universal-mcp-agents 0.1.17__py3-none-any.whl → 0.1.19__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Potentially problematic release.


This version of universal-mcp-agents might be problematic. Click here for more details.

@@ -9,7 +9,9 @@ from universal_mcp.agents.react import ReactAgent
9
9
  from universal_mcp.agents.simple import SimpleAgent
10
10
 
11
11
 
12
- def get_agent(agent_name: Literal["react", "simple", "builder", "bigtool", "codeact-script", "codeact-repl"]):
12
+ def get_agent(
13
+ agent_name: Literal["react", "simple", "builder", "bigtool", "codeact-script", "codeact-repl"],
14
+ ):
13
15
  if agent_name == "react":
14
16
  return ReactAgent
15
17
  elif agent_name == "simple":
@@ -49,8 +49,11 @@ class BaseAgent:
49
49
  run_metadata.update(metadata)
50
50
 
51
51
  run_config = {
52
+ "recursion_limit": 25,
52
53
  "configurable": {"thread_id": thread_id},
53
54
  "metadata": run_metadata,
55
+ "run_id": thread_id,
56
+ "run_name": self.name,
54
57
  }
55
58
 
56
59
  async for event, meta in self._graph.astream(
@@ -28,9 +28,6 @@ def run(name: str = "react"):
28
28
  "model": "anthropic/claude-sonnet-4-20250514",
29
29
  "registry": AgentrRegistry(client=client),
30
30
  "memory": MemorySaver(),
31
- "tools": {
32
- "google_mail": ["send_email"],
33
- },
34
31
  }
35
32
  agent_cls = get_agent(name)
36
33
  agent = agent_cls(name=name, **params)
@@ -19,12 +19,6 @@ async def main():
19
19
  memory=memory,
20
20
  )
21
21
  print("Starting agent...")
22
- # await agent.ainit()
23
- # await agent.run_interactive()
24
- # async for event in agent.stream(
25
- # user_input="Fetch unsubscribe links from my Gmail inbox for promo emails I have received in the last 7 days"
26
- # ):
27
- # print(event.content, end="")
28
22
  result = await agent.invoke(
29
23
  user_input="Fetch unsubscribe links from my Gmail inbox for promo emails I have received in the last 7 days"
30
24
  )
@@ -27,7 +27,7 @@ def smart_print(data: Any) -> None:
27
27
  Args:
28
28
  data: Either a dictionary with string keys, or a list of such dictionaries
29
29
  """
30
- print(light_copy(data)) # noqa
30
+ print(light_copy(data)) # noqa: T201
31
31
 
32
32
 
33
33
  def creative_writer(
@@ -275,105 +275,3 @@ def data_extractor(
275
275
  .invoke(prompt)
276
276
  )
277
277
  return cast(dict[str, Any], response)
278
-
279
-
280
- # news_articles_schema = {
281
- # "type": "object",
282
- # "properties": {
283
- # "articles": {
284
- # "type": "array",
285
- # "title": "Articles",
286
- # "description": "List of news articles",
287
- # "items": {
288
- # "type": "object",
289
- # "properties": {
290
- # "headline": {
291
- # "type": "string",
292
- # "title": "Headline",
293
- # "description": "The headline of the news article"
294
- # },
295
- # "url": {
296
- # "type": "string",
297
- # "title": "URL",
298
- # "description": "The URL of the news article"
299
- # }
300
- # },
301
- # "required": ["headline", "url"],
302
- # }
303
- # }
304
- # },
305
- # "required": ["articles"],
306
- # }
307
-
308
-
309
- # news_articles_schema = {
310
- # "title": "NewsArticleList",
311
- # "description": "A list of news articles with headlines and URLs",
312
- # "type": "object",
313
- # "properties": {
314
- # "articles": {
315
- # "type": "array",
316
- # "items": {
317
- # "type": "object",
318
- # "properties": {
319
- # "headline": {
320
- # "type": "string"
321
- # },
322
- # "url": {
323
- # "type": "string"
324
- # }
325
- # },
326
- # "required": ["headline", "url"]
327
- # }
328
- # }
329
- # },
330
- # "required": ["articles"]
331
- # }
332
- # model = init_chat_model(model="claude-4-sonnet-20250514", temperature=0)
333
- # structured_model = model.with_structured_output(news_articles_schema)
334
-
335
-
336
- # class TwitterComment(BaseModel):
337
- # skip: bool
338
- # reason: str
339
- # comment: str
340
-
341
- # twitter_comment_schema = {
342
- # "title": "TwitterComment",
343
- # "description": "A twitter comment to engage with followers",
344
- # "type": "object",
345
- # "properties": {
346
- # "skip": {
347
- # "type": "boolean"
348
- # },
349
- # "reason": {
350
- # "type": "string"
351
- # },
352
- # "comment": {
353
- # "type": "string"
354
- # },
355
- # "tagged_profiles": {
356
- # "type": "array",
357
- # "items": {
358
- # "type": "string"
359
- # }
360
- # }
361
- # },
362
- # "required": ["skip", "reason"]
363
- # }
364
-
365
- # comment = {
366
- # "tweet_id": "08109402",
367
- # "handle": "@iamnishant",
368
- # "text": "Hey really loved this tweet! Well said 💯"
369
- # }
370
-
371
- # comment_instructions = (
372
- # "Goal is to engage with my twitter followers who have commented on my tweets."
373
- # "Please generate a single line, context-aware, conversational reply for the given comment."
374
- # "- Use social media language (can use hinglish)."
375
- # "- Skip the reply, if the comment is too generic."
376
- # "- Also tag relevant people in the reply."
377
- # )
378
-
379
- # my_reply = call_llm(comment_instructions, comment, twitter_comment_schema)
@@ -2,18 +2,16 @@ import inspect
2
2
  import json
3
3
  import re
4
4
  from collections.abc import Callable
5
- from dataclasses import dataclass
6
- from pathlib import Path
7
5
  from typing import Literal, cast
8
6
 
9
- from langchain_core.messages import AIMessage, RemoveMessage, ToolMessage
7
+ from langchain_core.messages import AIMessage, ToolMessage
10
8
  from langchain_core.tools import StructuredTool
11
9
  from langchain_core.tools import tool as create_tool
12
10
  from langgraph.checkpoint.base import BaseCheckpointSaver
13
11
  from langgraph.graph import START, StateGraph
14
12
  from langgraph.types import Command, RetryPolicy
15
13
  from universal_mcp.tools.registry import ToolRegistry
16
- from universal_mcp.types import ToolFormat, ToolConfig
14
+ from universal_mcp.types import ToolConfig, ToolFormat
17
15
 
18
16
  from universal_mcp.agents.base import BaseAgent
19
17
  from universal_mcp.agents.codeact0.llm_tool import ai_classify, call_llm, data_extractor, smart_print
@@ -22,10 +20,14 @@ from universal_mcp.agents.codeact0.prompts import (
22
20
  )
23
21
  from universal_mcp.agents.codeact0.sandbox import eval_unsafe, execute_ipython_cell
24
22
  from universal_mcp.agents.codeact0.state import CodeActState
25
- from universal_mcp.agents.codeact0.tools import create_meta_tools, enter_playbook_mode, exit_playbook_mode, get_valid_tools
23
+ from universal_mcp.agents.codeact0.tools import (
24
+ create_meta_tools,
25
+ enter_playbook_mode,
26
+ get_valid_tools,
27
+ )
26
28
  from universal_mcp.agents.codeact0.utils import inject_context, smart_truncate
27
29
  from universal_mcp.agents.llm import load_chat_model
28
- from universal_mcp.agents.utils import filter_retry_on, get_message_text, convert_tool_ids_to_dict
30
+ from universal_mcp.agents.utils import convert_tool_ids_to_dict, filter_retry_on, get_message_text
29
31
 
