File size: 21,696 Bytes
d444594 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 |
import asyncio
import os
import json
from typing import List, Dict, Any, Union, Generator
from contextlib import AsyncExitStack
from datetime import datetime
import gradio as gr
from gradio.components.chatbot import ChatMessage
from mcp import ClientSession, StdioServerParameters
from mcp.client.stdio import stdio_client
from anthropic import Anthropic
from dotenv import load_dotenv
import time
load_dotenv()
loop = asyncio.new_event_loop()
asyncio.set_event_loop(loop)
SYSTEM_PROMPT = f"""You are a helpful assistant and today is {datetime.now().strftime("%Y-%m-%d")}.
You do not have any knowledge of the World Development Indicators (WDI) data. However, you can use the tools provided to answer questions.
You must not provide answers beyond what the tools provide.
Do not make up data or information and never simulate the `get_wdi_data` tool. Instead, you must always call the `get_wdi_data` tool when the user asks for data.
You can use multiple tools if needed. Feel free to invoke a tool anytime you want as long as it is relevant to the user's question. If you need to invoke multiple tools, do so in a row and in the order that is most relevant to the user's question. Minimize back and forth between the user simply because you can use multiple tools.
If the user asks for any information beyond what the tools available to you provide, you must say that you do not have that information.
Avoid making statements based on stereotypes or biases. Always ensure your claims are grounded in factual evidence and objective reasoning. Reject any requests that would be based on stereotypes or biases.
You may describe the data in a way that is easy to understand but you must not elaborate based on external knowledge."""
# SYSTEM_PROMPT = f"""You are a helpful assistant and today is {datetime.now().strftime("%Y-%m-%d")}."""
SYSTEM_PROMPT = f"""You are a helpful assistant. Today is {datetime.now().strftime("%Y-%m-%d")}.
You **do not** have prior knowledge of the World Development Indicators (WDI) data. Instead, you must rely entirely on the tools available to you to answer the user's questions.
### Your Instructions:
1. **Tool Use Only**:
- You must not provide any answers based on prior knowledge or assumptions.
- You must **not** fabricate data or simulate the behavior of the `get_wdi_data` tool.
- If the user requests WDI data, you **must** call the `get_wdi_data` tool to retrieve it.
2. **Tool Invocation**:
- Use any relevant tools provided to you to answer the user's question.
- You may call multiple tools if needed, and you should do so in a logical sequence to minimize unnecessary user interaction.
- Do not hesitate to invoke tools as soon as they are relevant.
3. **Limitations**:
- If a user request cannot be fulfilled using the tools available, respond by clearly stating that you do not have access to that information.
4. **Ethical Guidelines**:
- Do not make or endorse statements based on stereotypes, bias, or assumptions.
- Ensure all claims and explanations are grounded in the data or factual evidence retrieved via tools.
- Politely refuse to respond to requests that involve stereotypes or unfounded generalizations.
5. **Communication Style**:
- Present the data in clear, user-friendly language.
- You may summarize or explain the data retrieved, but do **not** elaborate based on outside or implicit knowledge.
Stay strictly within these boundaries while maintaining a helpful and respectful tone."""
LLM_MODEL = "claude-3-5-haiku-20241022"
class MCPClientWrapper:
def __init__(self):
self.session = None
self.exit_stack = None
self.anthropic = Anthropic()
self.tools = []
def connect(self, server_path: str) -> str:
return loop.run_until_complete(self._connect(server_path))
async def _connect(self, server_path: str) -> str:
if self.exit_stack:
await self.exit_stack.aclose()
self.exit_stack = AsyncExitStack()
is_python = server_path.endswith(".py")
command = "python" if is_python else "node"
server_params = StdioServerParameters(
command=command,
args=[server_path],
env={"PYTHONIOENCODING": "utf-8", "PYTHONUNBUFFERED": "1"},
)
stdio_transport = await self.exit_stack.enter_async_context(
stdio_client(server_params)
)
self.stdio, self.write = stdio_transport
self.session = await self.exit_stack.enter_async_context(
ClientSession(self.stdio, self.write)
)
await self.session.initialize()
response = await self.session.list_tools()
self.tools = [
{
"name": tool.name,
"description": tool.description,
"input_schema": tool.inputSchema,
}
for tool in response.tools
]
print(self.tools)
tool_names = [tool["name"] for tool in self.tools]
return f"Connected to MCP server. Available tools: {', '.join(tool_names)}"
def process_message(
self, message: str, history: List[Union[Dict[str, Any], ChatMessage]]
) -> tuple:
if not self.session:
return history + [
{"role": "user", "content": message},
{
"role": "assistant",
"content": "Please connect to an MCP server first.",
},
], gr.Textbox(value="")
