImageBlock{{MimeType: "image/png", Data: []byte("png-1")}},
	23	+	}
	24	+
	25	+	if !i.restoreQueuedMessageToEditor(message) {
	26	+	t.Fatal("queued message was not restored")
	27	+	}
	28	+	text, images := preparePromptWithClipboardImages(i.ed.Value(), i.clipboardImages)
	29	+
	30	+	if text != message.Text {
	31	+	t.Fatalf("resubmitted text = %q, want %q", text, message.Text)
	32	+	}
	33	+	if len(images) != 1 \|\| string(images[0].Data) != "png-1" {
	34	+	t.Fatalf("resubmitted images = %#v, want png-1", images)
	35	+	}
	36	+	}
	37	+
17	38		func TestPreparePromptWithClipboardImagesStripsPresentMarker(t *testing.T) {
18	39		pending := []clipboardImageAttachment{testClipboardImage("[clipboard image #1]", "png-1")}
19	40


@@ -4,6 +4,7 @@ import (
4	"strings"	4		"strings"
5	"testing"	5		"testing"
6		6
		7	+	"github.com/bnema/zut/packages/core"
7	"github.com/bnema/zut/packages/provider"	8		"github.com/bnema/zut/packages/provider"
8	)	9		)
9		10

@@ -14,6 +15,26 @@ func testClipboardImage(marker string, data string) clipboardImageAttachment {
14	}	15		}
15	}	16		}
16		17
		18	+	func TestRestoredQueuedImageDoesNotConsumeLiteralClipboardMarker(t *testing.T) {
		19	+	i := NewInteractive(InteractiveConfig{})
		20	+	message := core.QueuedMessage{
		21	+	Text: "keep literal [clipboard image #1]",
		22	+	Images: []provider.ImageBlock{{MimeType: "image/png", Data: []byte("png-1")}},
		23	+	}
		24	+
		25	+	if !i.restoreQueuedMessageToEditor(message) {
		26	+	t.Fatal("queued message was not restored")
		27	+	}
		28	+	text, images := preparePromptWithClipboardImages(i.ed.Value(), i.clipboardImages)
		29	+
		30	+	if text != message.Text {
		31	+	t.Fatalf("resubmitted text = %q, want %q", text, message.Text)
		32	+	}
		33	+	if len(images) != 1 \|\| string(images[0].Data) != "png-1" {
		34	+	t.Fatalf("resubmitted images = %#v, want png-1", images)
		35	+	}
		36	+	}
		37	+
17	func TestPreparePromptWithClipboardImagesStripsPresentMarker(t *testing.T) {	38		func TestPreparePromptWithClipboardImagesStripsPresentMarker(t *testing.T) {
18	pending := []clipboardImageAttachment{testClipboardImage("[clipboard image #1]", "png-1")}	39		pending := []clipboardImageAttachment{testClipboardImage("[clipboard image #1]", "png-1")}
19		40

M packages/agent/modes/compact_queue_test.go +2 -2


@@ -543,7 +543,7 @@ func TestRestoredHandoffReservesTurnBeforePersistence(t *testing.T) {
543	543		})
544	544		interactive.runCtx = context.Background()
545	545		interactive.mu.Lock()
546		-	interactive.queued = []string{"newer user request"}
	546	+	interactive.queued = []core.QueuedMessage{{Text: "newer user request"}}
547	547		interactive.mu.Unlock()
548	548
549	549		done := make(chan struct{})

@@ -631,7 +631,7 @@ func TestRestoredHandoffDefersToQueuedUserPrompt(t *testing.T) {
631	631		})
632	632		interactive.runCtx = context.Background()
633	633		interactive.mu.Lock()
634		-	interactive.queued = []string{"newer user request"}
	634	+	interactive.queued = []core.QueuedMessage{{Text: "newer user request"}}
635	635		interactive.mu.Unlock()
636	636
637	637		interactive.startRestoredCompactHandoff(context.Background())


@@ -543,7 +543,7 @@ func TestRestoredHandoffReservesTurnBeforePersistence(t *testing.T) {
543		})	543		})
544		interactive.runCtx = context.Background()	544		interactive.runCtx = context.Background()
545		interactive.mu.Lock()	545		interactive.mu.Lock()
546	-	interactive.queued = []string{"newer user request"}	546	+	interactive.queued = []core.QueuedMessage{{Text: "newer user request"}}
547		interactive.mu.Unlock()	547		interactive.mu.Unlock()
548			548
549		done := make(chan struct{})	549		done := make(chan struct{})

@@ -631,7 +631,7 @@ func TestRestoredHandoffDefersToQueuedUserPrompt(t *testing.T) {
631		})	631		})
632		interactive.runCtx = context.Background()	632		interactive.runCtx = context.Background()
633		interactive.mu.Lock()	633		interactive.mu.Lock()
634	-	interactive.queued = []string{"newer user request"}	634	+	interactive.queued = []core.QueuedMessage{{Text: "newer user request"}}
635		interactive.mu.Unlock()	635		interactive.mu.Unlock()
636			636
637		interactive.startRestoredCompactHandoff(context.Background())	637		interactive.startRestoredCompactHandoff(context.Background())

M packages/agent/modes/goal.go +2 -2


@@ -100,7 +100,7 @@ func (i *Interactive) startGoalContinuation(parent context.Context, message prov
100	100		i.queued = i.queued[1:]
101	101		i.mu.Unlock()
102	102		i.abandonGoalRun(run)
103		-	i.startTurn(parent, next)
	103	+	i.startTurnWithImages(parent, next.Text, next.Images)
104	104		return
105	105		}
106	106		if ag.QueuedMessageCount() > 0 {

@@ -134,7 +134,7 @@ func (i *Interactive) startGoalContinuation(parent context.Context, message prov
134	134		i.queued = i.queued[1:]
135	135		i.mu.Unlock()
136	136		i.abandonGoalRun(run)
137		-	i.startTurn(parent, next)
	137	+	i.startTurnWithImages(parent, next.Text, next.Images)
138	138		return
139	139		}
140	140		i.mu.Unlock()


@@ -100,7 +100,7 @@ func (i *Interactive) startGoalContinuation(parent context.Context, message prov
100		i.queued = i.queued[1:]	100		i.queued = i.queued[1:]
101		i.mu.Unlock()	101		i.mu.Unlock()
102		i.abandonGoalRun(run)	102		i.abandonGoalRun(run)
103	-	i.startTurn(parent, next)	103	+	i.startTurnWithImages(parent, next.Text, next.Images)
104		return	104		return
105		}	105		}
106		if ag.QueuedMessageCount() > 0 {	106		if ag.QueuedMessageCount() > 0 {

@@ -134,7 +134,7 @@ func (i *Interactive) startGoalContinuation(parent context.Context, message prov
134		i.queued = i.queued[1:]	134		i.queued = i.queued[1:]
135		i.mu.Unlock()	135		i.mu.Unlock()
136		i.abandonGoalRun(run)	136		i.abandonGoalRun(run)
137	-	i.startTurn(parent, next)	137	+	i.startTurnWithImages(parent, next.Text, next.Images)
138		return	138		return
139		}	139		}
140		i.mu.Unlock()	140		i.mu.Unlock()

M packages/agent/modes/interactive.go +6 -9


@@ -733,7 +733,7 @@ type Interactive struct {
733	733		// Messages typed while a turn is in flight. Each is delivered as
734	734		// its own follow-up turn once the current one finishes. Rendered
735	735		// above the status bar as "sliding in: ..." chips.
736		-	queued []string
	736	+	queued []core.QueuedMessage
737	737
738	738		// scheduled holds prompt-only follow-ups injected by the in-process
739	739		// scheduler. Unlike agent.QueueMessage, these always wait until the

@@ -1533,13 +1533,10 @@ const ctrlCExitWindow = 2 * time.Second
1533	1533		// agent cancel the active turn first via the same path the editor
1534	1534		// uses for typed slash commands.
1535	1535
1536		-	// SubmitOrQueue runs text immediately if the agent is idle, or
1537		-	// appends it to the pending queue if a turn is already in flight.
1538		-	// Used by the telegram bridge (and by the editor submit path) so
1539		-	// both input sources share the same "queue behind an active turn"
1540		-	// semantics. Images are ignored for now — only the text prompt is
1541		-	// forwarded — because the queued-prompt path is text-only; a
1542		-	// follow-up can expand the queue entry to carry images.
	1536	+	// SubmitOrQueue runs a text-and-image prompt immediately if the agent is idle,
	1537	+	// or appends it to the pending queue if a turn is already in flight. Used by
	1538	+	// the Telegram bridge and editor submit path so both input sources share the
	1539	+	// same queue semantics.
1543	1540
1544	1541		// CancelTurn aborts the active turn if one is running. Used by the
1545	1542		// telegram bridge when the paired user sends /stop.