30
32
  PLAYBOOK_PLANNING_PROMPT = """Now, you are tasked with creating a reusable playbook from the user's previous workflow.
31
33
 
@@ -48,7 +50,6 @@ Example:
48
50
  Now create a plan based on the conversation history. Enclose it between ``` and ```. Ask the user if the plan is okay."""
49
51
 
50
52
 
51
-
52
53
  PLAYBOOK_CONFIRMING_PROMPT = """Now, you are tasked with confirming the playbook plan. Return True if the user is happy with the plan, False otherwise. Do not say anything else in your response. The user response will be the last message in the chain.
53
54
  """
54
55
 
@@ -80,7 +81,7 @@ class CodeActPlaybookAgent(BaseAgent):
80
81
  memory=memory,
81
82
  **kwargs,
82
83
  )
83
- self.model_instance = load_chat_model(model, thinking=True)
84
+ self.model_instance = load_chat_model(model)
84
85
  self.tools_config = tools or []
85
86
  self.registry = registry
86
87
  self.playbook_registry = playbook_registry
@@ -92,20 +93,19 @@ class CodeActPlaybookAgent(BaseAgent):
92
93
  meta_tools = create_meta_tools(self.registry)
93
94
  additional_tools = [smart_print, data_extractor, ai_classify, call_llm, meta_tools["web_search"]]
94
95
  self.additional_tools = [t if isinstance(t, StructuredTool) else create_tool(t) for t in additional_tools]
96
+
95
97
  async def call_model(state: CodeActState) -> Command[Literal["sandbox", "execute_tools"]]:
96
98
  self.exported_tools = []
97
99
  if self.tools_config:
98
100
  # Convert dict format to list format if needed
99
101
  if isinstance(self.tools_config, dict):
100
102
  self.tools_config = [
101
- f"{provider}__{tool}"
102
- for provider, tools in self.tools_config.items()
103
- for tool in tools
103
+ f"{provider}__{tool}" for provider, tools in self.tools_config.items() for tool in tools
104
104
  ]
105
105
  if not self.registry:
106
106
  raise ValueError("Tools are configured but no registry is provided")
107
107
  # Langchain tools are fine
108
- self.tools_config.extend(state.get('selected_tool_ids',[]))
108
+ self.tools_config.extend(state.get("selected_tool_ids", []))
109
109
  self.exported_tools = await self.registry.export_tools(self.tools_config, ToolFormat.LANGCHAIN)
110
110
  self.final_instructions, self.tools_context = create_default_prompt(
111
111
  self.exported_tools, self.additional_tools, self.instructions
@@ -167,7 +167,7 @@ class CodeActPlaybookAgent(BaseAgent):
167
167
  )
168
168
  return Command(
169
169
  goto="playbook",
170
- update={"playbook_mode": "planning", "messages": [tool_message]}, #Entered Playbook mode
170
+ update={"playbook_mode": "planning", "messages": [tool_message]}, # Entered Playbook mode
171
171
  )
172
172
  elif tool_call["name"] == "execute_ipython_cell":
173
173
  return Command(goto="sandbox")
@@ -184,8 +184,13 @@ class CodeActPlaybookAgent(BaseAgent):
184
184
  ai_msg = f"Please login to the following app(s) using the following links and let me know in order to proceed:\n {links} "
185
185
  elif tool_call["name"] == "search_functions":
186
186
  tool_result = await meta_tools["search_functions"].ainvoke(tool_call["args"])
187
+ else:
188
+ raise Exception(
189
+ f"Unexpected tool call: {tool_call['name']}. "
190
+ "tool calls must be one of 'enter_playbook_mode', 'execute_ipython_cell', 'load_functions', or 'search_functions'"
191
+ )
187
192
  except Exception as e:
188
- tool_result = f"Error during {tool_call}: {e}"
193
+ tool_result = str(e)
189
194
 
190
195
  tool_message = ToolMessage(
191
196
  content=json.dumps(tool_result),
@@ -261,13 +266,12 @@ class CodeActPlaybookAgent(BaseAgent):
261
266
  response = cast(AIMessage, response)
262
267
  response_text = get_message_text(response)
263
268
  # Extract plan from response text between triple backticks
264
- plan_match = re.search(r'```(.*?)```', response_text, re.DOTALL)
269
+ plan_match = re.search(r"```(.*?)```", response_text, re.DOTALL)
265
270
  if plan_match:
266
271
  plan = plan_match.group(1).strip()
267
272
  else:
268
273
  plan = response_text.strip()
269
274
  return Command(update={"messages": [response], "playbook_mode": "confirming", "plan": plan})
270
-
271
275
 
272
276
  elif playbook_mode == "confirming":
273
277
  confirmation_instructions = self.instructions + PLAYBOOK_CONFIRMING_PROMPT
@@ -279,8 +283,6 @@ class CodeActPlaybookAgent(BaseAgent):
279
283
  else:
280
284
  return Command(goto="playbook", update={"playbook_mode": "planning"})
281
285
 
282
-
283
-
284
286
  elif playbook_mode == "generating":
285
287
  generating_instructions = self.instructions + PLAYBOOK_GENERATING_PROMPT
286
288
  messages = [{"role": "system", "content": generating_instructions}] + state["messages"]
@@ -327,25 +329,19 @@ class CodeActPlaybookAgent(BaseAgent):
327
329
  saved_note = f"Failed to save generated playbook as Agent '{function_name}': {e}"
328
330
 
329
331
  # Mock tool call for exit_playbook_mode (for testing/demonstration)
330
- mock_exit_tool_call = {
331
- "name": "exit_playbook_mode",
332
- "args": {},
333
- "id": "mock_exit_playbook_123"
334
- }
335
- mock_assistant_message = AIMessage(
336
- content=saved_note,
337
- tool_calls=[mock_exit_tool_call]
338
- )
332
+ mock_exit_tool_call = {"name": "exit_playbook_mode", "args": {}, "id": "mock_exit_playbook_123"}
333
+ mock_assistant_message = AIMessage(content=saved_note, tool_calls=[mock_exit_tool_call])
339
334
 
340
-
341
335
  # Mock tool response for exit_playbook_mode
342
336
  mock_exit_tool_response = ToolMessage(
343
337
  content=json.dumps(f"Exited Playbook Mode.{saved_note}"),
344
338
  name="exit_playbook_mode",
345
- tool_call_id="mock_exit_playbook_123"
339
+ tool_call_id="mock_exit_playbook_123",
346
340
  )
347
341
 
348
- return Command(update={"messages": [mock_assistant_message, mock_exit_tool_response], "playbook_mode": "normal"})
342
+ return Command(
343
+ update={"messages": [mock_assistant_message, mock_exit_tool_response], "playbook_mode": "normal"}
344
+ )
349
345
 