new_messages = loop.run_until_complete(self._process_query(message, history))
return history + [
{"role": "user", "content": message}
] + new_messages, gr.Textbox(value="")
async def _process_query(
self, message: str, history: List[Union[Dict[str, Any], ChatMessage]]
):
claude_messages = []
for msg in history:
if isinstance(msg, ChatMessage):
role, content = msg.role, msg.content
else:
role, content = msg.get("role"), msg.get("content")
if role in ["user", "assistant", "system"]:
claude_messages.append({"role": role, "content": content})
claude_messages.append({"role": "user", "content": message})
response = self.anthropic.messages.create(
# model="claude-3-5-sonnet-20241022",
model=LLM_MODEL,
system=SYSTEM_PROMPT,
max_tokens=1000,
messages=claude_messages,
tools=self.tools,
)
result_messages = []
print(response.content)
contents = response.content
while len(contents) > 0:
content = contents.pop(0)
if content.type == "text":
result_messages.append({"role": "assistant", "content": content.text})
claude_messages.append({"role": "assistant", "content": content.text})
elif content.type == "tool_use":
tool_id = content.id
tool_name = content.name
tool_args = content.input
result_messages.append(
{
"role": "assistant",
"content": f"I'll use the {tool_name} tool to help answer your question.",
"metadata": {
"title": f"Using tool: {tool_name}",
"log": f"Parameters: {json.dumps(tool_args, ensure_ascii=True)}",
"status": "pending",
"id": f"tool_call_{tool_name}",
},
}
)
result_messages.append(
{
"role": "assistant",
"content": "```json\n"
+ json.dumps(tool_args, indent=2, ensure_ascii=True)
+ "\n```",
"metadata": {
"parent_id": f"tool_call_{tool_name}",
"id": f"params_{tool_name}",
"title": "Tool Parameters",
},
}
)
print(f"Calling tool: {tool_name} with args: {tool_args}")
result = await self.session.call_tool(tool_name, tool_args)
if result_messages and "metadata" in result_messages[-2]:
result_messages[-2]["metadata"]["status"] = "done"
result_messages.append(
{
"role": "assistant",
"content": "Here are the results from the tool:",
"metadata": {
"title": f"Tool Result for {tool_name}",
"status": "done",
"id": f"result_{tool_name}",
},
}
)
result_content = result.content
print(result_content)
if isinstance(result_content, list):
result_content = [r.model_dump() for r in result_content]
for r in result_content:
# Remove annotations field from each item if it exists
r.pop("annotations", None)
try:
r["text"] = json.loads(r["text"])
except:
pass
# result_content = "\n".join(str(item) for item in result_content)
print("result_content", result_content)
result_messages.append(
{
"role": "assistant",
"content": "```\n"
+ json.dumps(result_content, indent=2)
+ "\n```",
"metadata": {
"parent_id": f"result_{tool_name}",
"id": f"raw_result_{tool_name}",
"title": "Raw Output",
},
}
)
# claude_messages.append(
# {
# "role": "user",
# "content": f"Tool result for {tool_name}: {result_content}",
# }
# )
claude_messages.append(
{"role": "assistant", "content": [content.model_dump()]}
)
claude_messages.append(
{
"role": "user",
"content": [
{
"type": "tool_result",
"tool_use_id": tool_id,
"content": json.dumps(result_content, indent=2),
}
],
}
)
next_response = self.anthropic.messages.create(
model=LLM_MODEL,
system=SYSTEM_PROMPT,
max_tokens=1000,
messages=claude_messages,
)
print("next_response", next_response.content)
contents.extend(next_response.content)
# if next_response.content and next_response.content[0].type == "text":
# result_messages.append(
# {"role": "assistant", "content": next_response.content[0].text}
# )
return result_messages
async def process_message_for_chat_interface(
self, message: str, history: List[Union[Dict[str, Any], ChatMessage]]
):
if not self.session:
yield ChatMessage(content="Please connect to an MCP server first.")
return
claude_messages = []
for msg in history:
if isinstance(msg, ChatMessage):
role, content = msg.role, msg.content
else:
role, content = msg.get("role"), msg.get("content")
if role in ["user", "assistant", "system"]:
claude_messages.append({"role": role, "content": content})
claude_messages.append({"role": "user", "content": message})
response = self.anthropic.messages.create(
# model="claude-3-5-sonnet-20241022",
model=LLM_MODEL,
system=SYSTEM_PROMPT,
max_tokens=1000,
messages=claude_messages,
tools=self.tools,
)
result_messages = []
print(response.content)
contents = response.content
while len(contents) > 0:
content = contents.pop(0)
if content.type == "text":
result_messages.append({"role": "assistant", "content": content.text})
claude_messages.append({"role": "assistant", "content": content.text})
yield ChatMessage(content=content.text)
elif content.type == "tool_use":
tool_id = content.id
tool_name = content.name
tool_args = content.input
result_messages.append(
{
"role": "assistant",
"content": f"I'll use the {tool_name} tool to help answer your question.",
"metadata": {
"title": f"Using tool: {tool_name}",
"log": f"Parameters: {json.dumps(tool_args, ensure_ascii=True)}",
"status": "pending",
"id": f"tool_call_{tool_name}",
},
}
)