@@ -1667,7 +1664,7 @@ const modalBackdropDimPercent = 50
1667	1664		// model call immediately instead of just opening an empty dialog.
1668	1665
1669	1666		// submitOrQueuePrompt submits a slash command's expanded prompt immediately,
1670		-	// or queues it behind the active turn using the normal text-only queue.
	1667	+	// or queues it behind the active turn.
1671	1668
1672	1669		// openSkillsDialog opens the skill inspector. The picker reflects
1673	1670		// whatever SkillSnapshot returns at call time, so edits to a

 	// Messages typed while a turn is in flight. Each is delivered as
 	// its own follow-up turn once the current one finishes. Rendered
 	// above the status bar as "sliding in: ..." chips.
-	queued []string
+	queued []core.QueuedMessage
 	// scheduled holds prompt-only follow-ups injected by the in-process
 	// scheduler. Unlike agent.QueueMessage, these always wait until the
 // agent cancel the active turn first via the same path the editor
 // uses for typed slash commands.
-// SubmitOrQueue runs text immediately if the agent is idle, or
+// SubmitOrQueue runs a text-and-image prompt immediately if the agent is idle,
-// appends it to the pending queue if a turn is already in flight.
+// or appends it to the pending queue if a turn is already in flight. Used by
-// Used by the telegram bridge (and by the editor submit path) so
+// the Telegram bridge and editor submit path so both input sources share the
-// both input sources share the same "queue behind an active turn"
+// same queue semantics.
-// semantics. Images are ignored for now — only the text prompt is
-// forwarded — because the queued-prompt path is text-only; a
-// follow-up can expand the queue entry to carry images.
 // CancelTurn aborts the active turn if one is running. Used by the
 // telegram bridge when the paired user sends /stop.
 // model call immediately instead of just opening an empty dialog.
 // submitOrQueuePrompt submits a slash command's expanded prompt immediately,
-// or queues it behind the active turn using the normal text-only queue.
+// or queues it behind the active turn.
 // openSkillsDialog opens the skill inspector. The picker reflects
 // whatever SkillSnapshot returns at call time, so edits to a

M packages/agent/modes/interactive_commands.go +4 -4


@@ -347,12 +347,12 @@ func (i *Interactive) runSlash(ctx context.Context, cmd string) (done bool) {
347	347		if ag != nil && !compacting {
348	348		i.mu.Lock()
349	349		handoff, persistHandoff = i.resetCompactContinuationLocked()
350		-	ag.QueueMessage(studyPrompt)
	350	+	ag.QueueMessage(studyPrompt, nil)
351	351		i.mu.Unlock()
352	352		} else {
353	353		i.mu.Lock()
354	354		handoff, persistHandoff = i.resetCompactContinuationLocked()
355		-	i.queued = append(i.queued, studyPrompt)
	355	+	i.queued = append(i.queued, core.QueuedMessage{Text: studyPrompt})
356	356		i.mu.Unlock()
357	357		}
358	358		if persistHandoff {

@@ -880,10 +880,10 @@ func (i *Interactive) submitOrQueuePrompt(ctx context.Context, prompt string) {
880	880		var persistHandoff bool
881	881		if i.compacting {
882	882		handoff, persistHandoff = i.resetCompactContinuationLocked()
883		-	i.queued = append(i.queued, prompt)
	883	+	i.queued = append(i.queued, core.QueuedMessage{Text: prompt})
884	884		} else {
885	885		handoff, persistHandoff = i.resetCompactContinuationLocked()
886		-	i.agent.QueueMessage(prompt)
	886	+	i.agent.QueueMessage(prompt, nil)
887	887		}
888	888		i.mu.Unlock()
889	889		if persistHandoff {


@@ -347,12 +347,12 @@ func (i *Interactive) runSlash(ctx context.Context, cmd string) (done bool) {
347		if ag != nil && !compacting {	347		if ag != nil && !compacting {
348		i.mu.Lock()	348		i.mu.Lock()
349		handoff, persistHandoff = i.resetCompactContinuationLocked()	349		handoff, persistHandoff = i.resetCompactContinuationLocked()
350	-	ag.QueueMessage(studyPrompt)	350	+	ag.QueueMessage(studyPrompt, nil)
351		i.mu.Unlock()	351		i.mu.Unlock()
352		} else {	352		} else {
353		i.mu.Lock()	353		i.mu.Lock()
354		handoff, persistHandoff = i.resetCompactContinuationLocked()	354		handoff, persistHandoff = i.resetCompactContinuationLocked()
355	-	i.queued = append(i.queued, studyPrompt)	355	+	i.queued = append(i.queued, core.QueuedMessage{Text: studyPrompt})
356		i.mu.Unlock()	356		i.mu.Unlock()
357		}	357		}
358		if persistHandoff {	358		if persistHandoff {

@@ -880,10 +880,10 @@ func (i *Interactive) submitOrQueuePrompt(ctx context.Context, prompt string) {
880		var persistHandoff bool	880		var persistHandoff bool
881		if i.compacting {	881		if i.compacting {
882		handoff, persistHandoff = i.resetCompactContinuationLocked()	882		handoff, persistHandoff = i.resetCompactContinuationLocked()
883	-	i.queued = append(i.queued, prompt)	883	+	i.queued = append(i.queued, core.QueuedMessage{Text: prompt})
884		} else {	884		} else {
885		handoff, persistHandoff = i.resetCompactContinuationLocked()	885		handoff, persistHandoff = i.resetCompactContinuationLocked()
886	-	i.agent.QueueMessage(prompt)	886	+	i.agent.QueueMessage(prompt, nil)
887		}	887		}
888		i.mu.Unlock()	888		i.mu.Unlock()
889		if persistHandoff {	889		if persistHandoff {

M packages/agent/modes/interactive_compaction.go +12 -6


@@ -5,6 +5,7 @@ import (
5	5		"encoding/json"
6	6		"fmt"
7	7
	8	+	"github.com/bnema/zut/packages/core"
8	9		"github.com/bnema/zut/packages/provider"
9	10		"github.com/bnema/zut/packages/tui"
10	11		)

@@ -109,7 +110,8 @@ func (i *Interactive) startRestoredCompactHandoff(parent context.Context) {
109	110		return
110	111		}
111	112
112		-	var next string
	113	+	var next core.QueuedMessage
	114	+	var hasNext bool
113	115		continueQueued := false
114	116		var resume compactHandoffResume
115	117		var handoff json.RawMessage

@@ -117,6 +119,7 @@ func (i *Interactive) startRestoredCompactHandoff(parent context.Context) {
117	119		switch {
118	120		case state.reason != compactContinuationForcedLength && len(i.queued) > 0:
119	121		next, i.queued = i.queued[0], i.queued[1:]
	122	+	hasNext = true
120	123		handoff, persistHandoff = i.resetCompactContinuationLocked()
121	124		case state.reason != compactContinuationForcedLength && ag.QueuedMessageCount() > 0:
122	125		continueQueued = true

@@ -129,7 +132,7 @@ func (i *Interactive) startRestoredCompactHandoff(parent context.Context) {
129	132		}
130	133		// Reserve the turn slot before persisting or dispatching so concurrent input
131	134		// can only queue behind the restored handoff, never start a competing turn.
132		-	starting := next != "" \|\| continueQueued \|\| resume == compactHandoffContinueExisting \|\| resume == compactHandoffAppendPrompt
	135	+	starting := hasNext \|\| continueQueued \|\| resume == compactHandoffContinueExisting \|\| resume == compactHandoffAppendPrompt
133	136		i.busy = starting
134	137		i.mu.Unlock()
135	138		if persistHandoff {

@@ -140,8 +143,8 @@ func (i *Interactive) startRestoredCompactHandoff(parent context.Context) {
140	143		return
141	144		}
142	145		switch {
143		-	case next != "":
144		-	i.startTurn(parent, next)
	146	+	case hasNext:
	147	+	i.startTurnWithImages(parent, next.Text, next.Images)
145	148		case continueQueued:
146	149		i.startTurnRequest(parent, "", nil, true, false)
147	150		case resume == compactHandoffContinueExisting:

@@ -173,7 +176,7 @@ func (i *Interactive) startAutoCompactContinuation(parent context.Context) {
173	176		if persistHandoff {
174	177		i.persistCompactHandoff(handoff)
175	178		}
176		-	i.startTurn(parent, next)
	179	+	i.startTurnWithImages(parent, next.Text, next.Images)
177	180		return
178	181		}
179	182		if i.compactContinuation.reason != compactContinuationForcedLength && ag.QueuedMessageCount() > 0 {

@@ -338,7 +341,10 @@ func (i *Interactive) runCompact(parent context.Context, request compactContinua
338	341		handoff, persistHandoff = i.setCompactContinuationLocked(compactContinuationState{reason: continuationReason})
339	342		continueAutomatically = true
340	343		case len(i.queued) > 0:
341		-	next, i.queued = i.queued[0], i.queued[1:]
	344	+	queued := i.queued[0]
	345	+	i.queued = i.queued[1:]
	346	+	next = queued.Text
	347	+	nextImages = queued.Images
342	348		hasNext = true
343	349		handoff, persistHandoff = i.resetCompactContinuationLocked()
344	350		case continuationReason == compactContinuationStructuralTail \|\| continuationReason == compactContinuationStatusRescue:

 	"encoding/json"
 	"fmt"
+	"github.com/bnema/zut/packages/core"
 	"github.com/bnema/zut/packages/provider"
 	"github.com/bnema/zut/packages/tui"
 )
 		return
 	}
-	var next string
+	var next core.QueuedMessage
+	var hasNext bool
 	continueQueued := false
 	var resume compactHandoffResume
 	var handoff json.RawMessage
 	switch {
 	case state.reason != compactContinuationForcedLength && len(i.queued) > 0:
 		next, i.queued = i.queued[0], i.queued[1:]
+		hasNext = true
 		handoff, persistHandoff = i.resetCompactContinuationLocked()
 	case state.reason != compactContinuationForcedLength && ag.QueuedMessageCount() > 0:
 		continueQueued = true
 	}
 	// Reserve the turn slot before persisting or dispatching so concurrent input
 	// can only queue behind the restored handoff, never start a competing turn.
-	starting := next != "" || continueQueued || resume == compactHandoffContinueExisting || resume == compactHandoffAppendPrompt
+	starting := hasNext || continueQueued || resume == compactHandoffContinueExisting || resume == compactHandoffAppendPrompt
 	i.busy = starting
 	i.mu.Unlock()
 	if persistHandoff {
 		return
 	}
 	switch {
-	case next != "":
+	case hasNext:
-		i.startTurn(parent, next)
+		i.startTurnWithImages(parent, next.Text, next.Images)
 	case continueQueued:
 		i.startTurnRequest(parent, "", nil, true, false)
 	case resume == compactHandoffContinueExisting:
 		if persistHandoff {
 			i.persistCompactHandoff(handoff)
 		}
-		i.startTurn(parent, next)
+		i.startTurnWithImages(parent, next.Text, next.Images)
 		return
 	}
 	if i.compactContinuation.reason != compactContinuationForcedLength && ag.QueuedMessageCount() > 0 {
 				handoff, persistHandoff = i.setCompactContinuationLocked(compactContinuationState{reason: continuationReason})
 				continueAutomatically = true
 			case len(i.queued) > 0:
-				next, i.queued = i.queued[0], i.queued[1:]
+				queued := i.queued[0]
+				i.queued = i.queued[1:]
+				next = queued.Text
+				nextImages = queued.Images
 				hasNext = true
 				handoff, persistHandoff = i.resetCompactContinuationLocked()
 			case continuationReason == compactContinuationStructuralTail || continuationReason == compactContinuationStatusRescue:

M packages/agent/modes/interactive_host.go +4 -5


@@ -256,7 +256,7 @@ func (i *Interactive) submitOrQueue(text string, images []provider.ImageBlock, u
256	256		// the manager is dormant behind a sealed worker wave, the coordinator makes
257	257		// this the next manager turn and carries any completed worker summary with it.
258	258		if userInput && i.coordinatorAcceptsUserInput() {
259		-	actions := i.applyCoordinator(orchestration.Event{Kind: orchestration.EventUserInput, Text: text})
	259	+	actions := i.applyCoordinator(orchestration.Event{Kind: orchestration.EventUserInput, Text: text, Images: images})
260	260		i.executeCoordinatorActions(actions)
261	261		// The coordinator has accepted the input for its sealed worker wave.
262	262		// It may intentionally defer an action until pending workers finish;

@@ -266,8 +266,7 @@ func (i *Interactive) submitOrQueue(text string, images []provider.ImageBlock, u
266	266		i.maybeStartSessionTitle(i.runCtx, text)
267	267		i.mu.Lock()
268	268		if i.busy {
269		-	// Queue text only; images are dropped for queued prompts. Keep the
270		-	// interactive mutex held through Agent.QueueMessage so turn teardown
	269	+	// Keep the interactive mutex held through Agent.QueueMessage so turn teardown
271	270		// cannot inspect the agent queue between the busy check and enqueue.
272	271		// Compaction uses the host queue because it has no active agent loop
273	272		// to drain Agent.QueueMessage entries.

@@ -275,10 +274,10 @@ func (i *Interactive) submitOrQueue(text string, images []provider.ImageBlock, u
275	274		var persistHandoff bool
276	275		if i.agent != nil && !i.compacting {
277	276		handoff, persistHandoff = i.resetCompactContinuationLocked()
278		-	i.agent.QueueMessage(text)
	277	+	i.agent.QueueMessage(text, images)
279	278		} else {
280	279		handoff, persistHandoff = i.resetCompactContinuationLocked()
281		-	i.queued = append(i.queued, text)
	280	+	i.queued = append(i.queued, core.QueuedMessage{Text: text, Images: images})
282	281		}
283	282		i.mu.Unlock()
284	283		if persistHandoff {

 	// the manager is dormant behind a sealed worker wave, the coordinator makes
 	// this the next manager turn and carries any completed worker summary with it.
 	if userInput && i.coordinatorAcceptsUserInput() {
-		actions := i.applyCoordinator(orchestration.Event{Kind: orchestration.EventUserInput, Text: text})
+		actions := i.applyCoordinator(orchestration.Event{Kind: orchestration.EventUserInput, Text: text, Images: images})
 		i.executeCoordinatorActions(actions)
 		// The coordinator has accepted the input for its sealed worker wave.
 		// It may intentionally defer an action until pending workers finish;
 	i.maybeStartSessionTitle(i.runCtx, text)
 	i.mu.Lock()
 	if i.busy {
-		// Queue text only; images are dropped for queued prompts. Keep the
+		// Keep the interactive mutex held through Agent.QueueMessage so turn teardown
-		// interactive mutex held through Agent.QueueMessage so turn teardown
 		// cannot inspect the agent queue between the busy check and enqueue.
 		// Compaction uses the host queue because it has no active agent loop
 		// to drain Agent.QueueMessage entries.
 		var persistHandoff bool
 		if i.agent != nil && !i.compacting {
 			handoff, persistHandoff = i.resetCompactContinuationLocked()
-			i.agent.QueueMessage(text)
+			i.agent.QueueMessage(text, images)
 		} else {
 			handoff, persistHandoff = i.resetCompactContinuationLocked()
-			i.queued = append(i.queued, text)
+			i.queued = append(i.queued, core.QueuedMessage{Text: text, Images: images})
 		}
 		i.mu.Unlock()
 		if persistHandoff {

M packages/agent/modes/interactive_host_test.go +20 -1


@@ -6,15 +6,31 @@ import (
6	6		"github.com/bnema/zut/packages/agent/internal/orchestration"
7	7		"github.com/bnema/zut/packages/agent/subagents"
8	8		"github.com/bnema/zut/packages/core"
	9	+	"github.com/bnema/zut/packages/provider"
9	10		)
10	11
	12	+	func TestSubmitOrQueuePreservesImagesInHostQueue(t *testing.T) {
	13	+	i := &Interactive{agent: &core.Agent{}, busy: true, compacting: true}
	14	+	image := provider.ImageBlock{MimeType: "image/png", Data: []byte("png-1")}
	15	+
	16	+	i.submitOrQueue("inspect", []provider.ImageBlock{image}, false)
	17	+
	18	+	if len(i.queued) != 1 \|\| i.queued[0].Text != "inspect" {
	19	+	t.Fatalf("queued messages = %#v, want one inspect prompt", i.queued)
	20	+	}
	21	+	if len(i.queued[0].Images) != 1 \|\| string(i.queued[0].Images[0].Data) != "png-1" {
	22	+	t.Fatalf("queued images = %#v, want png-1", i.queued[0].Images)
	23	+	}
	24	+	}
	25	+
11	26		func TestSubmitOrQueuePreservesSealedWorkerWaveWithPendingWorker(t *testing.T) {
12	27		i := &Interactive{agent: &core.Agent{}}
13	28		i.applyCoordinator(orchestration.Event{Kind: orchestration.EventManagerStarted})
14	29		i.applyCoordinator(orchestration.Event{Kind: orchestration.EventWorkerRegistered, WorkerID: "worker#1"})
15	30		i.applyCoordinator(orchestration.Event{Kind: orchestration.EventManagerFinished})
16	31
17		-	i.submitOrQueue("review the result", nil, true)
	32	+	image := provider.ImageBlock{MimeType: "image/png", Data: []byte("png-1")}
	33	+	i.submitOrQueue("review the result", []provider.ImageBlock{image}, true)
18	34
19	35		actions := i.applyCoordinator(orchestration.Event{
20	36		Kind: orchestration.EventWorkerFinished,

@@ -30,4 +46,7 @@ func TestSubmitOrQueuePreservesSealedWorkerWaveWithPendingWorker(t *testing.T) {
30	46		if got := actions[0].Text; got != "review the result" {
31	47		t.Fatalf("completion action text = %q, want queued user input", got)
32	48		}
	49	+	if len(actions[0].Images) != 1 \|\| string(actions[0].Images[0].Data) != "png-1" {
	50	+	t.Fatalf("completion action images = %#v, want queued image", actions[0].Images)
	51	+	}
33	52		}

 	"github.com/bnema/zut/packages/agent/internal/orchestration"
 	"github.com/bnema/zut/packages/agent/subagents"
 	"github.com/bnema/zut/packages/core"
+	"github.com/bnema/zut/packages/provider"
 )
+func TestSubmitOrQueuePreservesImagesInHostQueue(t *testing.T) {
+	i := &Interactive{agent: &core.Agent{}, busy: true, compacting: true}
+	image := provider.ImageBlock{MimeType: "image/png", Data: []byte("png-1")}
+	i.submitOrQueue("inspect", []provider.ImageBlock{image}, false)
+	if len(i.queued) != 1 || i.queued[0].Text != "inspect" {
+		t.Fatalf("queued messages = %#v, want one inspect prompt", i.queued)
+	}
+	if len(i.queued[0].Images) != 1 || string(i.queued[0].Images[0].Data) != "png-1" {
+		t.Fatalf("queued images = %#v, want png-1", i.queued[0].Images)
+	}
+}
 func TestSubmitOrQueuePreservesSealedWorkerWaveWithPendingWorker(t *testing.T) {
 	i := &Interactive{agent: &core.Agent{}}
 	i.applyCoordinator(orchestration.Event{Kind: orchestration.EventManagerStarted})
 	i.applyCoordinator(orchestration.Event{Kind: orchestration.EventWorkerRegistered, WorkerID: "worker#1"})
 	i.applyCoordinator(orchestration.Event{Kind: orchestration.EventManagerFinished})
-	i.submitOrQueue("review the result", nil, true)
+	image := provider.ImageBlock{MimeType: "image/png", Data: []byte("png-1")}
+	i.submitOrQueue("review the result", []provider.ImageBlock{image}, true)
 	actions := i.applyCoordinator(orchestration.Event{
 		Kind:     orchestration.EventWorkerFinished,
 	if got := actions[0].Text; got != "review the result" {
 		t.Fatalf("completion action text = %q, want queued user input", got)
 	}
+	if len(actions[0].Images) != 1 || string(actions[0].Images[0].Data) != "png-1" {
+		t.Fatalf("completion action images = %#v, want queued image", actions[0].Images)
+	}
 }