350
346
  def route_entry(state: CodeActState) -> Literal["call_model", "playbook"]:
351
347
  """Route to either normal mode or playbook creation"""
@@ -9,102 +9,41 @@ from universal_mcp.agents.codeact0.utils import schema_to_signature
9
9
  uneditable_prompt = """
10
10
  You are **Wingmen**, an AI Assistant created by AgentR — a creative, straight-forward, and direct principal software engineer with access to tools.
11
11
 
12
- ## Responsibilities
13
-
14
- - **Answer directly** if the task is simple (e.g. print, math, general knowledge).
15
- - For any task requiring logic, execution, or data handling, use `execute_ipython_cell`.
16
- - For writing or NLP tasks (summarizing, generating, extracting), always use AI functions via code never respond directly.
17
-
18
- ## Tool vs. Function: Required Separation
19
-
20
- You must clearly distinguish between tools (called via the tool calling API) and internal functions (used inside code blocks).
21
-
22
- ### Tools Must Be Called via Tool Calling API
23
-
24
- These must be called using **tool calling**, not from inside code blocks:
25
-
26
- - `execute_ipython_cell` For running any Python code or logic.
27
- - `search_functions` To discover available functions for a task.
28
- - `load_functions` To load a specific function by full ID.
29
-
30
- **Do not attempt to call these inside `python` code.**
31
- Use tool calling syntax for these operations.
32
-
33
- ### Functions Must Be Used Inside Code Blocks
34
-
35
- All other functions, including LLM functions, must always be used within code executed by `execute_ipython_cell`. These include:
36
-
37
- - `smart_print()` — For inspecting unknown data structures before looping.
38
- - `asyncio.run()` — For wrapping and executing asynchronous logic. You must not use await outside an async function. And the async function must be called by `asyncio.run()`.
39
- - Any functions for applications loaded via `load_functions`.
40
- - Any logic, data handling, writing, NLP, generation, summarization, or extraction functionality of LLMs.
41
-
42
- These must be called **inside a Python code block**, and that block must be executed using `execute_ipython_cell`.
43
-
44
- ## Tool/Function Usage Policy
45
-
46
- 1. **Always Use Tools/Functions for Required Tasks**
47
- Any searching, loading, or executing must be done using a tool/function call. Never answer manually if a tool/function is appropriate.
48
-
49
- 2. **Use Existing Functions First**
50
- Use existing functions if available. Otherwise, use `search_functions` with a concise query describing the task.
51
-
52
- 3. **Load Only Relevant Tools**
53
- When calling `load_functions`, include only relevant function IDs.
54
- - Prefer connected applications over unconnected ones.
55
- - If multiple functions match (i.e. if none are connected, or multiple are connected), ask the user to choose.
56
- - After loading a tool, you do not need to import/declare it again. It can be called directly in further cells.
57
-
58
- 4. **Follow First Turn Process Strictly**
59
- On the **first turn**, do only **one** of the following:
60
- - Handle directly (if trivial)
61
- - Use a tool/function (`execute_ipython_cell`, `search_functions`, etc.)
62
-
63
- **Do not extend the conversation on the first message.**
64
-
65
- ## Coding Rules
66
-
67
- - Use `smart_print()` to inspect unknown structures, especially those received from function outputs, before looping or branching.
68
- - Validate logic with a single item before processing lists or large inputs.
69
- - Try to achieve as much as possible in a single code block.
70
- - Use only pre-installed Python libraries. Do import them once before using.
71
- - Outer level functions, variables, classes, and imports declared previously can be used in later cells.
72
- - For all functions, call using keyword arguments only. DO NOT use any positional arguments.
73
-
74
- ### **Async Function Usage — Critical**
75
-
76
- When calling asynchronous functions:
77
- - You must define or use an **inner async function**.
78
- - Use `await` only **inside** that async function.
79
- - Run it using `asyncio.run(<function_name>())` **without** `await` at the outer level.
80
-
81
- **Wrong - Using `await` outside an async function**
82
- ```
83
- result = await some_async_function()
84
- ```
85
- **Wrong - Attaching await before asyncio.run**.
86
- `await asyncio.run(main())`
87
- These will raise SyntaxError: 'await' outside async function
88
- The correct method is the following-
89
- ```
90
- import asyncio
91
- async def some_async_function():
92
- ...
93
-
94
- async def main():
95
- result = await some_async_function()
96
- print(result)
97
-
98
- asyncio.run(main())
99
- #or
100
- result = asyncio.run(some_async_function(arg1 = <arg1>))
101
- ```
102
- ## Output Formatting
103
- - All code results must be returned in **Markdown**.
104
- - The user cannot see raw output, so format results clearly:
105
- - Use tables for structured data.
106
- - Provide links for files or images.
107
- - Be explicit in formatting to ensure readability.
12
+ Your job is to answer the user's question or perform the task they ask for.
13
+ - Answer simple questions (which do not require you to write any code or access any external resources) directly. Note that any operation that involves using ONLY print functions should be answered directly.
14
+ - For task requiring operations or access to external resources, you should achieve the task by executing Python code snippets.
15
+ - You have access to `execute_ipython_cell` tool that allows you to execute Python code in an IPython notebook cell.
16
+ - You also have access to two tools for finding and loading more python functions- `search_functions` and `load_functions`, which you must use for finding functions for using different external applications.
17
+ - Prefer pre-loaded or functions already available when possible.
18
+ - Prioritize connected applications over unconnected ones from the output of `search_functions`.
19
+ - When multiple apps are connected, or none of the apps are connected, ask the user to choose the application(s).
20
+ - In writing or natural language processing tasks DO NOT answer directly. Instead use `execute_ipython_cell` tool with the AI functions provided to you for tasks like summarizing, text generation, classification, data extraction from text or unstructured data, etc. Avoid hardcoded approaches to classification, data extraction.
21
+ - The code you write will be executed in a sandbox environment, and you can use the output of previous executions in your code. variables, functions, imports are retained.
22
+ - Read and understand the output of the previous code snippet and use it to answer the user's request. Note that the code output is NOT visible to the user, so after the task is complete, you have to give the output to the user in a markdown format.
23
+ - If needed, feel free to ask for more information from the user (without using the `execute_ipython_cell` tool) to clarify the task.
24
+
25
+ GUIDELINES for writing code:
26
+ - Variables defined at the top level of previous code snippets can be referenced in your code.
27
+ - External functions which return a dict or list[dict] are ambiguous. Therefore, you MUST explore the structure of the returned data using `smart_print()` statements before using it, printing keys and values. `smart_print` truncates long strings from data, preventing huge output logs.
28
+ - When an operation involves running a fixed set of steps on a list of items, run one run correctly and then use a for loop to run the steps on each item in the list.
29
+ - In a single code snippet, try to achieve as much as possible.
30
+ - You can only import libraries that come pre-installed with Python. For external functions, use the search and load tools to access them in the code.
31
+ - For displaying final results to the user, you must present your output in markdown format, including image links, so that they are rendered and displayed to the user. The code output is NOT visible to the user.
32
+ - Call all functions using keyword arguments only, never positional arguments.
33
+ - Async Functions (Critical): Use them only as follows-
34
+ Case 1: Top-level await without asyncio.run()
35
+ Wrap in async function and call with asyncio.run():
36
+ async def main():
37
+ result = await some_async_function()
38
+ return result
39
+ asyncio.run(main())
40
+ Case 2: Using asyncio.run() directly
41
+ If code already contains asyncio.run(), use as-is — do not wrap again:
42
+ asyncio.run(some_async_function())
43
+ Rules:
44
+ - Never use await outside an async function
45
+ - Never use await asyncio.run()
46
+ - Never nest asyncio.run() calls
108
47
  """
109
48
 