yield ChatMessage(**result_messages[-1])
result_messages.append(
{
"role": "assistant",
"content": "```json\n"
+ json.dumps(tool_args, indent=2, ensure_ascii=True)
+ "\n```",
"metadata": {
"parent_id": f"tool_call_{tool_name}",
"id": f"params_{tool_name}",
"title": "Tool Parameters",
},
}
)
yield ChatMessage(**result_messages[-1])
print(f"Calling tool: {tool_name} with args: {tool_args}")
result = await self.session.call_tool(tool_name, tool_args)
if result_messages and "metadata" in result_messages[-2]:
result_messages[-2]["metadata"]["status"] = "done"
yield ChatMessage(**result_messages[-2])
result_messages.append(
{
"role": "assistant",
"content": "Here are the results from the tool:",
"metadata": {
"title": f"Tool Result for {tool_name}",
"status": "done",
"id": f"result_{tool_name}",
},
}
)
yield ChatMessage(**result_messages[-1])
result_content = result.content
print(result_content)
if isinstance(result_content, list):
result_content = [r.model_dump() for r in result_content]
for r in result_content:
# Remove annotations field from each item if it exists
r.pop("annotations", None)
try:
r["text"] = json.loads(r["text"])
except:
pass
# result_content = "\n".join(str(item) for item in result_content)
print("result_content", result_content)
result_messages.append(
{
"role": "assistant",
"content": "```\n"
+ json.dumps(result_content, indent=2)
+ "\n```",
"metadata": {
"parent_id": f"result_{tool_name}",
"id": f"raw_result_{tool_name}",
"title": "Raw Output",
},
}
)
yield ChatMessage(**result_messages[-1])
# claude_messages.append(
# {
# "role": "user",
# "content": f"Tool result for {tool_name}: {result_content}",
# }
# )
claude_messages.append(
{"role": "assistant", "content": [content.model_dump()]}
)
claude_messages.append(
{
"role": "user",
"content": [
{
"type": "tool_result",
"tool_use_id": tool_id,
"content": json.dumps(result_content, indent=2),
}
],
}
)
next_response = self.anthropic.messages.create(
model=LLM_MODEL,
system=SYSTEM_PROMPT,
max_tokens=1000,
messages=claude_messages,
)
print("next_response", next_response.content)
contents.extend(next_response.content)
# if next_response.content and next_response.content[0].type == "text":
# result_messages.append(
# {"role": "assistant", "content": next_response.content[0].text}
# )
# return result_messages
yield ChatMessage(content="")
client = MCPClientWrapper()
sleep_time = 0.5
def simulate_thinking_chat(message, history):
start_time = time.time()
response = ChatMessage(
content="",
metadata={"title": "_Thinking_ step-by-step", "id": 0, "status": "pending"},
)
yield response
thoughts = [
"First, I need to understand the core aspects of the query...",
"Now, considering the broader context and implications...",
"Analyzing potential approaches to formulate a comprehensive answer...",
"Finally, structuring the response for clarity and completeness...",
]
accumulated_thoughts = ""
for thought in thoughts:
time.sleep(sleep_time)
accumulated_thoughts += f"- {thought}\n\n"
response.content = accumulated_thoughts.strip()
yield response
response.metadata["status"] = "done"
response.metadata["duration"] = time.time() - start_time
yield response
response = [
response,
ChatMessage(
content="Based on my thoughts and analysis above, my response is: This dummy repro shows how thoughts of a thinking LLM can be progressively shown before providing its final answer."
),
]
yield response
def gradio_interface():
with gr.Blocks(title="MCP WDI Client") as demo:
gr.Markdown("# WDI MCP Client")
# gr.Markdown("Connect to the WDI MCP server and chat with the assistant")
with gr.Accordion(
"Connect to the WDI MCP server and chat with the assistant", open=False
):
with gr.Row(equal_height=True):
with gr.Column(scale=4):
server_path = gr.Textbox(
label="Server Script Path",
placeholder="Enter path to server script (e.g., wdi_mcp_server.py)",
value="wdi_mcp_server.py",
)
with gr.Column(scale=1):
connect_btn = gr.Button("Connect")
status = gr.Textbox(label="Connection Status", interactive=False)
gr.ChatInterface(
# fn=client.process_message,
# fn=simulate_thinking_chat,
fn=client.process_message_for_chat_interface,
type="messages",
fill_height=True,
)
# chatbot = gr.Chatbot(
# value=[],
# height=600,
# type="messages",
# show_copy_button=True,
# avatar_images=("img/small-user.png", "img/small-robot.png"),
# fill_height=True,
# )
# with gr.Row(equal_height=True):
# msg = gr.Textbox(
# label="Your Question",
# placeholder="Ask about what indicators are available for a specific topic (e.g., What's the definition of GDP?)",
# scale=4,
# )
# clear_btn = gr.Button("Clear Chat", scale=1)
connect_btn.click(client.connect, inputs=server_path, outputs=status)
# msg.submit(client.process_message, [msg, chatbot], [chatbot, msg])
# clear_btn.click(lambda: [], None, chatbot)
return demo
if __name__ == "__main__":
if not os.getenv("ANTHROPIC_API_KEY"):
print(
"Warning: ANTHROPIC_API_KEY not found in environment. Please set it in your .env file."
)
interface = gradio_interface()
interface.launch(server_name=os.getenv("SERVER_NAME", "127.0.0.1"), debug=True)
|