M packages/agent/modes/interactive_input_routing.go +16 -22


@@ -578,27 +578,28 @@ func (i *Interactive) handleKey(ctx context.Context, k tui.Key) (done bool) {
578	578		i.mu.Lock()
579	579		busyCancel := i.busy && i.cancelTurn != nil
580	580		cancelTurn := i.cancelTurn
581		-	var restoreQueued string
	581	+	var restoreQueued core.QueuedMessage
	582	+	var hasRestoreQueued bool
582	583		var handoff json.RawMessage
583	584		var persistHandoff bool
584	585		if busyCancel {
585	586		// Keep the most recently queued follow-up as an editable draft
586	587		// instead of losing it with the cancelled turn's stale queue.
587	588		if i.agent != nil {
588		-	restoreQueued, _ = i.agent.PopQueuedMessage()
	589	+	restoreQueued, hasRestoreQueued = i.agent.PopQueuedMessage()
589	590		}
590		-	if restoreQueued == "" && len(i.queued) > 0 {
	591	+	if !hasRestoreQueued && len(i.queued) > 0 {
591	592		n := len(i.queued) - 1
592	593		restoreQueued = i.queued[n]
593	594		i.queued = i.queued[:n]
	595	+	hasRestoreQueued = true
594	596		}
595	597		handoff, persistHandoff = i.resetCompactContinuationLocked()
596	598		}
597	599		i.mu.Unlock()
598	600		if busyCancel {
599		-	if restoreQueued != "" {
600		-	i.ed.SetValue(restoreQueued)
601		-	i.inputHistoryIndex = -1
	601	+	if hasRestoreQueued {
	602	+	i.restoreQueuedMessageToEditor(restoreQueued)
602	603		}
603	604		i.updateActiveGoal(core.GoalPaused, "interrupted by user")
604	605		cancelTurn()

@@ -647,20 +648,20 @@ func (i *Interactive) handleKey(ctx context.Context, k tui.Key) (done bool) {
647	648		// the keypress falls through to the normal scroll behavior.
648	649		if k.Alt {
649	650		i.mu.Lock()
650		-	var text string
	651	+	var message core.QueuedMessage
	652	+	var ok bool
651	653		if i.agent != nil {
652		-	text, _ = i.agent.PopQueuedMessage()
	654	+	message, ok = i.agent.PopQueuedMessage()
653	655		}
654		-	if text == "" {
	656	+	if !ok {
655	657		if n := len(i.queued); n > 0 {
656		-	text = i.queued[n-1]
	658	+	message = i.queued[n-1]
657	659		i.queued = i.queued[:n-1]
	660	+	ok = true
658	661		}
659	662		}
660	663		i.mu.Unlock()
661		-	if text != "" {
662		-	i.ed.SetValue(text)
663		-	i.inputHistoryIndex = -1
	664	+	if ok && i.restoreQueuedMessageToEditor(message) {
664	665		i.invalidate()
665	666		return false
666	667		}

@@ -914,25 +915,18 @@ func (i *Interactive) handleKey(ctx context.Context, k tui.Key) (done bool) {
914	915		ag := i.agent
915	916		i.mu.Unlock()
916	917		if busy {
917		-	if len(images) > 0 {
918		-	i.mu.Lock()
919		-	i.statusErr = "can't queue clipboard images while a turn is running; wait for the current turn to finish"
920		-	i.mu.Unlock()
921		-	i.invalidate()
922		-	return false
923		-	}
924	918		clearSubmittedInput()
925	919		var handoff json.RawMessage
926	920		var persistHandoff bool
927	921		if ag != nil && !compacting {
928	922		i.mu.Lock()
929	923		handoff, persistHandoff = i.resetCompactContinuationLocked()
930		-	ag.QueueMessage(text)
	924	+	ag.QueueMessage(text, images)
931	925		i.mu.Unlock()
932	926		} else {
933	927		i.mu.Lock()
934	928		handoff, persistHandoff = i.resetCompactContinuationLocked()
935		-	i.queued = append(i.queued, text)
	929	+	i.queued = append(i.queued, core.QueuedMessage{Text: text, Images: images})
936	930		i.mu.Unlock()
937	931		}
938	932		if persistHandoff {

 		i.mu.Lock()
 		busyCancel := i.busy && i.cancelTurn != nil
 		cancelTurn := i.cancelTurn
-		var restoreQueued string
+		var restoreQueued core.QueuedMessage
+		var hasRestoreQueued bool
 		var handoff json.RawMessage
 		var persistHandoff bool
 		if busyCancel {
 			// Keep the most recently queued follow-up as an editable draft
 			// instead of losing it with the cancelled turn's stale queue.
 			if i.agent != nil {
-				restoreQueued, _ = i.agent.PopQueuedMessage()
+				restoreQueued, hasRestoreQueued = i.agent.PopQueuedMessage()
 			}
-			if restoreQueued == "" && len(i.queued) > 0 {
+			if !hasRestoreQueued && len(i.queued) > 0 {
 				n := len(i.queued) - 1
 				restoreQueued = i.queued[n]
 				i.queued = i.queued[:n]
+				hasRestoreQueued = true
 			}
 			handoff, persistHandoff = i.resetCompactContinuationLocked()
 		}
 		i.mu.Unlock()
 		if busyCancel {
-			if restoreQueued != "" {
+			if hasRestoreQueued {
-				i.ed.SetValue(restoreQueued)
+				i.restoreQueuedMessageToEditor(restoreQueued)
-				i.inputHistoryIndex = -1
 			}
 			i.updateActiveGoal(core.GoalPaused, "interrupted by user")
 			cancelTurn()
 		// the keypress falls through to the normal scroll behavior.
 		if k.Alt {
 			i.mu.Lock()
-			var text string
+			var message core.QueuedMessage
+			var ok bool
 			if i.agent != nil {
-				text, _ = i.agent.PopQueuedMessage()
+				message, ok = i.agent.PopQueuedMessage()
 			}
-			if text == "" {
+			if !ok {
 				if n := len(i.queued); n > 0 {
-					text = i.queued[n-1]
+					message = i.queued[n-1]
 					i.queued = i.queued[:n-1]
+					ok = true
 				}
 			}
 			i.mu.Unlock()
-			if text != "" {
+			if ok && i.restoreQueuedMessageToEditor(message) {
-				i.ed.SetValue(text)
-				i.inputHistoryIndex = -1
 				i.invalidate()
 				return false
 			}
 		ag := i.agent
 		i.mu.Unlock()
 		if busy {
-			if len(images) > 0 {
-				i.mu.Lock()
-				i.statusErr = "can't queue clipboard images while a turn is running; wait for the current turn to finish"
-				i.mu.Unlock()
-				i.invalidate()
-				return false
-			}
 			clearSubmittedInput()
 			var handoff json.RawMessage
 			var persistHandoff bool
 			if ag != nil && !compacting {
 				i.mu.Lock()
 				handoff, persistHandoff = i.resetCompactContinuationLocked()
-				ag.QueueMessage(text)
+				ag.QueueMessage(text, images)
 				i.mu.Unlock()
 			} else {
 				i.mu.Lock()
 				handoff, persistHandoff = i.resetCompactContinuationLocked()
-				i.queued = append(i.queued, text)
+				i.queued = append(i.queued, core.QueuedMessage{Text: text, Images: images})
 				i.mu.Unlock()
 			}
 			if persistHandoff {