110
49
 
@@ -6,18 +6,19 @@ from langchain_core.tools import tool
6
6
  from universal_mcp.tools.registry import ToolRegistry
7
7
  from universal_mcp.types import ToolFormat
8
8
 
9
- MAX_LENGHT=100
9
+ MAX_LENGHT = 100
10
+
10
11
 
11
12
  def enter_playbook_mode():
12
13
  """Call this function to enter playbook mode. Playbook mode is when the user wants to store a repeated task as a script with some inputs for the future."""
13
14
  return
14
15
 
16
+
15
17
  def exit_playbook_mode():
16
18
  """Call this function to exit playbook mode. Playbook mode is when the user wants to store a repeated task as a script with some inputs for the future."""
17
19
  return
18
20
 
19
21
 
20
-
21
22
  def create_meta_tools(tool_registry: ToolRegistry) -> dict[str, Any]:
22
23
  """Create the meta tools for searching and loading tools"""
23
24
 
@@ -46,7 +47,7 @@ def create_meta_tools(tool_registry: ToolRegistry) -> dict[str, Any]:
46
47
  for tool in tools_list:
47
48
  app = tool["id"].split("__")[0]
48
49
  tool_id = tool["id"]
49
-
50
+
50
51
  # Check if within limit and add to set (automatically deduplicates)
51
52
  if len(app_tools[app]) < MAX_LENGTH:
52
53
  cleaned_desc = tool["description"].split("Context:")[0].strip()
@@ -54,6 +55,9 @@ def create_meta_tools(tool_registry: ToolRegistry) -> dict[str, Any]:
54
55
 
55
56
  # Build result string efficiently
56
57
  result_parts = []
58
+ apps_in_results = set(app_tools.keys())
59
+ connected_apps_in_results = apps_in_results.intersection(connected_apps)
60
+
57
61
  for app, tools in app_tools.items():
58
62
  app_status = "connected" if app in connected_apps else "NOT connected"
59
63
  result_parts.append(f"Tools from {app} (status: {app_status} by user):")
@@ -62,6 +66,13 @@ def create_meta_tools(tool_registry: ToolRegistry) -> dict[str, Any]:
62
66
  result_parts.append(f" - {tool}")
63
67
  result_parts.append("") # Empty line between apps
64
68
 
69
+ # Add connection status information
70
+ if len(connected_apps_in_results) == 0 and len(apps_in_results) > 0:
71
+ result_parts.append("Connection Status: None of the apps in the results are connected. You must ask the user to choose the application.")
72
+ elif len(connected_apps_in_results) > 1:
73
+ connected_list = ", ".join(connected_apps_in_results)
74
+ result_parts.append(f"Connection Status: Multiple apps are connected ({connected_list}). You must ask the user to select which application they want to use.")
75
+
65
76
  result_parts.append("Call load_functions to select the required functions only.")
66
77
  return "\n".join(result_parts)
67
78
 
@@ -103,29 +114,30 @@ def create_meta_tools(tool_registry: ToolRegistry) -> dict[str, Any]:
103
114
  return f"Successfully loaded {len(tool_ids)} functions: {tool_ids}"
104
115
 
105
116
  @tool
106
- async def web_search(query: str) -> list:
107
- """Search the web for the given query and return structured search results.
117
+ async def web_search(query: str) -> dict:
118
+ """
119
+ Get an LLM answer to a question informed by Exa search results.
108
120
 
109
- Do not use for app-specific searches (for example, reddit or linkedin searches
121
+ This tool performs an Exa `/answer` request, which:
122
+ 1. Provides a **direct answer** for factual queries (e.g., "What is the capital of France?" → "Paris")
123
+ 2. Generates a **summary with citations** for open-ended questions
124
+ (e.g., "What is the state of AI in healthcare?" → A detailed summary with source links)
110
125
 
126
+ Args:
127
+ query (str): The question or topic to answer.
111
128
  Returns:
112
- list: A list of up to 10 search result dictionaries, each containing:
113
- - id (str): Unique identifier, typically the URL
114
- - title (str): The title/headline of the search result
115
- - url (str): The web URL of the result
116
- - publishedDate (str): ISO 8601 formatted date (e.g., "2025-01-01T00:00:00.000Z")
117
- - author (str): Author name (may be empty string)
118
- - summary (str): Text summary/snippet of the content
119
- - image (str): URL to associated image (if available)
120
-
121
- Example:
122
- results = await web_search(query="python programming")
129
+ dict: A structured response containing only:
130
+ - answer (str): Generated answer
131
+ - citations (list[dict]): List of cited sources
123
132
  """
124
- await tool_registry.export_tools(["exa__search_with_filters"], ToolFormat.LANGCHAIN)
125
- response = await tool_registry.call_tool(
126
- "exa__search_with_filters", {"query": query, "contents": {"summary": True}}
127
- )
128
- return response["results"]
133
+ await tool_registry.export_tools(["exa__answer"], ToolFormat.LANGCHAIN)
134
+ response = await tool_registry.call_tool("exa__answer", {"query": query, "text": True})
135
+
136
+ # Extract only desired fields
137
+ return {
138
+ "answer": response.get("answer"),
139
+ "citations": response.get("citations", []),
140
+ }
129
141
 
130
142
  return {"search_functions": search_functions, "load_functions": load_functions, "web_search": web_search}
131
143
 
