feat: introduce MCP server, cross-platform runtime, and Showcase UI reorganization
- Showcase UI Reorganization: * Move web UI dashboard from root directory into apps/showcase_ui/ for cleaner project structure * Add 'artemis ui' CLI command (artemis/interfaces/cli/commands/ui.py) and launcher scripts for the interactive showcase console - Standalone MCP Server & Notifiers (mcp_server/): * Implement dedicated MCP server package with mobile automation tools (device_state, inspect_trace, task_manager, task_runner) * Add multi-channel notifiers (AgentAPI, Desktop notifications, Webhooks, File output, and Composite routing) * Introduce background task runner and trace store for autonomous mobile execution monitoring - Cross-Platform Runtime & Toolchain Support: * Add artemis/platform/ with OS-specific implementations for Linux, macOS (Darwin), and Windows * Add artemis/runtime/supervisor.py for resilient process and agent runtime supervision * Add artemis/toolchain/ with dependency resolution and SDK descriptors * Add cross-platform launcher scripts (start.sh, start.bat) and installer scripts (scripts/install_deps.*) - Core Enhancements & Diagnostics: * Enhance coordinate normalization, OCR API fallback, and perception pipeline * Add credential validation (credentials_validator.py) and system diagnostics * Update bilingual documentation (README.md, README_CN.md) and architecture diagrams - Test Suite Expansion: * Add integration tests for cloud gateway, remote ADB, and Artemis login * Add unit tests covering core diagnostics, platform detection, toolchain, credentials validator, optional OCR, and MCP tools
This commit is contained in:
somew1nd committed
1 parent
78e478f502
commit
9a5aaf99b7
249 files changed
+37226
-9339
No files matched your search
@@ -0,0 +1,465 @@
|
||||
# Copyright 2026 Google LLC
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
"""MCP Tool: mobile_inspect_trace."""
|
||||
|
||||
import json
|
||||
import math
|
||||
import os
|
||||
import sqlite3
|
||||
import sys
|
||||
from typing import Any
|
||||
from PIL import Image, ImageDraw
|
||||
|
||||
from mcp_server.base import mcp
|
||||
from mcp_server.utils import env_utils, trace_store
|
||||
from artemis.utils.task_tree import (
|
||||
_render_step_detailed,
|
||||
build_plan_and_history,
|
||||
)
|
||||
|
||||
|
||||
def _draw_action_overlay(bg_image_path: str, action_obj: dict, output_path: str) -> bool:
|
||||
"""Draws the action (tap, swipe, etc.) on the image and saves it."""
|
||||
try:
|
||||
with Image.open(bg_image_path) as img:
|
||||
draw_img = img.convert("RGBA")
|
||||
draw = ImageDraw.Draw(draw_img)
|
||||
|
||||
act_type = action_obj.get("action") or action_obj.get("name")
|
||||
coords = action_obj.get("coordinates") or action_obj.get("target")
|
||||
|
||||
if not act_type or not coords:
|
||||
return False
|
||||
|
||||
fill_color = (255, 0, 0, 80)
|
||||
outline_color = (255, 0, 0, 255)
|
||||
outline_width = 4
|
||||
|
||||
if act_type in (
|
||||
"tap",
|
||||
"click",
|
||||
"long_press",
|
||||
"focus_and_input_text",
|
||||
"focus_and_clear_text",
|
||||
):
|
||||
if isinstance(coords, list) and len(coords) == 2:
|
||||
x, y = coords[0], coords[1]
|
||||
r = 25
|
||||
draw.ellipse(
|
||||
[x - r, y - r, x + r, y + r],
|
||||
fill=fill_color,
|
||||
outline=outline_color,
|
||||
width=outline_width,
|
||||
)
|
||||
r_center = 4
|
||||
draw.ellipse(
|
||||
[
|
||||
x - r_center,
|
||||
y - r_center,
|
||||
x + r_center,
|
||||
y + r_center,
|
||||
],
|
||||
fill=outline_color,
|
||||
outline=outline_color,
|
||||
)
|
||||
elif act_type == "swipe":
|
||||
if isinstance(coords, list) and len(coords) == 4:
|
||||
x1, y1, x2, y2 = coords[0], coords[1], coords[2], coords[3]
|
||||