M packages/agent/modes/interactive_queue_test.go +64 -4


@@ -2,16 +2,76 @@ package modes
2	2
3	3		import (
4	4		"context"
	5	+	"strings"
5	6		"testing"
6	7
7	8		"github.com/bnema/zut/packages/core"
	9	+	"github.com/bnema/zut/packages/provider"
8	10		"github.com/bnema/zut/packages/tui"
9	11		)
10	12
	13	+	func TestQueuedMessageSummaryKeepsImageIndicatorWhenTextIsTruncated(t *testing.T) {
	14	+	message := core.QueuedMessage{
	15	+	Text: strings.Repeat("x", 100),
	16	+	Images: []provider.ImageBlock{{MimeType: "image/png", Data: []byte("png-1")}},
	17	+	}
	18	+
	19	+	summary := queuedMessageSummary(message, 24)
	20	+
	21	+	if !strings.HasSuffix(summary, " [image]") {
	22	+	t.Fatalf("summary = %q, want visible image indicator", summary)
	23	+	}
	24	+	if len([]rune(summary)) > 24 {
	25	+	t.Fatalf("summary width = %d, want at most 24", len([]rune(summary)))
	26	+	}
	27	+	}
	28	+
	29	+	func TestBusySubmitQueuesClipboardImagePrompt(t *testing.T) {
	30	+	agent := core.NewAgent(nil, "test-model", "", nil)
	31	+	i := NewInteractive(InteractiveConfig{Agent: agent})
	32	+	i.mu.Lock()
	33	+	i.busy = true
	34	+	i.mu.Unlock()
	35	+	i.ed.SetValue("inspect [clipboard image #1]")
	36	+	i.clipboardImages = []clipboardImageAttachment{testClipboardImage("[clipboard image #1]", "png-1")}
	37	+
	38	+	i.handleKey(context.Background(), tui.Key{Kind: tui.KeyEnter})
	39	+
	40	+	queued := agent.PendingQueuedMessages()
	41	+	if len(queued) != 1 \|\| queued[0].Text != "inspect" {
	42	+	t.Fatalf("queued messages = %#v, want one inspect prompt", queued)
	43	+	}
	44	+	if len(queued[0].Images) != 1 \|\| string(queued[0].Images[0].Data) != "png-1" {
	45	+	t.Fatalf("queued images = %#v, want png-1", queued[0].Images)
	46	+	}
	47	+	if !i.ed.IsEmpty() \|\| len(i.clipboardImages) != 0 {
	48	+	t.Fatal("submitted image prompt remained in the editor")
	49	+	}
	50	+	}
	51	+
	52	+	func TestSlideBackRestoresQueuedImagesToEditor(t *testing.T) {
	53	+	agent := core.NewAgent(nil, "test-model", "", nil)
	54	+	image := testClipboardImage("unused", "png-1").Image
	55	+	agent.QueueMessage("inspect", []provider.ImageBlock{image})
	56	+	i := NewInteractive(InteractiveConfig{Agent: agent})
	57	+	i.mu.Lock()
	58	+	i.busy = true
	59	+	i.mu.Unlock()
	60	+
	61	+	i.handleKey(context.Background(), tui.Key{Kind: tui.KeyUp, Alt: true})
	62	+
	63	+	if got := i.ed.Value(); got != "inspect [clipboard image #1]" {
	64	+	t.Fatalf("editor = %q, want restored image marker", got)
	65	+	}
	66	+	if len(i.clipboardImages) != 1 \|\| string(i.clipboardImages[0].Image.Data) != "png-1" {
	67	+	t.Fatalf("clipboard images = %#v, want png-1", i.clipboardImages)
	68	+	}
	69	+	}
	70	+
11	71		func TestEscapeRestoresMostRecentQueuedMessageToEditor(t *testing.T) {
12	72		agent := core.NewAgent(nil, "test-model", "", nil)
13		-	agent.QueueMessage("older follow-up")
14		-	agent.QueueMessage("recover this draft")
	73	+	agent.QueueMessage("older follow-up", nil)
	74	+	agent.QueueMessage("recover this draft", nil)
15	75
16	76		i := NewInteractive(InteractiveConfig{Agent: agent})
17	77		i.ed.SetValue("existing draft")

@@ -30,7 +90,7 @@ func TestEscapeRestoresMostRecentQueuedMessageToEditor(t *testing.T) {
30	90		if got, want := i.ed.Value(), "recover this draft"; got != want {
31	91		t.Errorf("editor = %q, want %q", got, want)
32	92		}
33		-	if got := agent.PendingQueuedMessages(); len(got) != 1 \|\| got[0] != "older follow-up" {
	93	+	if got := agent.PendingQueuedMessages(); len(got) != 1 \|\| got[0].Text != "older follow-up" {
34	94		t.Errorf("remaining queued messages = %v, want [older follow-up]", got)
35	95		}
36	96		}

@@ -41,7 +101,7 @@ func TestEscapeRestoresHostQueuedMessageToEditor(t *testing.T) {
41	101		i.mu.Lock()
42	102		i.busy = true
43	103		i.cancelTurn = func() { cancelled++ }
44		-	i.queued = []string{"recover this draft"}
	104	+	i.queued = []core.QueuedMessage{{Text: "recover this draft"}}
45	105		i.mu.Unlock()
46	106
47	107		i.handleKey(context.Background(), tui.Key{Kind: tui.KeyEsc})


@@ -2,16 +2,76 @@ package modes
2			2
3		import (	3		import (
4		"context"	4		"context"
			5	+	"strings"
5		"testing"	6		"testing"
6			7
7		"github.com/bnema/zut/packages/core"	8		"github.com/bnema/zut/packages/core"
			9	+	"github.com/bnema/zut/packages/provider"
8		"github.com/bnema/zut/packages/tui"	10		"github.com/bnema/zut/packages/tui"
9		)	11		)
10			12
			13	+	func TestQueuedMessageSummaryKeepsImageIndicatorWhenTextIsTruncated(t *testing.T) {
			14	+	message := core.QueuedMessage{
			15	+	Text: strings.Repeat("x", 100),
			16	+	Images: []provider.ImageBlock{{MimeType: "image/png", Data: []byte("png-1")}},
			17	+	}
			18	+
			19	+	summary := queuedMessageSummary(message, 24)
			20	+
			21	+	if !strings.HasSuffix(summary, " [image]") {
			22	+	t.Fatalf("summary = %q, want visible image indicator", summary)
			23	+	}
			24	+	if len([]rune(summary)) > 24 {
			25	+	t.Fatalf("summary width = %d, want at most 24", len([]rune(summary)))
			26	+	}
			27	+	}
			28	+
			29	+	func TestBusySubmitQueuesClipboardImagePrompt(t *testing.T) {
			30	+	agent := core.NewAgent(nil, "test-model", "", nil)
			31	+	i := NewInteractive(InteractiveConfig{Agent: agent})
			32	+	i.mu.Lock()
			33	+	i.busy = true
			34	+	i.mu.Unlock()
			35	+	i.ed.SetValue("inspect [clipboard image #1]")
			36	+	i.clipboardImages = []clipboardImageAttachment{testClipboardImage("[clipboard image #1]", "png-1")}
			37	+
			38	+	i.handleKey(context.Background(), tui.Key{Kind: tui.KeyEnter})
			39	+
			40	+	queued := agent.PendingQueuedMessages()
			41	+	if len(queued) != 1 \|\| queued[0].Text != "inspect" {
			42	+	t.Fatalf("queued messages = %#v, want one inspect prompt", queued)
			43	+	}
			44	+	if len(queued[0].Images) != 1 \|\| string(queued[0].Images[0].Data) != "png-1" {
			45	+	t.Fatalf("queued images = %#v, want png-1", queued[0].Images)
			46	+	}
			47	+	if !i.ed.IsEmpty() \|\| len(i.clipboardImages) != 0 {
			48	+	t.Fatal("submitted image prompt remained in the editor")
			49	+	}
			50	+	}
			51	+
			52	+	func TestSlideBackRestoresQueuedImagesToEditor(t *testing.T) {
			53	+	agent := core.NewAgent(nil, "test-model", "", nil)
			54	+	image := testClipboardImage("unused", "png-1").Image
			55	+	agent.QueueMessage("inspect", []provider.ImageBlock{image})
			56	+	i := NewInteractive(InteractiveConfig{Agent: agent})
			57	+	i.mu.Lock()
			58	+	i.busy = true
			59	+	i.mu.Unlock()
			60	+
			61	+	i.handleKey(context.Background(), tui.Key{Kind: tui.KeyUp, Alt: true})
			62	+
			63	+	if got := i.ed.Value(); got != "inspect [clipboard image #1]" {
			64	+	t.Fatalf("editor = %q, want restored image marker", got)
			65	+	}
			66	+	if len(i.clipboardImages) != 1 \|\| string(i.clipboardImages[0].Image.Data) != "png-1" {
			67	+	t.Fatalf("clipboard images = %#v, want png-1", i.clipboardImages)
			68	+	}
			69	+	}
			70	+
11		func TestEscapeRestoresMostRecentQueuedMessageToEditor(t *testing.T) {	71		func TestEscapeRestoresMostRecentQueuedMessageToEditor(t *testing.T) {
12		agent := core.NewAgent(nil, "test-model", "", nil)	72		agent := core.NewAgent(nil, "test-model", "", nil)
13	-	agent.QueueMessage("older follow-up")	73	+	agent.QueueMessage("older follow-up", nil)
14	-	agent.QueueMessage("recover this draft")	74	+	agent.QueueMessage("recover this draft", nil)
15			75
16		i := NewInteractive(InteractiveConfig{Agent: agent})	76		i := NewInteractive(InteractiveConfig{Agent: agent})
17		i.ed.SetValue("existing draft")	77		i.ed.SetValue("existing draft")

@@ -30,7 +90,7 @@ func TestEscapeRestoresMostRecentQueuedMessageToEditor(t *testing.T) {
30		if got, want := i.ed.Value(), "recover this draft"; got != want {	90		if got, want := i.ed.Value(), "recover this draft"; got != want {
31		t.Errorf("editor = %q, want %q", got, want)	91		t.Errorf("editor = %q, want %q", got, want)
32		}	92		}
33	-	if got := agent.PendingQueuedMessages(); len(got) != 1 \|\| got[0] != "older follow-up" {	93	+	if got := agent.PendingQueuedMessages(); len(got) != 1 \|\| got[0].Text != "older follow-up" {
34		t.Errorf("remaining queued messages = %v, want [older follow-up]", got)	94		t.Errorf("remaining queued messages = %v, want [older follow-up]", got)
35		}	95		}
36		}	96		}

@@ -41,7 +101,7 @@ func TestEscapeRestoresHostQueuedMessageToEditor(t *testing.T) {
41		i.mu.Lock()	101		i.mu.Lock()
42		i.busy = true	102		i.busy = true
43		i.cancelTurn = func() { cancelled++ }	103		i.cancelTurn = func() { cancelled++ }
44	-	i.queued = []string{"recover this draft"}	104	+	i.queued = []core.QueuedMessage{{Text: "recover this draft"}}
45		i.mu.Unlock()	105		i.mu.Unlock()
46			106
47		i.handleKey(context.Background(), tui.Key{Kind: tui.KeyEsc})	107		i.handleKey(context.Background(), tui.Key{Kind: tui.KeyEsc})