@@ -6,7 +6,7 @@ from typing import Any
6
6
 
7
7
  from langchain_core.messages import BaseMessage
8
8
 
9
- MAX_CHARS = 700
9
+ MAX_CHARS = 5000
10
10
 
11
11
 
12
12
  def light_copy(data):
@@ -0,0 +1,90 @@
1
+ import contextlib
2
+ import inspect
3
+ import io
4
+ import queue
5
+ import re
6
+ import socket
7
+ import threading
8
+ import types
9
+ from typing import Any
10
+
11
+ from universal_mcp.agents.codeact0.utils import derive_context
12
+
13
+
14
+ class Sandbox:
15
+ """
16
+ A class to execute code safely in a sandboxed environment with a timeout.
17
+ """
18
+
19
+ def __init__(self, timeout: int = 180):
20
+ """
21
+ Initializes the Sandbox.
22
+ Args:
23
+ timeout: The timeout for code execution in seconds.
24
+ """
25
+ self.timeout = timeout
26
+ self._locals: dict[str, Any] = {}
27
+ self.add_context: dict[str, Any] = {}
28
+
29
+ def run(self, code: str) -> tuple[str, dict[str, Any], dict[str, Any]]:
30
+ """
31
+ Execute code safely with a timeout.
32
+ - Returns (output_str, filtered_locals_dict, new_add_context)
33
+ - Errors or timeout are returned as output_str.
34
+ - Previous variables in _locals persist across calls.
35
+ """
36
+
37
+ EXCLUDE_TYPES = (
38
+ types.ModuleType,
39
+ type(re.match("", "")),
40
+ type(threading.Lock()),
41
+ type(threading.RLock()),
42
+ threading.Event,
43
+ threading.Condition,
44
+ threading.Semaphore,
45
+ queue.Queue,
46
+ socket.socket,
47
+ io.IOBase,
48
+ )
49
+
50
+ result_container = {"output": "<no output>"}
51
+
52
+ def target():
53
+ try:
54
+ with contextlib.redirect_stdout(io.StringIO()) as f:
55
+ exec(code, self._locals, self._locals)
56
+ result_container["output"] = f.getvalue() or "<code ran, no output printed to stdout>"
57
+ except Exception as e:
58
+ result_container["output"] = "Error during execution: " + str(e)
59
+
60
+ thread = threading.Thread(target=target)
61
+ thread.start()
62
+ thread.join(self.timeout)
63
+
64
+ if thread.is_alive():
65
+ result_container["output"] = f"Code timeout: code execution exceeded {self.timeout} seconds."
66
+
67
+ # Filter locals for picklable/storable variables
68
+ all_vars = {}
69
+ for key, value in self._locals.items():
70
+ if key == "__builtins__":
71
+ continue
72
+ if inspect.iscoroutine(value) or inspect.iscoroutinefunction(value):
73
+ continue
74
+ if inspect.isasyncgen(value) or inspect.isasyncgenfunction(value):
75
+ continue
76
+ if isinstance(value, EXCLUDE_TYPES):
77
+ continue
78
+ if not callable(value) or not hasattr(value, "__name__"):
79
+ all_vars[key] = value
80
+
81
+ self._locals = all_vars
82
+
83
+ # Safely derive context
84
+ try:
85
+ self.add_context = derive_context(code, self.add_context)
86
+ except Exception:
87
+ # Keep the old context if derivation fails
88
+ pass
89
+
90
+ return result_container["output"], self._locals, self.add_context
File without changes
@@ -0,0 +1,160 @@
1
+ import fnmatch
2
+ import os
3
+ import pathlib
4
+ import re
5
+ import uuid
6
+
7
+ from loguru import logger
8
+ from universal_mcp.applications.application import BaseApplication
9
+
10
+
11
+ class FileSystemApp(BaseApplication):
12
+ """
13
+ A class to safely interact with the filesystem within a specified working directory.
14
+ """
15
+
16
+ def __init__(self, working_dir: str | None = None, **kwargs):
17
+ """
18
+ Initializes the FileSystemApp with a working directory.
19
+
20
+ Args:
21
+ working_dir: The absolute path to the directory where all operations will be performed.
22
+ """
23
+ super().__init__(name="Filesystem")
24
+
25
+ self.set_working_dir(working_dir or f"/tmp/{uuid.uuid4()}")
26
+
27
+ def set_working_dir(self, working_dir: str):
28
+ self.working_dir = pathlib.Path(working_dir).absolute()
29
+ # Create dir if not exists
30
+ self.working_dir.mkdir(parents=True, exist_ok=True)
31
+
32
+ def _is_safe_path(self, path: str) -> bool:
33
+ """
34
+ Checks if the given path is within the working directory.
35
+
36
+ Args:
37
+ path: The path to check.
38
+
39
+ Returns:
40
+ True if the path is safe, False otherwise.
41
+ """
42
+ common_path = os.path.commonpath([self.working_dir, path])
43
+ return common_path == str(self.working_dir)
44
+
45
+ def create_file(self, path: str, content: str = "") -> None:
46
+ """
47
+ Creates a file with the given content.
48
+
49
+ Args:
50
+ path: The relative path to the file to create.
51
+ content: The content to write to the file.
52
+
53
+ Raises:
54
+ ValueError: If the path is outside the working directory.
55
+ """
56
+ if not self._is_safe_path(path):
57
+ error = f"Path is outside the working directory: {path} vs {self.working_dir}"
58
+ logger.error(error)
59
+ raise ValueError(error)
60
+
61
+ full_path = os.path.join(self.working_dir, path)
62
+ os.makedirs(os.path.dirname(full_path), exist_ok=True)
63
+ with open(full_path, "w") as f:
64
+ f.write(content)
65
+
66
+ def read_file(self, path: str) -> str:
67
+ """
68
+ Reads the content of a file.
69
+
70
+ Args:
71
+ path: The relative path to the file to read.
72
+
73
+ Returns:
74
+ The content of the file.
75
+
76
+ Raises:
77
+ ValueError: If the path is outside the working directory.
78
+ FileNotFoundError: If the file does not exist.
79
+ """
80
+ if not self._is_safe_path(path):
81
+ raise ValueError("Path is outside the working directory.")
82
+
83
+ full_path = os.path.join(self.working_dir, path)
84
+ if not os.path.exists(full_path):
85
+ raise FileNotFoundError(f"File not found: {full_path}")
86
+
87
+ with open(full_path) as f:
88
+ return f.read()
89
+
90
+ def list_files(self, path: str = ".", recursive: bool = False) -> list[str]:
91
+ """
92
+ Lists files in a directory.
93
+
94
+ Args:
95
+ path: The relative path to the directory to list.
96
+ recursive: Whether to list files recursively.
97
+
98
+ Returns:
99
+ A list of file paths.
100
+
101
+ Raises:
102
+ ValueError: If the path is outside the working directory.
103
+ """
104
+ if not self._is_safe_path(path):
105
+ raise ValueError("Path is outside the working directory.")
106
+
107
+ full_path = os.path.join(self.working_dir, path)
108
+ if not os.path.isdir(full_path):
109
+ raise ValueError(f"Path '{path}' is not a directory.")
110
+
111
+ files = []
112
+ if recursive:
113
+ for root, _, filenames in os.walk(full_path):
114
+ for filename in filenames:
115
+ files.append(os.path.relpath(os.path.join(root, filename), self.working_dir))
116
+ else:
117
+ for item in os.listdir(full_path):
118
+ item_path = os.path.join(full_path, item)
119
+ if os.path.isfile(item_path):
120
+ files.append(os.path.relpath(item_path, self.working_dir))
121
+ return files
122
+
123
+ def grep(self, pattern: str, path: str = ".", file_pattern: str = "*") -> list[str]:
124
+ """
125
+ Searches for a pattern in files.
126
+
127
+ Args:
128
+ pattern: The regex pattern to search for.
129
+ path: The relative path to the directory to search in.
130
+ file_pattern: A glob pattern to filter files to search.
131
+
132
+ Returns:
133
+ A list of strings with "file:line_number:line" for each match.
134
+
135
+ Raises:
136
+ ValueError: If the path is outside the working directory.
137
+ """
138
+ if not self._is_safe_path(path):
139
+ raise ValueError("Path is outside the working directory.")
140
+
141
+ full_path = os.path.join(self.working_dir, path)
142
+ if not os.path.isdir(full_path):
143
+ raise ValueError(f"Path '{path}' is not a directory.")
144
+
145
+ matches = []
146
+ for root, _, filenames in os.walk(full_path):
147
+ for filename in fnmatch.filter(filenames, file_pattern):
148
+ file_path = os.path.join(root, filename)
149
+ try:
150
+ with open(file_path, errors="ignore") as f:
151
+ for i, line in enumerate(f, 1):
152
+ if re.search(pattern, line):
153
+ relative_path = os.path.relpath(file_path, self.working_dir)
154
+ matches.append(f"{relative_path}:{i}:{line.strip()}")
155
+ except OSError:
156
+ continue # Skip files that can't be opened
157
+ return matches
158
+
159
+ def list_tools(self):
160
+ return [self.create_file, self.grep, self.list_files, self.read_file]
@@ -1,3 +1,3 @@
1
- from .app import LLMApp
1
+ from .app import LlmApp
2
2
 