draw.line([x1, y1, x2, y2], fill=outline_color, width=6)
|
||||
r_start = 8
|
||||
draw.ellipse(
|
||||
[
|
||||
x1 - r_start,
|
||||
y1 - r_start,
|
||||
x1 + r_start,
|
||||
y1 + r_start,
|
||||
],
|
||||
fill=outline_color,
|
||||
)
|
||||
dx = x2 - x1
|
||||
dy = y2 - y1
|
||||
angle = math.atan2(dy, dx)
|
||||
arrow_length = 30
|
||||
arrow_angle = math.pi / 6
|
||||
|
||||
p1_x = x2 - arrow_length * math.cos(angle - arrow_angle)
|
||||
p1_y = y2 - arrow_length * math.sin(angle - arrow_angle)
|
||||
p2_x = x2 - arrow_length * math.cos(angle + arrow_angle)
|
||||
p2_y = y2 - arrow_length * math.sin(angle + arrow_angle)
|
||||
|
||||
draw.polygon([x2, y2, p1_x, p1_y, p2_x, p2_y], fill=outline_color)
|
||||
|
||||
final_img = draw_img.convert("RGB")
|
||||
final_img.save(output_path, "JPEG", quality=85)
|
||||
return True
|
||||
except Exception as e:
|
||||
print(f"Error drawing action overlay: {e}", file=sys.stderr)
|
||||
return False
|
||||
|
||||
|
||||
@mcp.tool()
|
||||
async def mobile_inspect_trace(action: str, trace_id: str, step_number: int | None = None) -> Any:
|
||||
"""A specialized trace data retrieval tool designed to extract execution details of mobile automation tasks.
|
||||
|
||||
Use this tool DURING a running task or AFTER it has finished (or failed) to
|
||||
monitor progress, verify answers, diagnose agent errors, and debug.
|
||||
It provides comprehensive runtime information, raw agent outputs, and visual
|
||||
screenshots for both **Flash** and **Pro** tasks.
|
||||
|
||||
### Available Actions:
|
||||
- **'view_summary'**: Retrieves a high-level execution summary across all
|
||||
steps.
|
||||
* **Pro Model**: Displays a hierarchical task plan summary, showing the
|
||||
status and descriptions of all completed, active, and pending steps.
|
||||
* **Flash Model**: Displays an all-in-one execution chain containing the
|
||||
complete reasoning (thoughts/motivation) and action taken (`click`,
|
||||
`input_text`, etc.) for every single step.
|
||||
|
||||
- **'view_step_screenshots'**: Retrieves local file paths of screenshots for
|
||||
a specific step (supports both Flash and Pro). Returns:
|
||||
1. `before_screenshot`: The original device screenshot observed by the
|
||||
agent right before taking the action.
|
||||
2. `after_screenshot`: A post-action screenshot capturing the failure or
|
||||
mismatch state (if the action failed or was intercepted). Note: This is
|
||||
conditional/optional and will be `None` (`null`) for normal/successful
|
||||
steps where no extra post-action screenshot was captured.
|
||||
3. `action_overlay_screenshot`: The `before_screenshot` with the agent's
|
||||
action visually marked (e.g., red circle for taps, red arrow for
|
||||
swipes). CRITICAL for verifying if the agent tapped the correct UI
|
||||
element.
|
||||
|
||||
- **'view_step_details'**: Retrieves deep runtime logs, raw VLM thoughts,
|
||||
and exact actions for a specific step.
|
||||
* **Pro Model**: Fully supported. Returns detailed per-step diagnostic
|
||||
data.
|
||||
* **Flash Model**: DO NOT USE. (Flash reasoning and actions are already
|
||||
fully included in 'view_summary', making this action redundant).
|
||||
|
||||
Args:
|
||||
action: STR. **REQUIRED**. The retrieval action to perform
|
||||
(`"view_summary"`, `"view_step_screenshots"`, or
|
||||
`"view_step_details"`).
|
||||
trace_id: STR. **REQUIRED**. The unique session identifier of the task
|
||||
(returned by `mobile_run_task`).
|
||||
step_number: INT. **CONDITIONAL**. The 1-indexed step number to query. -
|
||||
You MUST provide this if `action="view_step_screenshots"` or
|
||||
`action="view_step_details"`. - You MUST omit or leave this empty if
|
||||
`action="view_summary"`.