M packages/agent/modes/interactive_rendering.go +21 -2


@@ -8,10 +8,29 @@ import (
8	8
9	9		"github.com/bnema/zut/packages/agent/extproto"
10	10		"github.com/bnema/zut/packages/agent/skills"
	11	+	"github.com/bnema/zut/packages/core"
11	12		"github.com/bnema/zut/packages/provider"
12	13		"github.com/bnema/zut/packages/tui"
13	14		)
14	15
	16	+	func queuedMessageSummary(message core.QueuedMessage, maxWidth int) string {
	17	+	if len(message.Images) == 0 {
	18	+	return truncateLine(message.Text, maxWidth)
	19	+	}
	20	+	imageLabel := "[image]"
	21	+	if len(message.Images) > 1 {
	22	+	imageLabel = fmt.Sprintf("[%d images]", len(message.Images))
	23	+	}
	24	+	if message.Text == "" \|\| maxWidth <= len(imageLabel) {
	25	+	return truncateLine(imageLabel, maxWidth)
	26	+	}
	27	+	text := truncateLine(message.Text, maxWidth-len(imageLabel)-1)
	28	+	if text == "" {
	29	+	return imageLabel
	30	+	}
	31	+	return text + " " + imageLabel
	32	+	}
	33	+
15	34		func (i *Interactive) cachedChatLocked(cols int) []string {
16	35		// A busy frame contains mutable streaming/tool state. Keep the stable
17	36		// transcript cache, but never return a full-frame cache entry that would

@@ -749,7 +768,7 @@ func (i *Interactive) redraw() {
749	768		// in flight. Shown directly above the status bar so they're close
750	769		// to the editor but don't push the chat around.
751	770		var queue []string
752		-	queued := append([]string(nil), i.queued...)
	771	+	queued := append([]core.QueuedMessage(nil), i.queued...)
753	772		if i.agent != nil {
754	773		queued = append(queued, i.agent.PendingQueuedMessages()...)
755	774		}

@@ -757,7 +776,7 @@ func (i *Interactive) redraw() {
757	776		queue = append(queue, "")
758	777		for _, q := range queued {
759	778		label := i.cfg.Theme.FGColor(i.cfg.Theme.Accent, " sliding in: ")
760		-	text := truncateLine(q, mainCols-17)
	779	+	text := queuedMessageSummary(q, mainCols-17)
761	780		queue = append(queue, label+i.cfg.Theme.FGColor(i.cfg.Theme.Muted, text))
762	781		}
763	782		// Hint row, rendered in the same muted tone as the model

 	"github.com/bnema/zut/packages/agent/extproto"
 	"github.com/bnema/zut/packages/agent/skills"
+	"github.com/bnema/zut/packages/core"
 	"github.com/bnema/zut/packages/provider"
 	"github.com/bnema/zut/packages/tui"
 )
+func queuedMessageSummary(message core.QueuedMessage, maxWidth int) string {
+	if len(message.Images) == 0 {
+		return truncateLine(message.Text, maxWidth)
+	}
+	imageLabel := "[image]"
+	if len(message.Images) > 1 {
+		imageLabel = fmt.Sprintf("[%d images]", len(message.Images))
+	}
+	if message.Text == "" || maxWidth <= len(imageLabel) {
+		return truncateLine(imageLabel, maxWidth)
+	}
+	text := truncateLine(message.Text, maxWidth-len(imageLabel)-1)
+	if text == "" {
+		return imageLabel
+	}
+	return text + " " + imageLabel
+}
 func (i *Interactive) cachedChatLocked(cols int) []string {
 	// A busy frame contains mutable streaming/tool state. Keep the stable
 	// transcript cache, but never return a full-frame cache entry that would
 	// in flight. Shown directly above the status bar so they're close
 	// to the editor but don't push the chat around.
 	var queue []string
-	queued := append([]string(nil), i.queued...)
+	queued := append([]core.QueuedMessage(nil), i.queued...)
 	if i.agent != nil {
 		queued = append(queued, i.agent.PendingQueuedMessages()...)
 	}
 		queue = append(queue, "")
 		for _, q := range queued {
 			label := i.cfg.Theme.FGColor(i.cfg.Theme.Accent, "  sliding in: ")
-			text := truncateLine(q, mainCols-17)
+			text := queuedMessageSummary(q, mainCols-17)
 			queue = append(queue, label+i.cfg.Theme.FGColor(i.cfg.Theme.Muted, text))
 		}
 		// Hint row, rendered in the same muted tone as the model

M packages/agent/modes/interactive_resident_subagents.go +2 -2


@@ -190,8 +190,8 @@ func (i *Interactive) executeCoordinatorActions(actions []orchestration.Action)
190	190		prompt = update
191	191		}
192	192		}
193		-	if prompt != "" {
194		-	i.submitOrQueue(prompt, nil, false)
	193	+	if prompt != "" \|\| len(action.Images) != 0 {
	194	+	i.submitOrQueue(prompt, action.Images, false)
195	195		} else if action.Reason == orchestration.WakeGoal {
196	196		parent := i.runCtx
197	197		if parent == nil {


@@ -190,8 +190,8 @@ func (i *Interactive) executeCoordinatorActions(actions []orchestration.Action)
190		prompt = update	190		prompt = update
191		}	191		}
192		}	192		}
193	-	if prompt != "" {	193	+	if prompt != "" \|\| len(action.Images) != 0 {
194	-	i.submitOrQueue(prompt, nil, false)	194	+	i.submitOrQueue(prompt, action.Images, false)
195		} else if action.Reason == orchestration.WakeGoal {	195		} else if action.Reason == orchestration.WakeGoal {
196		parent := i.runCtx	196		parent := i.runCtx
197		if parent == nil {	197		if parent == nil {

M packages/agent/modes/interactive_turn.go +2 -2


@@ -331,7 +331,7 @@ func (i *Interactive) startTurnRequest(parent context.Context, prompt string, im
331	331		// deciding whether this continuation may spend another rescue attempt.
332	332		statusRescueActive := i.compactContinuation.reason == compactContinuationStatusRescue
333	333		// Pop the next queued message, if any, and relaunch.
334		-	var next string
	334	+	var next core.QueuedMessage
335	335		var hasNext bool
336	336		if !awaitingPre && len(i.queued) > 0 && ctx.Err() == nil && err == nil {
337	337		next, i.queued = i.queued[0], i.queued[1:]

@@ -409,7 +409,7 @@ func (i *Interactive) startTurnRequest(parent context.Context, prompt string, im
409	409		}
410	410		switch {
411	411		case hasNext:
412		-	i.startTurn(parent, next)
	412	+	i.startTurnWithImages(parent, next.Text, next.Images)
413	413		case hasScheduled:
414	414		// The scheduler holds the session transition read lock until this
415	415		// acknowledgement, so accepting here binds the prompt to the agent


@@ -331,7 +331,7 @@ func (i *Interactive) startTurnRequest(parent context.Context, prompt string, im
331		// deciding whether this continuation may spend another rescue attempt.	331		// deciding whether this continuation may spend another rescue attempt.
332		statusRescueActive := i.compactContinuation.reason == compactContinuationStatusRescue	332		statusRescueActive := i.compactContinuation.reason == compactContinuationStatusRescue
333		// Pop the next queued message, if any, and relaunch.	333		// Pop the next queued message, if any, and relaunch.
334	-	var next string	334	+	var next core.QueuedMessage
335		var hasNext bool	335		var hasNext bool
336		if !awaitingPre && len(i.queued) > 0 && ctx.Err() == nil && err == nil {	336		if !awaitingPre && len(i.queued) > 0 && ctx.Err() == nil && err == nil {
337		next, i.queued = i.queued[0], i.queued[1:]	337		next, i.queued = i.queued[0], i.queued[1:]

@@ -409,7 +409,7 @@ func (i *Interactive) startTurnRequest(parent context.Context, prompt string, im
409		}	409		}
410		switch {	410		switch {
411		case hasNext:	411		case hasNext:
412	-	i.startTurn(parent, next)	412	+	i.startTurnWithImages(parent, next.Text, next.Images)
413		case hasScheduled:	413		case hasScheduled:
414		// The scheduler holds the session transition read lock until this	414		// The scheduler holds the session transition read lock until this
415		// acknowledgement, so accepting here binds the prompt to the agent	415		// acknowledgement, so accepting here binds the prompt to the agent

M packages/core/agent.go +46 -20


@@ -24,8 +24,14 @@ const (
24	24		DefaultStreamMaxRetries = 5
25	25		)
26	26
	27	+	// QueuedMessage is a user prompt waiting for the next safe model-call boundary.
	28	+	type QueuedMessage struct {
	29	+	Text string
	30	+	Images []provider.ImageBlock
	31	+	}
	32	+
27	33		type queuedMessage struct {
28		-	text string
	34	+	message QueuedMessage
29	35		accepted time.Time
30	36		}
31	37