3
- __all__ = ["LLMApp"]
3
+ __all__ = ["LlmApp"]
@@ -2,10 +2,11 @@ import json
2
2
  from typing import Any, Literal, cast
3
3
 
4
4
  from langchain.chat_models import init_chat_model
5
- from langchain_openai import AzureChatOpenAI
6
5
  from pydantic import BaseModel, Field
7
6
  from universal_mcp.applications.application import BaseApplication
8
7
 
8
+ from universal_mcp.agents.llm import load_chat_model
9
+
9
10
  MAX_RETRIES = 3
10
11
 
11
12
 
@@ -28,7 +29,7 @@ def _get_context_as_string(source: Any | list[Any] | dict[str, Any]) -> str:
28
29
  return "\n".join(f"<{k}>\n{str(v)}\n</{k}>" for k, v in source.items())
29
30
 
30
31
 
31
- class LLMApp(BaseApplication):
32
+ class LlmApp(BaseApplication):
32
33
  """
33
34
  An application for leveraging Large Language Models (LLMs) for advanced text processing tasks.
34
35
  """
@@ -40,22 +41,34 @@ class LLMApp(BaseApplication):
40
41
  def generate_text(
41
42
  self,
42
43
  task: str,
43
- context: Any | list[Any] | dict[str, Any],
44
+ context: str | list[str] | dict[str, str] = "",
44
45
  tone: str = "normal",
45
46
  output_format: Literal["markdown", "html", "plain"] = "markdown",
46
47
  length: Literal["very-short", "concise", "normal", "long"] = "concise",
47
48
  ) -> str:
48
49
  """
49
- Generates well-written text for a high-level task using the provided context.
50
-
51
- Use this function for creative writing, summarization, and other text generation tasks.
50
+ Given a high-level writing task and context, returns a well-written text
51
+ that achieves the task, given the context.
52
+
53
+ Example Call:
54
+ generate_text("Summarize this website with the goal of making it easy to understand.", web_content)
55
+ generate_text("Make a markdown table summarizing the key differences between doc_1 and doc_2.", {"doc_1": str(doc_1), "doc_2": str(doc_2)})
56
+ generate_text("Summarize all the provided documents.", [doc_1, doc_2, doc_3])
57
+
58
+ Important:
59
+ - Include specifics of the goal in the context verbatim.
60
+ - Be precise and direct in the task, and include as much context as possible.
61
+ - Include relevant high-level goals or intent in the task.
62
+ - You can provide multiple documents as input, and reference them in the task.
63
+ - You MUST provide the contents of any source documents to `generate_text`.
64
+ - NEVER use `generate_text` to produce JSON for a Pydantic model.
52
65
 
53
66
  Args:
54
67
  task: The main writing task or directive.
55
- context: A single string, list of strings, or dictionary mapping labels to content.
56
- tone: The desired tone of the output (e.g., "formal", "casual", "technical").
68
+ context: A single string, list of strings, or dict mapping labels to content.
69
+ tone: The desired tone of the output (e.g., "normal", "flirty", "formal", "casual", "crisp", "poetic", "technical", "internet-chat", "smartass", etc.).
57
70
  output_format: The desired output format ('markdown', 'html', 'plain').
58
- length: The desired length of the output ('very-short', 'concise', 'normal', 'long').
71
+ length: Desired length of the output ('very-short', 'concise', 'normal', 'long').
59
72
 
60
73
  Returns:
61
74
  The generated text as a string.
@@ -78,31 +91,63 @@ class LLMApp(BaseApplication):
78
91
 
79
92
  full_prompt = f"{prompt}\n\nContext:\n{context_str}\n\n"
80
93
 
81
- model = AzureChatOpenAI(model="gpt-4o", temperature=0.7)
94
+ model = load_chat_model("azure/gpt-5-mini")
82
95
  response = model.with_retry(stop_after_attempt=MAX_RETRIES).invoke(full_prompt)
83
96
  return str(response.content)
84
97
 
85
98
  def classify_data(
86
99
  self,
87
- task: str,
100
+ classification_task_and_requirements: str,
88
101
  context: Any | list[Any] | dict[str, Any],
89
102
  class_descriptions: dict[str, str],
90
103
  ) -> dict[str, Any]:
91
104
  """
92
- Classifies data into one of several categories based on a given task and context.
105
+ Classifies and compares data based on given requirements.
106
+
107
+ Use `classify_data` for tasks which need to classify data into one of many categories.
108
+ If making multiple binary classifications, call `classify_data` for each.
109
+
110
+ Guidance:
111
+ - Prefer to use classify_data operations to compare strings, rather than string ops.
112
+ - Prefer to include an "Unsure" category for classification tasks.
113
+ - The `class_descriptions` dict argument MUST be a map from possible class names to a precise description.
114
+ - Use precise and specific class names and concise descriptions.
115
+ - Pass ALL relevant context, preferably as a dict mapping labels to content.
116
+ - Returned dict maps each possible class name to a probability.
117
+
118
+ Example Usage:
119
+ classification_task_and_requirements = "Does the document contain an address?"
120
+ class_descriptions = {
121
+ "Is_Address": "Valid addresses usually have street names, city, and zip codes.",
122
+ "Not_Address": "Not valid addresses."
123
+ }
124
+ classification = classify_data(
125
+ classification_task_and_requirements,
126
+ {"address": extracted_address},
127
+ class_descriptions
128
+ )
129
+ if classification["probabilities"]["Is_Address"] > 0.5:
130
+ ...
93
131
 
94
132
  Args:
95
- task: The classification question and any specific rules or requirements.
96
- context: The data to be classified, provided as a string, list, or dictionary.
97
- class_descriptions: A dictionary mapping class names to their descriptions.
133
+ classification_task_and_requirements: The classification question and rules.
134
+ context: The data to classify (string, list, or dict).
135
+ class_descriptions: Mapping from class names to descriptions.
136
+
137
+ Tags:
138
+ important
98
139
 
99
140
  Returns:
100
- A dictionary containing the classification probabilities, the reasoning, and the top class.
141
+ dict: {
142
+ "probabilities": dict[str, float],
143
+ "reason": str,
144
+ "top_class": str,
145
+ }
101
146
  """
102
147
  context_str = _get_context_as_string(context)
103
148
 
