refactor(agent): remove legacy image mode prompt injection

nanobot/agent/tools/image_generation.py

@@ -8,7 +8,6 @@ from typing import TYPE_CHECKING, Any
 from pydantic import Field
 
 from nanobot.agent.tools.base import Tool, ToolResult, tool_parameters
-from nanobot.agent.tools.context import RequestContext
 from nanobot.agent.tools.schema import (
     ArraySchema,
     IntegerSchema,
@@ -22,7 +21,6 @@ from nanobot.providers.image_generation import (
     ImageGenerationProvider,
     get_image_gen_provider,
 )
-from nanobot.runtime_context import RuntimeContextBlock
 from nanobot.security.workspace_access import current_tool_workspace
 from nanobot.security.workspace_policy import WorkspaceBoundaryError, resolve_allowed_path
 from nanobot.utils.artifacts import (
@@ -31,7 +29,6 @@ from nanobot.utils.artifacts import (
     store_generated_image_artifact,
 )
 from nanobot.utils.helpers import detect_image_mime
-from nanobot.utils.image_generation_intent import image_generation_runtime_context
 
 if TYPE_CHECKING:
     from nanobot.config.schema import ProviderConfig
@@ -119,18 +116,6 @@ class ImageGenerationTool(Tool):
             "or user image paths as reference_images."
         )
 
-    def runtime_context_provider(self):
-        return self._provide_runtime_context
-
-    async def _provide_runtime_context(
-        self,
-        request: RequestContext,
-    ) -> RuntimeContextBlock | None:
-        content = image_generation_runtime_context(request.metadata)
-        if not content:
-            return None
-        return RuntimeContextBlock(source="image_generation", content=content)
-
     def _provider_config(self) -> ProviderConfig | None:
         return self.provider_configs.get(self.config.provider)
 

nanobot/channels/websocket.py

@@ -782,13 +782,6 @@ class WebSocketChannel(BaseChannel):
                 metadata["mcp_presets"] = mcp_presets
             metadata[WORKSPACE_SCOPE_METADATA_KEY] = scope.metadata()
             self._workspaces.persist_scope(cid, scope)
-            image_generation = envelope.get("image_generation")
-            if isinstance(image_generation, dict) and image_generation.get("enabled") is True:
-                aspect_ratio = image_generation.get("aspect_ratio")
-                metadata["image_generation"] = {
-                    "enabled": True,
-                    "aspect_ratio": aspect_ratio if isinstance(aspect_ratio, str) else None,
-                }
             if metadata.get("webui") is True and self.is_allowed(client_id):
                 self._transcripts.append_user_message(
                     cid,

nanobot/utils/image_generation_intent.py

@@ -1,33 +0,0 @@
-"""Helpers for WebUI image-generation intent metadata."""
-
-from __future__ import annotations
-
-from typing import Any
-
-IMAGE_GENERATION_METADATA_KEY = "image_generation"
-
-
-def image_generation_prompt(content: str, metadata: dict[str, Any] | None) -> str:
-    """Decorate a user prompt when WebUI image mode is enabled."""
-    runtime_context = image_generation_runtime_context(metadata)
-    return f"{content}\n\n{runtime_context}" if runtime_context else content
-
-
-def image_generation_runtime_context(metadata: dict[str, Any] | None) -> str:
-    """Return the model-only instruction for WebUI image generation mode."""
-    raw = (metadata or {}).get(IMAGE_GENERATION_METADATA_KEY)
-    if not isinstance(raw, dict) or raw.get("enabled") is not True:
-        return ""
-
-    aspect_ratio = raw.get("aspect_ratio")
-    if isinstance(aspect_ratio, str) and aspect_ratio.strip():
-        instruction = (
-            "The user selected WebUI image generation mode. Use the generate_image tool. "
-            f"When calling generate_image, pass aspect_ratio={aspect_ratio!r}."
-        )
-    else:
-        instruction = (
-            "The user selected WebUI image generation mode. Use the generate_image tool. "
-            "Choose the most suitable aspect_ratio yourself from the prompt and intended use."
-        )
-    return f"[WebUI image generation instruction: {instruction}]"

tests/agent/test_loop_image_generation_media.py

@@ -11,9 +11,8 @@ from nanobot.bus.events import InboundMessage
 from nanobot.bus.queue import MessageBus
 from nanobot.config.loader import set_config_path
 from nanobot.config.schema import ImageGenerationToolConfig, ProviderConfig, ToolsConfig
-from nanobot.providers.base import LLMProvider, LLMResponse, ToolCallRequest
+from nanobot.providers.base import LLMResponse, ToolCallRequest
 from nanobot.providers.image_generation import GeneratedImageResponse
-from nanobot.runtime_context import public_history_message
 