|
||||
"""
|
||||
project_root = env_utils.get_project_root()
|
||||
db_path = os.path.join(trace_store.TRACES_DIR, "data_engine.db")
|
||||
if not os.path.exists(db_path):
|
||||
db_path = os.path.join(project_root, "traces", "data_engine.db")
|
||||
|
||||
if not os.path.exists(db_path):
|
||||
return {
|
||||
"error": "Database not found",
|
||||
"message": f"Data engine database does not exist at {db_path} yet.",
|
||||
}
|
||||
|
||||
# 1. Handle Action: "view_summary"
|
||||
if action == "view_summary":
|
||||
plan_content = None
|
||||
plan_path = os.path.join(trace_store.get_trace_dir(trace_id), "notes", "task_plan.md")
|
||||
if not os.path.exists(plan_path):
|
||||
plan_path = os.path.join(trace_store.TRACES_DIR, "notes", "task_plan.md")
|
||||
if os.path.exists(plan_path):
|
||||
try:
|
||||
with open(plan_path, encoding="utf-8") as f:
|
||||
plan_content = f.read()
|
||||
except Exception:
|
||||
pass
|
||||
|
||||
try:
|
||||
conn = sqlite3.connect(db_path)
|
||||
conn.row_factory = sqlite3.Row
|
||||
cursor = conn.cursor()
|
||||
|
||||
cursor.execute(
|
||||
"SELECT start_time FROM sessions WHERE session_id = ?",
|
||||
(trace_id,),
|
||||
)
|
||||
session_row = cursor.fetchone()
|
||||
session_start_time = session_row["start_time"] if session_row else None
|
||||
|
||||
cursor.execute(
|
||||
"SELECT * FROM steps WHERE session_id = ? ORDER BY step_number ASC",
|
||||
(trace_id,),
|
||||
)
|
||||
rows = cursor.fetchall()
|
||||
|
||||
steps = []
|
||||
for r in rows:
|
||||
step_dict = dict(r)
|
||||
for col in ("action_taken", "last_execution_result", "extra_metadata"):
|
||||
if step_dict.get(col):
|
||||
try:
|
||||
step_dict[col] = json.loads(step_dict[col])
|
||||
except Exception:
|
||||
pass
|
||||
|
||||
cursor.execute(
|
||||
"SELECT type, name, status, timestamp, duration, payload FROM traces WHERE step_id = ? ORDER BY timestamp ASC",
|
||||
(step_dict["step_id"],),
|
||||
)
|
||||
trace_rows = cursor.fetchall()
|
||||
events = []
|
||||
for tr in trace_rows:
|
||||
t_dict = dict(tr)
|
||||
if t_dict.get("payload"):
|
||||
try:
|
||||
t_dict["payload"] = json.loads(t_dict["payload"])
|
||||
except Exception:
|
||||
pass
|
||||
events.append(t_dict)
|
||||
step_dict["interleaved_events"] = events
|
||||
|
||||
if session_start_time and step_dict.get("timestamp"):
|
||||
relative_time_val = step_dict["timestamp"] - session_start_time
|
||||
step_dict["relative_time"] = f"{relative_time_val:.1f}s"
|
||||
else:
|
||||
step_dict["relative_time"] = "N/A"
|
||||
|
||||
steps.append(step_dict)
|
||||
conn.close()
|
||||
except Exception as e:
|
||||
return {
|
||||
"error": "Database query error",
|
||||
"message": f"Failed to retrieve steps from database: {e}",
|
||||
}
|
||||
|
||||
status_data = trace_store.read_status(trace_id)
|
||||
is_flash = status_data and status_data.get("model", "").lower() == "flash"
|
||||
|
||||
if is_flash:
|
||||
summary_lines = [
|
||||
"# ARTEMIS Flash Execution Summary",
|
||||
f"**Session ID:** `{trace_id}` | **Model:** Flash\n",
|
||||
"---",
|
||||
"## Step-by-Step Execution Chain\n",
|
||||
]
|
||||
if not steps:
|
||||
summary_lines.append("_No action steps have been recorded yet._")
|
||||
else:
|
||||
for s in steps:
|
||||
s_num = s.get("step_number", "?")