@@ -274,18 +280,18 @@ func (a *Agent) FastModeEnabled() bool {
274	280		return a.FastMode
275	281		}
276	282
277		-	// QueueMessage queues text to be injected as a user message at the
278		-	// next safe boundary of the active agent loop. It is non-blocking in
279		-	// the sense that it never waits for model/tool work; it only takes
280		-	// the transcript mutex briefly. Empty/whitespace-only messages are
281		-	// ignored.
282		-	func (a *Agent) QueueMessage(text string) bool {
	283	+	// QueueMessage queues a text and image prompt to be injected as a user message
	284	+	// at the next safe boundary of the active agent loop. It is non-blocking in the
	285	+	// sense that it never waits for model/tool work; it only takes the transcript
	286	+	// mutex briefly. Prompts without text or images are ignored.
	287	+	func (a *Agent) QueueMessage(text string, images []provider.ImageBlock) bool {
283	288		text = strings.TrimSpace(text)
284		-	if text == "" {
	289	+	if text == "" && len(images) == 0 {
285	290		return false
286	291		}
	292	+	message := cloneQueuedMessage(QueuedMessage{Text: text, Images: images})
287	293		a.mu.Lock()
288		-	a.queued = append(a.queued, queuedMessage{text: text, accepted: time.Now()})
	294	+	a.queued = append(a.queued, queuedMessage{message: message, accepted: time.Now()})
289	295		a.mu.Unlock()
290	296		return true
291	297		}

@@ -293,12 +299,12 @@ func (a *Agent) QueueMessage(text string) bool {
293	299		// PendingQueuedMessages returns a snapshot of user messages waiting
294	300		// to be injected. Used by hosts to render the visible "sliding in"
295	301		// chips without consuming them.
296		-	func (a *Agent) PendingQueuedMessages() []string {
	302	+	func (a *Agent) PendingQueuedMessages() []QueuedMessage {
297	303		a.mu.Lock()
298	304		defer a.mu.Unlock()
299		-	out := make([]string, len(a.queued))
	305	+	out := make([]QueuedMessage, len(a.queued))
300	306		for i, queued := range a.queued {
301		-	out[i] = queued.text
	307	+	out[i] = cloneQueuedMessage(queued.message)
302	308		}
303	309		return out
304	310		}

@@ -313,30 +319,43 @@ func (a *Agent) QueuedMessageCount() int {
313	319
314	320		// PopQueuedMessage removes and returns the most recently queued
315	321		// message. Hosts use this for the slide-back keybinding.
316		-	func (a *Agent) PopQueuedMessage() (string, bool) {
	322	+	func (a *Agent) PopQueuedMessage() (QueuedMessage, bool) {
317	323		a.mu.Lock()
318	324		defer a.mu.Unlock()
319	325		n := len(a.queued)
320	326		if n == 0 {
321		-	return "", false
	327	+	return QueuedMessage{}, false
322	328		}
323		-	text := a.queued[n-1].text
	329	+	message := cloneQueuedMessage(a.queued[n-1].message)
324	330		a.queued = a.queued[:n-1]
325		-	return text, true
	331	+	return message, true
326	332		}
327	333
328	334		// DrainQueuedMessages discards and returns every queued message.
329	335		// Hosts use this on explicit cancel/clear so stale follow-ups do
330	336		// not run after the user aborted the turn.
331		-	func (a *Agent) DrainQueuedMessages() []string {
	337	+	func (a *Agent) DrainQueuedMessages() []QueuedMessage {
332	338		queued := a.drainQueuedMessages()
333		-	out := make([]string, len(queued))
	339	+	out := make([]QueuedMessage, len(queued))
334	340		for i, message := range queued {
335		-	out[i] = message.text
	341	+	out[i] = cloneQueuedMessage(message.message)
336	342		}
337	343		return out
338	344		}
339	345
	346	+	func cloneQueuedMessage(message QueuedMessage) QueuedMessage {
	347	+	clone := QueuedMessage{Text: message.Text}
	348	+	if len(message.Images) == 0 {
	349	+	return clone
	350	+	}
	351	+	clone.Images = make([]provider.ImageBlock, len(message.Images))
	352	+	for i, image := range message.Images {
	353	+	clone.Images[i] = image
	354	+	clone.Images[i].Data = append([]byte(nil), image.Data...)
	355	+	}
	356	+	return clone
	357	+	}
	358	+
340	359		func (a *Agent) drainQueuedMessages() []queuedMessage {
341	360		a.mu.Lock()
342	361		defer a.mu.Unlock()

@@ -351,9 +370,16 @@ func (a *Agent) appendQueuedAsUser(messages []queuedMessage, sink func(AgentEven
351	370		if accepted.IsZero() {
352	371		accepted = time.Now()
353	372		}
	373	+	content := make([]provider.Content, 0, 1+len(queued.message.Images))
	374	+	if queued.message.Text != "" {
	375	+	content = append(content, provider.TextBlock{Text: queued.message.Text})
	376	+	}
	377	+	for _, image := range queued.message.Images {
	378	+	content = append(content, image)
	379	+	}
354	380		msg := provider.Message{
355	381		Role: provider.RoleUser,
356		-	Content: []provider.Content{provider.TextBlock{Text: queued.text}},
	382	+	Content: content,
357	383		Time: accepted,
358	384		}
359	385		a.mu.Lock()

 	DefaultStreamMaxRetries  = 5
 )
+// QueuedMessage is a user prompt waiting for the next safe model-call boundary.
+type QueuedMessage struct {
+	Text   string
+	Images []provider.ImageBlock
+}
 type queuedMessage struct {
-	text     string
+	message  QueuedMessage
 	accepted time.Time
 }
 	return a.FastMode
 }
-// QueueMessage queues text to be injected as a user message at the
+// QueueMessage queues a text and image prompt to be injected as a user message
-// next safe boundary of the active agent loop. It is non-blocking in
+// at the next safe boundary of the active agent loop. It is non-blocking in the
-// the sense that it never waits for model/tool work; it only takes
+// sense that it never waits for model/tool work; it only takes the transcript
-// the transcript mutex briefly. Empty/whitespace-only messages are
+// mutex briefly. Prompts without text or images are ignored.
-// ignored.
+func (a *Agent) QueueMessage(text string, images []provider.ImageBlock) bool {
-func (a *Agent) QueueMessage(text string) bool {
 	text = strings.TrimSpace(text)
-	if text == "" {
+	if text == "" && len(images) == 0 {
 		return false
 	}
+	message := cloneQueuedMessage(QueuedMessage{Text: text, Images: images})
 	a.mu.Lock()
-	a.queued = append(a.queued, queuedMessage{text: text, accepted: time.Now()})
+	a.queued = append(a.queued, queuedMessage{message: message, accepted: time.Now()})
 	a.mu.Unlock()
 	return true
 }
 // PendingQueuedMessages returns a snapshot of user messages waiting
 // to be injected. Used by hosts to render the visible "sliding in"
 // chips without consuming them.
-func (a *Agent) PendingQueuedMessages() []string {
+func (a *Agent) PendingQueuedMessages() []QueuedMessage {
 	a.mu.Lock()
 	defer a.mu.Unlock()
-	out := make([]string, len(a.queued))
+	out := make([]QueuedMessage, len(a.queued))
 	for i, queued := range a.queued {
-		out[i] = queued.text
+		out[i] = cloneQueuedMessage(queued.message)
 	}
 	return out
 }
 // PopQueuedMessage removes and returns the most recently queued
 // message. Hosts use this for the slide-back keybinding.
-func (a *Agent) PopQueuedMessage() (string, bool) {
+func (a *Agent) PopQueuedMessage() (QueuedMessage, bool) {
 	a.mu.Lock()
 	defer a.mu.Unlock()
 	n := len(a.queued)
 	if n == 0 {
-		return "", false
+		return QueuedMessage{}, false
 	}
-	text := a.queued[n-1].text
+	message := cloneQueuedMessage(a.queued[n-1].message)
 	a.queued = a.queued[:n-1]
-	return text, true
+	return message, true
 }
 // DrainQueuedMessages discards and returns every queued message.
 // Hosts use this on explicit cancel/clear so stale follow-ups do
 // not run after the user aborted the turn.
-func (a *Agent) DrainQueuedMessages() []string {
+func (a *Agent) DrainQueuedMessages() []QueuedMessage {
 	queued := a.drainQueuedMessages()
-	out := make([]string, len(queued))
+	out := make([]QueuedMessage, len(queued))
 	for i, message := range queued {
-		out[i] = message.text
+		out[i] = cloneQueuedMessage(message.message)
 	}
 	return out
 }
+func cloneQueuedMessage(message QueuedMessage) QueuedMessage {
+	clone := QueuedMessage{Text: message.Text}
+	if len(message.Images) == 0 {
+		return clone
+	}
+	clone.Images = make([]provider.ImageBlock, len(message.Images))
+	for i, image := range message.Images {
+		clone.Images[i] = image
+		clone.Images[i].Data = append([]byte(nil), image.Data...)
+	}
+	return clone
+}
 func (a *Agent) drainQueuedMessages() []queuedMessage {
 	a.mu.Lock()
 	defer a.mu.Unlock()
 		if accepted.IsZero() {
 			accepted = time.Now()
 		}
+		content := make([]provider.Content, 0, 1+len(queued.message.Images))
+		if queued.message.Text != "" {
+			content = append(content, provider.TextBlock{Text: queued.message.Text})
+		}
+		for _, image := range queued.message.Images {
+			content = append(content, image)
+		}
 		msg := provider.Message{
 			Role:    provider.RoleUser,
-			Content: []provider.Content{provider.TextBlock{Text: queued.text}},
+			Content: content,
 			Time:    accepted,
 		}
 		a.mu.Lock()

M packages/core/queue_test.go +45 -9


@@ -100,7 +100,7 @@ func TestQueuedMessageInjectedAfterToolBatchBeforeNextModelCall(t *testing.T) {
100	100		}()
101	101
102	102		<-tool.started
103		-	if !a.QueueMessage("also do Y") {
	103	+	if !a.QueueMessage("also do Y", nil) {
104	104		t.Fatal("QueueMessage returned false")
105	105		}
106	106		close(tool.release)