 PNG_DATA_URL = (
     "data:image/png;base64,"
@@ -86,56 +85,3 @@ async def test_outbound_no_longer_carries_generated_media(
     # OutboundMessage no longer carries generated media —
     # the LLM sends images via the message tool instead.
     assert result.media == []
-
-
-@pytest.mark.asyncio
-async def test_image_mode_instruction_is_persisted_as_next_turn_prefix(tmp_path: Path) -> None:
-    provider = MagicMock()
-    provider.get_default_model.return_value = "test-model"
-    provider.generation.max_tokens = 4096
-    provider.chat_with_retry = AsyncMock(side_effect=[
-        LLMResponse(content="first answer"),
-        LLMResponse(content="second answer"),
-    ])
-    loop = AgentLoop(
-        bus=MessageBus(),
-        provider=provider,
-        workspace=tmp_path,
-        model="test-model",
-        tools_config=ToolsConfig(
-            image_generation=ImageGenerationToolConfig(enabled=True),
-        ),
-    )
-    loop.consolidator.maybe_consolidate_by_tokens = AsyncMock(return_value=False)
-
-    await loop._process_message(InboundMessage(
-        channel="websocket",
-        sender_id="user",
-        chat_id="chat-image-prefix",
-        content="draw a fox",
-        metadata={
-            "image_generation": {
-                "enabled": True,
-                "aspect_ratio": "16:9",
-            },
-        },
-    ))
-    await loop._process_message(InboundMessage(
-        channel="websocket",
-        sender_id="user",
-        chat_id="chat-image-prefix",
-        content="thanks",
-    ))
-
-    first_wire = LLMProvider._sanitize_empty_content(
-        provider.chat_with_retry.await_args_list[0].kwargs["messages"]
-    )
-    second_wire = LLMProvider._sanitize_empty_content(
-        provider.chat_with_retry.await_args_list[1].kwargs["messages"]
-    )
-    assert second_wire[: len(first_wire)] == first_wire
-    assert "aspect_ratio='16:9'" in first_wire[1]["content"]
-
-    persisted = loop.sessions.get_or_create("websocket:chat-image-prefix").messages[0]
-    assert persisted["content"] == first_wire[1]["content"]
-    assert public_history_message(persisted)["content"] == "draw a fox"

tests/utils/test_image_generation_intent.py

@@ -1,25 +0,0 @@
-from nanobot.utils.image_generation_intent import image_generation_prompt
-
-
-def test_image_generation_prompt_ignores_plain_messages() -> None:
-    assert image_generation_prompt("hello", {}) == "hello"
-
-
-def test_image_generation_prompt_uses_auto_aspect_instruction() -> None:
-    prompt = image_generation_prompt(
-        "Draw a poster",
-        {"image_generation": {"enabled": True, "aspect_ratio": None}},
-    )
-
-    assert "Draw a poster" in prompt
-    assert "Use the generate_image tool" in prompt
-    assert "Choose the most suitable aspect_ratio yourself" in prompt
-
-
-def test_image_generation_prompt_uses_selected_aspect_ratio() -> None:
-    prompt = image_generation_prompt(
-        "Draw a banner",
-        {"image_generation": {"enabled": True, "aspect_ratio": "16:9"}},
-    )
-
-    assert "aspect_ratio='16:9'" in prompt

webui/src/hooks/useNanobotStream.ts

@@ -13,7 +13,6 @@ import type { StreamError } from "@/lib/nanobot-client";
 import type {
   InboundEvent,
   OutboundCliAppMention,
-  OutboundImageGeneration,
   OutboundMcpPresetMention,
   OutboundMedia,
   GoalStateWsPayload,
@@ -477,7 +476,6 @@ export interface SendImage {
 }
 