104
149
  prompt = (
105
- f"{task}\n\n"
150
+ f"{classification_task_and_requirements}\n\n"
106
151
  f"This is a classification task.\nPossible classes and descriptions:\n"
107
152
  f"{json.dumps(class_descriptions, indent=2)}\n\n"
108
153
  f"Context:\n{context_str}\n\n"
@@ -125,25 +170,61 @@ class LLMApp(BaseApplication):
125
170
 
126
171
  def extract_data(
127
172
  self,
128
- task: str,
173
+ extraction_task: str,
129
174
  source: Any | list[Any] | dict[str, Any],
130
175
  output_schema: dict[str, Any],
131
176
  ) -> dict[str, Any]:
132
177
  """
133
- Extracts structured data from unstructured text based on a provided JSON schema.
178
+ Extracts structured data from unstructured data (documents, webpages, images, large bodies of text),
179
+ returning a dictionary matching the given output_schema.
180
+
181
+ You MUST anticipate Exception raised for unextractable data; skip this item if applicable.
182
+
183
+ Strongly prefer to:
184
+ - Be comprehensive, specific, and precise on the data you want to extract.
185
+ - Use optional fields everywhere.
186
+ - Extract multiple items from each source unless otherwise specified.
187
+ - The more specific your extraction task and output_schema are, the better the results.
134
188
 
135
189
  Args:
136
- task: A description of the data to be extracted.
137
- source: The unstructured data to extract from (e.g., document, webpage content).
138
- output_schema: A valid JSON schema with a 'title' and 'description'.
190
+ extraction_task: The directive describing what to extract.
191
+ source: The unstructured data to extract from.
192
+ output_schema: must be a valid JSON schema with top-level 'title' and 'description' keys.
139
193
 
140
194
  Returns:
141
195
  A dictionary containing the extracted data, matching the provided schema.
196
+
197
+ Example:
198
+ news_articles_schema = {
199
+ "title": "NewsArticleList",
200
+ "description": "A list of news articles with headlines and URLs",
201
+ "type": "object",
202
+ "properties": {
203
+ "articles": {
204
+ "type": "array",
205
+ "items": {
206
+ "type": "object",
207
+ "properties": {
208
+ "headline": {
209
+ "type": "string"
210
+ },
211
+ "url": {
212
+ "type": "string"
213
+ }
214
+ },
215
+ "required": ["headline", "url"]
216
+ }
217
+ }
218
+ },
219
+ "required": ["articles"]
220
+ }
221
+
222
+ news_articles = extract_data("Extract headlines and their corresponding URLs.", content, news_articles_schema)
142
223
  """
143
224
  context_str = _get_context_as_string(source)
144
225
 