@@ -131,10 +131,46 @@ func queueTestContains(xs []string, want string) bool {
131	131		return false
132	132		}
133	133
	134	+	func TestQueuedImageMessageInjectedAtNextSafeBoundary(t *testing.T) {
	135	+	client := &queueFakeClient{}
	136	+	tool := &blockingTool{started: make(chan struct{}), release: make(chan struct{})}
	137	+	a := NewAgent(client, "fake-model", "system", Registry{"echo": tool})
	138	+
	139	+	var queued provider.Message
	140	+	done := make(chan error, 1)
	141	+	go func() {
	142	+	done <- a.Prompt(context.Background(), "inspect later", nil, func(ev AgentEvent) {
	143	+	if user, ok := ev.(EvUserMessage); ok && len(user.Message.Content) > 0 {
	144	+	if _, isImage := user.Message.Content[0].(provider.ImageBlock); isImage {
	145	+	queued = user.Message
	146	+	}
	147	+	}
	148	+	})
	149	+	}()
	150	+
	151	+	<-tool.started
	152	+	image := provider.ImageBlock{MimeType: "image/png", Data: []byte("png-bytes")}
	153	+	if !a.QueueMessage("", []provider.ImageBlock{image}) {
	154	+	t.Fatal("QueueMessage returned false for image-only prompt")
	155	+	}
	156	+	close(tool.release)
	157	+
	158	+	if err := <-done; err != nil {
	159	+	t.Fatalf("Prompt returned %v", err)
	160	+	}
	161	+	if len(queued.Content) != 1 {
	162	+	t.Fatalf("queued content = %#v, want one image", queued.Content)
	163	+	}
	164	+	got, ok := queued.Content[0].(provider.ImageBlock)
	165	+	if !ok \|\| got.MimeType != image.MimeType \|\| string(got.Data) != string(image.Data) {
	166	+	t.Fatalf("queued image = %#v, want %#v", queued.Content[0], image)
	167	+	}
	168	+	}
	169	+
134	170		func TestQueuedMessageKeepsAcceptedTimeAcrossDrain(t *testing.T) {
135	171		a := NewAgent(nil, "fake", "", Registry{})
136	172		before := time.Now()
137		-	if !a.QueueMessage("queued") {
	173	+	if !a.QueueMessage("queued", nil) {
138	174		t.Fatal("QueueMessage returned false")
139	175		}
140	176		after := time.Now()

@@ -150,18 +186,18 @@ func TestQueuedMessageKeepsAcceptedTimeAcrossDrain(t *testing.T) {
150	186
151	187		func TestQueueMessageSnapshotPopAndDrain(t *testing.T) {
152	188		a := NewAgent(nil, "fake", "", Registry{})
153		-	if a.QueueMessage(" ") {
	189	+	if a.QueueMessage(" ", nil) {
154	190		t.Fatal("blank queue message accepted")
155	191		}
156		-	a.QueueMessage("one")
157		-	a.QueueMessage("two")
158		-	if got := a.PendingQueuedMessages(); len(got) != 2 \|\| got[0] != "one" \|\| got[1] != "two" {
	192	+	a.QueueMessage("one", nil)
	193	+	a.QueueMessage("two", nil)
	194	+	if got := a.PendingQueuedMessages(); len(got) != 2 \|\| got[0].Text != "one" \|\| got[1].Text != "two" {
159	195		t.Fatalf("PendingQueuedMessages = %v; want [one two]", got)
160	196		}
161		-	if text, ok := a.PopQueuedMessage(); !ok \|\| text != "two" {
162		-	t.Fatalf("PopQueuedMessage = %q,%v; want two,true", text, ok)
	197	+	if message, ok := a.PopQueuedMessage(); !ok \|\| message.Text != "two" {
	198	+	t.Fatalf("PopQueuedMessage = %#v,%v; want two,true", message, ok)
163	199		}
164		-	if got := a.DrainQueuedMessages(); len(got) != 1 \|\| got[0] != "one" {
	200	+	if got := a.DrainQueuedMessages(); len(got) != 1 \|\| got[0].Text != "one" {
165	201		t.Fatalf("DrainQueuedMessages = %v; want [one]", got)
166	202		}
167	203		if got := a.QueuedMessageCount(); got != 0 {


@@ -100,7 +100,7 @@ func TestQueuedMessageInjectedAfterToolBatchBeforeNextModelCall(t *testing.T) {
100		}()	100		}()
101			101
102		<-tool.started	102		<-tool.started
103	-	if !a.QueueMessage("also do Y") {	103	+	if !a.QueueMessage("also do Y", nil) {
104		t.Fatal("QueueMessage returned false")	104		t.Fatal("QueueMessage returned false")
105		}	105		}
106		close(tool.release)	106		close(tool.release)

@@ -131,10 +131,46 @@ func queueTestContains(xs []string, want string) bool {
131		return false	131		return false
132		}	132		}
133			133
			134	+	func TestQueuedImageMessageInjectedAtNextSafeBoundary(t *testing.T) {
			135	+	client := &queueFakeClient{}
			136	+	tool := &blockingTool{started: make(chan struct{}), release: make(chan struct{})}
			137	+	a := NewAgent(client, "fake-model", "system", Registry{"echo": tool})
			138	+
			139	+	var queued provider.Message
			140	+	done := make(chan error, 1)
			141	+	go func() {
			142	+	done <- a.Prompt(context.Background(), "inspect later", nil, func(ev AgentEvent) {
			143	+	if user, ok := ev.(EvUserMessage); ok && len(user.Message.Content) > 0 {
			144	+	if _, isImage := user.Message.Content[0].(provider.ImageBlock); isImage {
			145	+	queued = user.Message
			146	+	}
			147	+	}
			148	+	})
			149	+	}()
			150	+
			151	+	<-tool.started
			152	+	image := provider.ImageBlock{MimeType: "image/png", Data: []byte("png-bytes")}
			153	+	if !a.QueueMessage("", []provider.ImageBlock{image}) {
			154	+	t.Fatal("QueueMessage returned false for image-only prompt")
			155	+	}
			156	+	close(tool.release)
			157	+
			158	+	if err := <-done; err != nil {
			159	+	t.Fatalf("Prompt returned %v", err)
			160	+	}
			161	+	if len(queued.Content) != 1 {
			162	+	t.Fatalf("queued content = %#v, want one image", queued.Content)
			163	+	}
			164	+	got, ok := queued.Content[0].(provider.ImageBlock)
			165	+	if !ok \|\| got.MimeType != image.MimeType \|\| string(got.Data) != string(image.Data) {
			166	+	t.Fatalf("queued image = %#v, want %#v", queued.Content[0], image)
			167	+	}
			168	+	}
			169	+
134		func TestQueuedMessageKeepsAcceptedTimeAcrossDrain(t *testing.T) {	170		func TestQueuedMessageKeepsAcceptedTimeAcrossDrain(t *testing.T) {
135		a := NewAgent(nil, "fake", "", Registry{})	171		a := NewAgent(nil, "fake", "", Registry{})
136		before := time.Now()	172		before := time.Now()
137	-	if !a.QueueMessage("queued") {	173	+	if !a.QueueMessage("queued", nil) {
138		t.Fatal("QueueMessage returned false")	174		t.Fatal("QueueMessage returned false")
139		}	175		}
140		after := time.Now()	176		after := time.Now()

@@ -150,18 +186,18 @@ func TestQueuedMessageKeepsAcceptedTimeAcrossDrain(t *testing.T) {
150			186
151		func TestQueueMessageSnapshotPopAndDrain(t *testing.T) {	187		func TestQueueMessageSnapshotPopAndDrain(t *testing.T) {
152		a := NewAgent(nil, "fake", "", Registry{})	188		a := NewAgent(nil, "fake", "", Registry{})
153	-	if a.QueueMessage(" ") {	189	+	if a.QueueMessage(" ", nil) {
154		t.Fatal("blank queue message accepted")	190		t.Fatal("blank queue message accepted")
155		}	191		}
156	-	a.QueueMessage("one")	192	+	a.QueueMessage("one", nil)
157	-	a.QueueMessage("two")	193	+	a.QueueMessage("two", nil)
158	-	if got := a.PendingQueuedMessages(); len(got) != 2 \|\| got[0] != "one" \|\| got[1] != "two" {	194	+	if got := a.PendingQueuedMessages(); len(got) != 2 \|\| got[0].Text != "one" \|\| got[1].Text != "two" {
159		t.Fatalf("PendingQueuedMessages = %v; want [one two]", got)	195		t.Fatalf("PendingQueuedMessages = %v; want [one two]", got)
160		}	196		}
161	-	if text, ok := a.PopQueuedMessage(); !ok \|\| text != "two" {	197	+	if message, ok := a.PopQueuedMessage(); !ok \|\| message.Text != "two" {
162	-	t.Fatalf("PopQueuedMessage = %q,%v; want two,true", text, ok)	198	+	t.Fatalf("PopQueuedMessage = %#v,%v; want two,true", message, ok)
163		}	199		}
164	-	if got := a.DrainQueuedMessages(); len(got) != 1 \|\| got[0] != "one" {	200	+	if got := a.DrainQueuedMessages(); len(got) != 1 \|\| got[0].Text != "one" {
165		t.Fatalf("DrainQueuedMessages = %v; want [one]", got)	201		t.Fatalf("DrainQueuedMessages = %v; want [one]", got)
166		}	202		}
167		if got := a.QueuedMessageCount(); got != 0 {	203		if got := a.QueuedMessageCount(); got != 0 {