|
||||
rel_t = s.get("relative_time", "N/A")
|
||||
|
||||
dur_val = s.get("duration")
|
||||
if dur_val is None and s.get("interleaved_events"):
|
||||
events = s["interleaved_events"]
|
||||
timestamps = [e.get("timestamp", 0) for e in events if e.get("timestamp")]
|
||||
if timestamps:
|
||||
dur_val = max(timestamps) - min(timestamps)
|
||||
dur_str = f"{dur_val:.1f}s" if dur_val is not None and dur_val > 0 else "< 0.5s"
|
||||
|
||||
thought = s.get("operator_raw_thinking")
|
||||
if not thought or not thought.strip():
|
||||
for ev in s.get("interleaved_events", []):
|
||||
if ev.get("type") in ("raw_thinking", "llm_call") and ev.get("payload"):
|
||||
p = ev["payload"]
|
||||
if isinstance(p, dict) and p.get("response_text"):
|
||||
thought = p["response_text"]
|
||||
break
|
||||
if not thought or not thought.strip():
|
||||
thought = "(Direct action without preliminary text reasoning)"
|
||||
|
||||
act = s.get("action_taken") or "N/A"
|
||||
if isinstance(act, dict) and "action" in act:
|
||||
act_str = f"{act['action']}({act.get('args', act.get('coordinates', ''))})"
|
||||
else:
|
||||
act_str = str(act)
|
||||
res = s.get("last_execution_result") or "N/A"
|
||||
summary_lines.append(f"### Step {s_num} (at {rel_t}, duration {dur_str})")
|
||||
summary_lines.append(f"- **Thinking / Motivation:** {thought}")
|
||||
summary_lines.append(f"- **Action Taken:** `{act_str}` -> `{res}`\n")
|
||||
return "\n".join(summary_lines)
|
||||
|
||||
try:
|
||||
summary_markdown = build_plan_and_history(
|
||||
task_plan=plan_content or "No task plan created yet.",
|
||||
steps=steps,
|
||||
current_subgoal_hash="default",
|
||||
min_summaries=len(steps),
|
||||
last_n_detailed=0,
|
||||
all_detailed=False,
|
||||
)
|
||||
return summary_markdown
|
||||
except Exception as e:
|
||||
return {
|
||||
"error": "Formatting error",
|
||||
"message": f"Failed to format plan and history using Artemis task tree: {e}",
|
||||
}
|
||||
|
||||
# 2. Handle Actions requiring step_number
|
||||
elif action in ("view_step_screenshots", "view_step_details"):
|
||||
status_data = trace_store.read_status(trace_id)
|
||||
is_flash = status_data and status_data.get("model", "").lower() == "flash"
|
||||
|
||||
if action == "view_step_details" and is_flash:
|
||||
return {
|
||||
"error": "Not applicable for Flash model",
|
||||
"message": (
|
||||
"Action 'view_step_details' is not needed for Flash model runs. "
|
||||
"For Flash tasks, 'view_summary' already outputs the complete "
|
||||
"step-by-step reasoning (thoughts) and executed actions for the entire run."
|
||||
),
|
||||
}
|
||||
|
||||
if step_number is None:
|
||||
return {
|
||||
"error": "Missing parameter",
|
||||
"message": f"Parameter 'step_number' is required for action '{action}'.",
|
||||
}
|
||||
|
||||
try:
|
||||
conn = sqlite3.connect(db_path)
|
||||
conn.row_factory = sqlite3.Row
|
||||
cursor = conn.cursor()
|
||||
cursor.execute(
|
||||
"SELECT * FROM steps WHERE session_id = ? AND step_number = ?",
|
||||
(trace_id, step_number),
|
||||
)
|
||||
row = cursor.fetchone()
|
||||
|
||||
if not row:
|
||||
conn.close()
|
||||
return {
|
||||
"error": "Step not found",
|
||||
"message": f"Step number {step_number} not found for trace '{trace_id}'.",
|
||||
}
|
||||
|
||||
step_dict = dict(row)
|
||||
for col in ("action_taken", "last_execution_result", "extra_metadata"):
|
||||
if step_dict.get(col):
|
||||
try:
|
||||
step_dict[col] = json.loads(step_dict[col])
|
||||
except Exception:
|
||||
pass
|
||||
|
||||
cursor.execute(
|
||||
"SELECT start_time FROM sessions WHERE session_id = ?",
|
||||
(trace_id,),
|
||||
)
|
||||
session_row = cursor.fetchone()
|
||||
session_start_time = session_row["start_time"] if session_row else None