145
226
  prompt = (
146
- f"{task}\n\n"
227
+ f"{extraction_task}\n\n"
147
228
  f"Context:\n{context_str}\n\n"
148
229
  "Return ONLY a valid JSON object that conforms to the provided schema, with no extra text."
149
230
  )
@@ -156,3 +237,64 @@ class LLMApp(BaseApplication):
156
237
  .invoke(prompt)
157
238
  )
158
239
  return cast(dict[str, Any], response)
240
+
241
+ def call_llm(
242
+ self,
243
+ task_instructions: str,
244
+ context: Any | list[Any] | dict[str, Any],
245
+ output_schema: dict[str, Any],
246
+ ) -> dict[str, Any]:
247
+ """
248
+ Call a Large Language Model (LLM) with an instruction and contextual information,
249
+ returning a dictionary matching the given output_schema.
250
+ Can be used for tasks like creative writing, llm reasoning based content generation, etc.
251
+
252
+ You MUST anticipate Exceptions in reasoning based tasks which will lead to some empty fields
253
+ in the returned output; skip this item if applicable.
254
+
255
+ General Guidelines:
256
+ - Be comprehensive, specific, and precise on the task instructions.
257
+ - Include as much context as possible.
258
+ - You can provide multiple items in context, and reference them in the task.
259
+ - Include relevant high-level goals or intent in the task.
260
+ - In the output_schema, use required field wherever necessary.
261
+ - The more specific your task instructions and output_schema are, the better the results.
262
+
263
+ Guidelines for content generation tasks:
264
+ - Feel free to add instructions for tone, length, and format (markdown, html, plain-text, xml)
265
+ - Some examples of tone are: "normal", "flirty", "formal", "casual", "crisp", "poetic", "technical", "internet-chat", "smartass", etc.
266
+ - Prefer length to be concise by default. Other examples are: "very-short", "concise", "normal", "long", "2-3 lines", etc.
267
+ - In format prefer plain-text but you can also use markdown and html wherever useful.
268
+
269
+ Args:
270
+ task_instructions: The main directive for the LLM (e.g., "Summarize the article" or "Extract key entities").
271
+ context:
272
+ A dictionary containing named text elements that provide additional
273
+ information for the LLM. Keys are labels (e.g., 'article', 'transcript'),
274
+ values are strings of content.
275
+ output_schema: must be a valid JSON schema with top-level 'title' and 'description' keys.
276
+
277
+ Returns:
278
+ dict: Parsed JSON object matching the desired output_schema.
279
+
280
+ """
281
+ context_str = _get_context_as_string(context)
282
+
283
+ prompt = f"{task_instructions}\n\nContext:\n{context_str}\n\nReturn ONLY a valid JSON object, no extra text."
284
+
285
+ model = init_chat_model(model="claude-4-sonnet-20250514", temperature=0)
286
+
287
+ response = (
288
+ model.with_structured_output(schema=output_schema, method="json_mode")
289
+ .with_retry(stop_after_attempt=MAX_RETRIES)
290
+ .invoke(prompt)
291
+ )
292
+ return cast(dict[str, Any], response)
293
+
294
+ def list_tools(self):
295
+ return [
296
+ self.generate_text,
297
+ self.classify_data,
298
+ self.extract_data,
299
+ self.call_llm,
300
+ ]
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: universal-mcp-agents
3
- Version: 0.1.17
3
+ Version: 0.1.19
4
4
  Summary: Add your description here
5
5
  Project-URL: Homepage, https://github.com/universal-mcp/applications
6
6
  Project-URL: Repository, https://github.com/universal-mcp/applications
@@ -1,9 +1,10 @@
1
- universal_mcp/agents/__init__.py,sha256=5aE4zVgem6ehzfRrt5QqE6gLi7949vySZAn_uFuaU7Q,1252
2
- universal_mcp/agents/base.py,sha256=GmagwWFJdPpp_yLeAfsrr4fu-zzxh7CgIlBJbTgFgWM,7072
3
- universal_mcp/agents/cli.py,sha256=AG9e4iSX3GazT537573YrYT1wSaZYOr42rrYQ7xP3YA,1016
1
+ universal_mcp/agents/__init__.py,sha256=NuS_JOKYeHfAkNA4xRyvxdGZP_tXKfvsmoc2fhkMrj8,1259
2
+ universal_mcp/agents/base.py,sha256=hrlvSv0TLS-ShaKofpHD3kishciql7SSbTKCUHKGObc,7175
3
+ universal_mcp/agents/cli.py,sha256=bXdpgxsOMjclm1STHJgx10ocX9EebQ11DrxH0p6KMZk,943
4
4
  universal_mcp/agents/hil.py,sha256=_5PCK6q0goGm8qylJq44aSp2MadP-yCPvhOJYKqWLMo,3808
5
5
  universal_mcp/agents/llm.py,sha256=hVRwjZs3MHl5_3BWedmurs2Jt1oZDfFX0Zj9F8KH7fk,1787
6
6
  universal_mcp/agents/react.py,sha256=8XQvJ0HLVgc-K0qn9Ml48WGcgUGuIKtL67HatlT6Da0,3334
7
+ universal_mcp/agents/sandbox.py,sha256=Int2O8JNFPlB8c7gb86KRxlNbuV0zdz5_NCo_GMcCds,2876
7
8
  universal_mcp/agents/simple.py,sha256=NSATg5TWzsRNS7V3LFiDG28WSOCIwCdcC1g7NRwg2nM,2095
8
9
  universal_mcp/agents/utils.py,sha256=P6W9k6XAOBp6tdjC2VTP4tE0B2M4-b1EDmr-ylJ47Pw,7765
9
10
  universal_mcp/agents/bigtool/__init__.py,sha256=mZG8dsaCVyKlm82otxtiTA225GIFLUCUUYPEIPF24uw,2299
@@ -28,23 +29,25 @@ universal_mcp/agents/codeact/sandbox.py,sha256=NjN6ISj8psFtHf8V0w24ChJdUMUWkq7Or
28
29
  universal_mcp/agents/codeact/state.py,sha256=WTPfpxDlGRnlr5tZuXMg_KU7GS7TZbnrIKslOvZLbQI,565
29
30
  universal_mcp/agents/codeact/utils.py,sha256=JUbT_HYGS_D1BzmzoVpORIe7SGur1KgJguTZ_1tZ4JY,1918
30
31
  universal_mcp/agents/codeact0/__init__.py,sha256=ebKkpgg-0UnsvDtagEJ2tMer1VsfhmEE5KJcFzUk9fU,133
31
- universal_mcp/agents/codeact0/__main__.py,sha256=V2wLWW9ym3rtiSvPEs-N0Mki7G5dYHzV5dAsAoF-ygQ,1148
32
+ universal_mcp/agents/codeact0/__main__.py,sha256=xeqNuawP9M8JVAnkhLesalnpI_TakC49ATJaSCzCsYs,880
32
33
  universal_mcp/agents/codeact0/agent.py,sha256=9BInAQr3csES-XHSscmeJlYJ3-wQUHPvLOf-6wFILUU,6695
33
34
  universal_mcp/agents/codeact0/config.py,sha256=H-1woj_nhSDwf15F63WYn723y4qlRefXzGxuH81uYF0,2215
34
35
  universal_mcp/agents/codeact0/langgraph_agent.py,sha256=ehjMV_Z1118pCFWB_Sa5H7XnUp0udsbUHjfjXjhIQM8,435
35
- universal_mcp/agents/codeact0/llm_tool.py,sha256=fydA4BbSnhG3OLMR_VEtJxTPTmRK_mMvXGsUCO-5Mig,13829
36
- universal_mcp/agents/codeact0/playbook_agent.py,sha256=YNuWQ_xK9KlOtb4kkHkxLl3CEJfmM3sUf8dwaSPyYcg,18089
37
- universal_mcp/agents/codeact0/prompts.py,sha256=j8HxA3Rp-EZsms9qMBcRmFrUjeySrG1IWjqrNFXZZn8,9457
36
+ universal_mcp/agents/codeact0/llm_tool.py,sha256=q-hiqkKtjVmpyNceFoRgo7hvKh4HtQf_I1VudRUEPR0,11075
37
+ universal_mcp/agents/codeact0/playbook_agent.py,sha256=6ePcpEOrHxoNoaAdKdg9i7Yi6hcJOxBAo1MavCn8J6A,18081
38
+ universal_mcp/agents/codeact0/prompts.py,sha256=2MF0J371Ib7cDXqpW6Ei_CwBpRFmK95neLh9QM4emIY,8708
38
39
  universal_mcp/agents/codeact0/sandbox.py,sha256=zMgHrWnQYkSkJb2MzfXvT3euCc4hvqzBE_EbX2_iLxA,3142
39
40
  universal_mcp/agents/codeact0/state.py,sha256=Y-Rzn_S7--aXH18KPvyhqDqOOB-miu1lsAmLgmMlaAg,1259
40
- universal_mcp/agents/codeact0/tools.py,sha256=Adh9Go87JnIk5U_rtBOHglFRPowk2JmRK9bumR4zJoo,7751
41
- universal_mcp/agents/codeact0/utils.py,sha256=ACncBI9-JJ4Qe_4VHWnSziY3P2nE6Vb9pj71NPeVZ0o,15943
41
+ universal_mcp/agents/codeact0/tools.py,sha256=qVZLq1YlVKABZdpEqFAzLo04DTFd1ZJi18atfzmxEb8,8374
42
+ universal_mcp/agents/codeact0/utils.py,sha256=jAZItSd3KGDkY9PquSWRIFCj9N26K9Kt0HKQ_jwvvSQ,15944
42
43
  universal_mcp/agents/shared/__main__.py,sha256=XxH5qGDpgFWfq7fwQfgKULXGiUgeTp_YKfcxftuVZq8,1452
43
44
  universal_mcp/agents/shared/prompts.py,sha256=yjP3zbbuKi87qCj21qwTTicz8TqtkKgnyGSeEjMu3ho,3761
44
45
  universal_mcp/agents/shared/tool_node.py,sha256=DC9F-Ri28Pam0u3sXWNODVgmj9PtAEUb5qP1qOoGgfs,9169
45
- universal_mcp/applications/llm/__init__.py,sha256=xnpxq4Wl_pevvwtSUtEwcty8_d61ywO1V2YnEXyCREY,46
46
- universal_mcp/applications/llm/app.py,sha256=iNLU6z2LRZc01GfSKvV0vNzT1LhKAjq_UrSJEmjthjw,6032
46
+ universal_mcp/applications/filesystem/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
47
+ universal_mcp/applications/filesystem/app.py,sha256=0TRjjm8YnslVRSmfkXI7qQOAlqWlD1eEn8Jm0xBeigs,5561
48
+ universal_mcp/applications/llm/__init__.py,sha256=_XGRxN3O1--ZS5joAsPf8IlI9Qa6negsJrwJ5VJXno0,46
49
+ universal_mcp/applications/llm/app.py,sha256=oqX3byvlFRmeRo4jJJxUBGy-iTDGm2fplMEKA2pcMtw,12743
47
50
  universal_mcp/applications/ui/app.py,sha256=c7OkZsO2fRtndgAzAQbKu-1xXRuRp9Kjgml57YD2NR4,9459
48
- universal_mcp_agents-0.1.17.dist-info/METADATA,sha256=GFVLLOc3wPsDCvYnLV7w1klLJJIITEax5ipX_MScI98,878
49
- universal_mcp_agents-0.1.17.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
50
- universal_mcp_agents-0.1.17.dist-info/RECORD,,
51
+ universal_mcp_agents-0.1.19.dist-info/METADATA,sha256=imopF36ZWDrHP3tweT0GtvIlKdNg_CLa9yZrXvv3s9s,878
52
+ universal_mcp_agents-0.1.19.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
53
+ universal_mcp_agents-0.1.19.dist-info/RECORD,,