 export interface SendOptions {
-  imageGeneration?: OutboundImageGeneration;
   cliApps?: OutboundCliAppMention[];
   mcpPresets?: OutboundMcpPresetMention[];
   workspaceScope?: WorkspaceScopePayload | null;

webui/src/lib/nanobot-client.ts

@@ -3,7 +3,6 @@ import type {
   InboundEvent,
   Outbound,
   OutboundCliAppMention,
-  OutboundImageGeneration,
   OutboundMcpPresetMention,
   OutboundMedia,
   GoalStateWsPayload,
@@ -385,7 +384,6 @@ export class NanobotClient {
     content: string,
     media?: OutboundMedia[],
     options?: {
-      imageGeneration?: OutboundImageGeneration;
       cliApps?: OutboundCliAppMention[];
       mcpPresets?: OutboundMcpPresetMention[];
       workspaceScope?: WorkspaceScopePayload | null;
@@ -398,7 +396,6 @@ export class NanobotClient {
       chat_id: chatId,
       content,
       ...(media && media.length > 0 ? { media } : {}),
-      ...(options?.imageGeneration ? { image_generation: options.imageGeneration } : {}),
       ...(options?.cliApps?.length ? { cli_apps: options.cliApps } : {}),
       ...(options?.mcpPresets?.length ? { mcp_presets: options.mcpPresets } : {}),
       ...(options?.workspaceScope ? { workspace_scope: options.workspaceScope } : {}),

webui/src/lib/types.ts

@@ -932,11 +932,6 @@ export interface OutboundMedia {
   name?: string;
 }
 
-export interface OutboundImageGeneration {
-  enabled: true;
-  aspect_ratio?: string | null;
-}
-
 export interface OutboundCliAppMention {
   name: string;
   display_name?: string;
@@ -998,7 +993,6 @@ export type Outbound =
       chat_id: string;
       content: string;
       media?: OutboundMedia[];
-      image_generation?: OutboundImageGeneration;
       cli_apps?: OutboundCliAppMention[];
       mcp_presets?: OutboundMcpPresetMention[];
       workspace_scope?: WorkspaceScopePayload;

webui/src/tests/nanobot-client.test.ts

@@ -528,33 +528,6 @@ describe("NanobotClient", () => {
     });
   });
 
-  it("includes image generation options in outbound messages", () => {
-    const client = new NanobotClient({
-      url: "ws://test",
-      reconnect: false,
-      socketFactory: (url) => new FakeSocket(url) as unknown as WebSocket,
-    });
-    client.connect();
-    lastSocket().fakeOpen();
-
-    client.sendMessage(
-      "chat-img",
-      "draw a banner",
-      undefined,
-      { imageGeneration: { enabled: true, aspect_ratio: "16:9" } },
-    );
-
-    expect(lastSocket().sent).toContain(
-      JSON.stringify({
-        type: "message",
-        chat_id: "chat-img",
-        content: "draw a banner",
-        image_generation: { enabled: true, aspect_ratio: "16:9" },
-        webui: true,
-      }),
-    );
-  });
-
   it("includes CLI app attachments in outbound messages", () => {
     const client = new NanobotClient({
       url: "ws://test",

webui/src/tests/useNanobotStream.test.tsx

@@ -1640,31 +1640,6 @@ describe("useNanobotStream", () => {
     expect(result.current.messages[0].media).toHaveLength(1);
   });
 
-  it("passes image generation options to the websocket client", () => {
-    const fake = fakeClient();
-    const { result } = renderHook(() => useNanobotStream("chat-img", EMPTY_MESSAGES), {
-      wrapper: wrap(fake.client),
-    });
-
-    act(() => {
-      result.current.send(
-        "draw a square icon",
-        undefined,
-        { imageGeneration: { enabled: true, aspect_ratio: "1:1" } },
-      );
-    });
-
-    expect(fake.client.sendMessage).toHaveBeenCalledWith(
-      "chat-img",
-      "draw a square icon",
-      undefined,
-      expect.objectContaining({
-        imageGeneration: { enabled: true, aspect_ratio: "1:1" },
-        turnId: expect.any(String),
-      }),
-    );
-  });
-
   it("stops the active turn without adding a user slash command bubble", () => {
     const fake = fakeClient();
     const { result } = renderHook(() => useNanobotStream("chat-stop", EMPTY_MESSAGES), {