|
||||
|
||||
cursor.execute(
|
||||
"SELECT type, name, status, timestamp, duration, payload FROM traces WHERE step_id = ? ORDER BY timestamp ASC",
|
||||
(step_dict["step_id"],),
|
||||
)
|
||||
trace_rows = cursor.fetchall()
|
||||
events = []
|
||||
for tr in trace_rows:
|
||||
t_dict = dict(tr)
|
||||
if t_dict.get("payload"):
|
||||
try:
|
||||
t_dict["payload"] = json.loads(t_dict["payload"])
|
||||
except Exception:
|
||||
pass
|
||||
events.append(t_dict)
|
||||
step_dict["interleaved_events"] = events
|
||||
conn.close()
|
||||
except Exception as e:
|
||||
return {
|
||||
"error": "Database error",
|
||||
"message": f"Database query failed: {e}",
|
||||
}
|
||||
|
||||
# 3. Handle Action: "view_step_screenshots"
|
||||
if action == "view_step_screenshots":
|
||||
images_dir = os.path.join(trace_store.TRACES_DIR, "images")
|
||||
if not os.path.exists(images_dir):
|
||||
images_dir = os.path.join(project_root, "traces", "images")
|
||||
|
||||
pre_image = None
|
||||
post_image = None
|
||||
overlay_image = None
|
||||
|
||||
if step_dict.get("pre_image_name"):
|
||||
name = step_dict["pre_image_name"]
|
||||
if not name.endswith(".jpg"):
|
||||
name += ".jpg"
|
||||
path = os.path.join(images_dir, name)
|
||||
if os.path.exists(path):
|
||||
pre_image = f"file://{path}"
|
||||
|
||||
overlay_dir = trace_store.get_trace_dir(trace_id)
|
||||
os.makedirs(overlay_dir, exist_ok=True)
|
||||
overlay_filename = f"step_{step_dict['step_number']}_overlay.jpg"
|
||||
overlay_abs_path = os.path.join(overlay_dir, overlay_filename)
|
||||
|
||||
action_obj = step_dict.get("action_taken")
|
||||
action_parsed = (
|
||||
action_obj[0] if isinstance(action_obj, list) and action_obj else action_obj
|
||||
)
|
||||
|
||||
if isinstance(action_parsed, dict) and action_parsed.get("action"):
|
||||
if not os.path.exists(overlay_abs_path):
|
||||
success_draw = _draw_action_overlay(
|
||||
path, action_parsed, overlay_abs_path
|
||||
)
|
||||
if success_draw:
|
||||
overlay_image = f"file://{overlay_abs_path}"
|
||||
else:
|
||||
overlay_image = f"file://{overlay_abs_path}"
|
||||
|
||||
if step_dict.get("post_image_name"):
|
||||
name = step_dict["post_image_name"]
|
||||
if not name.endswith(".jpg"):
|
||||
name += ".jpg"
|
||||
path = os.path.join(images_dir, name)
|
||||
if os.path.exists(path):
|
||||
post_image = f"file://{path}"
|
||||
|
||||
return {
|
||||
"trace_id": trace_id,
|
||||
"step_number": step_dict["step_number"],
|
||||
"before_screenshot": pre_image,
|
||||
"after_screenshot": post_image,
|
||||
"action_overlay_screenshot": overlay_image,
|
||||
}
|
||||
|
||||
# 4. Handle Action: "view_step_details"
|
||||
elif action == "view_step_details":
|
||||
relative_time_str = "N/A"
|
||||
if session_start_time and step_dict.get("timestamp"):
|
||||
relative_time_val = step_dict["timestamp"] - session_start_time
|
||||
relative_time_str = f"{relative_time_val:.1f}s"
|
||||
|
||||
rendered_text = _render_step_detailed(
|
||||
step=step_dict,
|
||||
relative_time=relative_time_str,
|
||||
summary=step_dict.get("summary"),
|
||||
action=step_dict.get("action_taken"),
|
||||
result=step_dict.get("last_execution_result"),
|
||||
is_most_recent=False,
|
||||
for_failure_analyzer=False,
|
||||
)
|
||||
|
||||
return {
|
||||
"trace_id": trace_id,
|
||||
"step_number": step_dict["step_number"],
|
||||
"details": rendered_text,
|
||||
}
|
||||
|
||||
else:
|
||||
return {
|
||||
"error": "Invalid action",
|
||||
"message": (
|
||||
f"Action '{action}' is not supported. Supported actions: "
|
||||
"'view_summary', 'view_step_screenshots', 'view_step_details'."
|
||||
),
|
||||
}
|
||||
Reference in new issue
Block a user