Feat/visual grounding delegation

artemis/agents/explorer/constants.py

@@ -33,6 +33,12 @@
     " coordinate, so you can act on it right away."
 )
 
+ASK_EXPLORER_ULTRA_DESCRIPTION = (
+    ASK_EXPLORER_DESCRIPTION
+    + " Explorer can also inspect pixel data to identify color codes and"
+    " disambiguate visually similar targets like subtle color swatches."
+)
+
 ASK_EXPLORER_QUERY_DESCRIPTION = (
     "What to find, described the way you see it on the screenshot: visible"
     " text, icon shape, color, position, or nearby landmarks (e.g. 'gear icon"

artemis/agents/explorer/explorer.json

@@ -29,7 +29,7 @@
     "**Finalization**: Submit coordinates using `submit_answer`. In your final turn all other tools are stripped away, leaving only `submit_answer`, so never postpone the submission past the point where you are reasonably confident.",
     "**Tool Classes**: `Perception` tools extract and locate elements, coordinates and text on the screen; the `Finalization` tool (`submit_answer`) submits your ranked candidates and terminates the search.",
     {
-      "text": "**Transformation Tool**: `ask_image_processor` performs pixel-level image editing (cropping, upscaling, masking, contrast changes) and registers the resulting images in the Image Pool. It is slow and expensive; use it only when perception on the original screenshot is insufficient, and call it sequentially because later perception steps depend on its output.",
+      "text": "**Transformation Tool**: `ask_image_processor` performs pixel-level image editing and analysis (cropping, upscaling, masking, color/contrast adjustments, or pixel sampling) and registers the resulting images in the Image Pool. It is slow and expensive; use it only when perception on the original screenshot is insufficient, and call it sequentially because later perception steps depend on its output.",
       "requires": ["ask_image_processor"]
     },
     {

artemis/agents/explorer/perception_tools.py

@@ -630,3 +630,4 @@ async def exec_inspect_region(
                 "text": f"Error: Region inspection failed: {e}",
                 "image_path": None,
             }
+

artemis/agents/explorer/tool_declarations.py

@@ -95,7 +95,10 @@
         "type": "function",
         "function": {
             "name": "ask_image_processor",
-            "description": ("[Transformation] A scripting tool for pixel-level image processing."),
+            "description": (
+                "[Transformation] A scripting tool for pixel-level image processing,"
+                " transformations, and spot pixel/color inspection."
+            ),
             "parameters": {
                 "type": "object",
                 "properties": {
@@ -274,10 +277,10 @@
     types.FunctionDeclaration(
         name="ask_image_processor",
         description=(
-            "[Transformation] A scripting tool for pixel-level image"
-            " processing. CAUTION: This tool is slow and expensive. Returns"
-            " a summary of the operations performed and new image IDs,"
-            " along with new labels and screen coordinates."
+            "[Transformation] A scripting tool for pixel-level image processing,"
+            " transformations, and spot pixel/color inspection. CAUTION: This tool is"
+            " slow and expensive. Returns a summary of the operations performed and"
+            " new image IDs, along with new labels and screen coordinates."
         ),
         parameters=types.Schema(
             type=types.Type.OBJECT,

artemis/agents/operator/prompts.py

@@ -39,6 +39,7 @@
 
 
 from artemis.agents.prompt_assembly import render_tool_enum, resolve_available
+from artemis.config import resolve_explorer_version
 from artemis.mcp.action_specs import OPERATOR_SHELL_ORDER
 
 
@@ -260,12 +261,14 @@ def _checkpoint_max_repairs(ctx: ArtemisContext) -> int:
 def _grammar_render_context(ctx: ArtemisContext) -> dict:
     """Template variables shared by the static and legacy system renders."""
     midway, final, verification_active = _operator_grammar_flags(ctx)
+    explorer_tier = resolve_explorer_version(ctx, agent_or_profile_name="operator")
     return {
         "plan_grammar": render_plan_grammar_spec(midway=midway, final=final),
         "verification_active": verification_active,
         "checks_active": midway or final,
         "midway_checks_active": midway,
         "checkpoint_max_repairs": _checkpoint_max_repairs(ctx),
+        "explorer_tier": explorer_tier,
     }
 
 

artemis/tools/explorer_tool.py

@@ -44,6 +44,7 @@
     ASK_EXPLORER_DESCRIPTION,
     ASK_EXPLORER_QUERY_DESCRIPTION,
     ASK_EXPLORER_TOOL_NAME,
+    ASK_EXPLORER_ULTRA_DESCRIPTION,
 )
 from artemis.agents.explorer.explorer import Explorer
 from artemis.agents.explorer.geometry import (
@@ -516,9 +517,15 @@ def __init__(
     ):
         self.version = version
         self.agent_name = agent_name
+        if description is None:
+            description = (
+                ASK_EXPLORER_ULTRA_DESCRIPTION
+                if version == "ultra"
+                else ASK_EXPLORER_DESCRIPTION
+            )
         super().__init__(
             name=ASK_EXPLORER_TOOL_NAME,
-            description=description or ASK_EXPLORER_DESCRIPTION,
+            description=description,
             args_schema=AskExplorerArgs,
             category=category,
         )
@@ -558,7 +565,10 @@ def get_ask_explorer_tool(
     agent_name: str = "operator",
 ) -> BaseTool:
     """Exports ``ask_explorer`` as a LangChain tool bound to ``ctx``."""
-    return AskExplorerTool(version=version, agent_name=agent_name).to_langchain_tool(ctx)
+    resolved_version = resolve_explorer_version(
+        ctx, explicit_version=version, agent_or_profile_name=agent_name
+    )
+    return AskExplorerTool(version=resolved_version, agent_name=agent_name).to_langchain_tool(ctx)
 
 
 ask_explorer_wrapper = ToolWrapper(

artemis/tools/image_processor_tool.py

@@ -63,7 +63,7 @@ def __init__(self):
             name="ask_image_processor",
             description=(
                 "Call this tool to ask the image processor agent to run"
-                " complex image transformations or modifications."
+                " complex image transformations, modifications, or pixel/color measurements."
             ),
             args_schema=AskVisionCoderArgs,
             category="custom",

tests/unit/tools/test_explorer_tool.py

@@ -24,6 +24,7 @@
     ASK_EXPLORER_CONTEXT_FEEDBACK_DESCRIPTION,
     ASK_EXPLORER_DESCRIPTION,
     ASK_EXPLORER_QUERY_DESCRIPTION,
+    ASK_EXPLORER_ULTRA_DESCRIPTION,
 )
 from artemis.context import ArtemisContext
 from artemis.graph.state import State
@@ -118,7 +119,12 @@ def test_tool_contract_is_tier_agnostic():
 @pytest.mark.parametrize("version", [None, "flash", "pro", "ultra"])
 def test_description_does_not_depend_on_the_tier(version):
     tool = AskExplorerTool(version=version)
-    assert tool.description == ASK_EXPLORER_DESCRIPTION
+    expected = (
+        ASK_EXPLORER_ULTRA_DESCRIPTION
+        if version == "ultra"
+        else ASK_EXPLORER_DESCRIPTION
+    )
+    assert tool.description == expected
     assert tool.version == version