refactor: Remove all imports from deepagents_cli & pull in custom tools (#894)

apps/agent/agent/prompt.py

@@ -0,0 +1,14 @@
+SYSTEM_PROMPT = """### Current Working Directory
+
+You are operating in a **remote Linux sandbox** at `{working_dir}`.
+
+All code execution and file operations happen in this sandbox environment.
+
+**Important:**
+- Use `{working_dir}` as your working directory for all operations
+
+"""
+
+
+def construct_system_prompt(working_dir: str) -> str:
+    return SYSTEM_PROMPT.format(working_dir=working_dir)

apps/agent/agent/server.py

@@ -6,19 +6,16 @@
 import logging
 import os
 import warnings
-from collections.abc import Sequence
 from typing import Any
 
 logger = logging.getLogger(__name__)
 
 from langchain.agents.middleware import AgentState, after_agent, after_model, before_model
-from langchain.agents.middleware.types import AgentMiddleware
-from langchain.tools import BaseTool
-from langchain_core.language_models import BaseChatModel
 from langgraph.config import get_config, get_store
 from langgraph.graph.state import RunnableConfig
 from langgraph.pregel import Pregel
 from langgraph.runtime import Runtime
+from langgraph_sdk import get_client
 
 warnings.filterwarnings("ignore", module="langchain_core._api.deprecation")
 
@@ -30,12 +27,11 @@ warnings.filterwarnings("ignore", message=".*Pydantic V1.*", category=UserWarnin
 # Now safe to import agent (which imports LangChain modules)
 from deepagents import create_deep_agent
 from deepagents.backends.sandbox import SandboxBackendProtocol
-from deepagents_cli.agent import get_system_prompt
-from deepagents_cli.config import config, settings
-from deepagents_cli.tools import fetch_url, http_request, web_search
 
 # Local import for encryption
 from .encryption import decrypt_token
+from .prompt import construct_system_prompt
+from .tools import fetch_url, http_request
 
 
 def _get_langsmith_api_key() -> str | None:
@@ -88,69 +84,6 @@ def _create_langsmith_sandbox(
     )
 
 
-def create_server_agent(
-    model: str | BaseChatModel | None,
-    assistant_id: str,
-    *,
-    tools: list[BaseTool] | None = None,
-    sandbox: SandboxBackendProtocol | None = None,
-    sandbox_type: str | None = None,
-    system_prompt: str | None = None,
-    auto_approve: bool = True,  # noqa: ARG001 - Always True for Open SWE
-    working_dir: str | None = None,
-    middleware: Sequence[AgentMiddleware] = (),
-) -> Pregel:
-    """Create a server-mode agent for Open SWE.
-
-    This creates an agent configured for server/cloud deployment with sandbox
-    support and custom middleware. Always runs with auto_approve=True.
-
-    Args:
-        model: LLM model to use. Can be None for introspection-only mode.
-        assistant_id: Agent identifier for memory/state storage
-        tools: Additional tools to provide to agent
-        sandbox: Optional sandbox backend for remote execution (e.g., LangSmithBackend).
-        sandbox_type: Type of sandbox provider ("langsmith").
-                     Used for system prompt generation.
-        system_prompt: Override the default system prompt. If None, generates one
-                      based on sandbox_type and assistant_id.
-        working_dir: Override the default working directory (e.g., cloned repo path).
-                    Used in system prompt to tell the agent where to operate.
-        middleware: Sequence of middleware to apply to the agent.
-
-    Returns:
-        Configured LangGraph Pregel instance ready for execution
-    """
-    agent_tools = tools or []
-
-    # Get or use custom system prompt
-    if system_prompt is None:
-        if sandbox_type is not None:
-            system_prompt = get_system_prompt(
-                assistant_id=assistant_id,
-                sandbox_type=sandbox_type,
-                working_dir=working_dir,
-            )
-        else:
-            # Only happens when thread_id is None / not actually running
-            system_prompt = ""
-
-    return create_deep_agent(
-        model=model,
-        system_prompt=system_prompt,
-        tools=agent_tools,
-        backend=sandbox,
-        middleware=middleware,
-        interrupt_on={},  # Always auto-approve for Open SWE
-    ).with_config(config)
-
-
-tools = [http_request, fetch_url]
-if settings.has_tavily:
-    tools.append(web_search)
-
-from langgraph_sdk import get_client
-
 client = get_client()
 
 SANDBOX_CREATING = "__creating__"
@@ -867,11 +800,16 @@ def graph_loaded_for_execution(config: RunnableConfig) -> bool:
     )
 
 
+DEFAULT_RECURSION_LIMIT = 1_000
+
+
 async def get_agent(config: RunnableConfig) -> Pregel:  # noqa: PLR0915
     """Get or create an agent with a sandbox for the given thread."""
     thread_id = config["configurable"].get("thread_id", None)
     logger.info("get_agent called for thread %s", thread_id)
 
+    config["recursion_limit"] = DEFAULT_RECURSION_LIMIT
+
     repo_config = config["configurable"].get("repo", {})
     repo_owner = repo_config.get("owner")
     repo_name = repo_config.get("name")
@@ -883,14 +821,10 @@ async def get_agent(config: RunnableConfig) -> Pregel:  # noqa: PLR0915
 
     if thread_id is None or not graph_loaded_for_execution(config):
         logger.info("No thread_id or not for execution, returning agent without sandbox")
-        return create_server_agent(
-            model=None,
-            assistant_id="agent",
-            tools=tools,
-            sandbox=None,
-            sandbox_type=None,
-            auto_approve=True,
-        )
+        return create_deep_agent(
+            system_prompt="",
+            tools=[],
+        ).with_config(config)
 
     sandbox_id = await _get_sandbox_id_from_metadata(thread_id)
 
@@ -938,14 +872,10 @@ async def get_agent(config: RunnableConfig) -> Pregel:  # noqa: PLR0915
         logger.info("Connecting to existing sandbox %s", sandbox_id)
         try:
             # Connect to existing sandbox without context manager cleanup
-            sandbox_backend = await asyncio.to_thread(
-                _create_langsmith_sandbox, sandbox_id
-            )
+            sandbox_backend = await asyncio.to_thread(_create_langsmith_sandbox, sandbox_id)
             logger.info("Connected to existing sandbox %s", sandbox_id)
         except Exception:
-            logger.warning(
-                "Failed to connect to existing sandbox %s, creating new one", sandbox_id
-            )
+            logger.warning("Failed to connect to existing sandbox %s, creating new one", sandbox_id)
             # Reset sandbox_id and create a new sandbox
             await client.threads.update(
                 thread_id=thread_id,
@@ -962,9 +892,7 @@ async def get_agent(config: RunnableConfig) -> Pregel:  # noqa: PLR0915
                 )
             except Exception:
                 logger.exception("Failed to create replacement sandbox")
-                await client.threads.update(
-                    thread_id=thread_id, metadata={"sandbox_id": None}
-                )
+                await client.threads.update(thread_id=thread_id, metadata={"sandbox_id": None})
                 raise
 
         thread = await client.threads.get(thread_id=thread_id)
@@ -983,17 +911,14 @@ async def get_agent(config: RunnableConfig) -> Pregel:  # noqa: PLR0915
     _SANDBOX_BACKENDS[thread_id] = sandbox_backend
 
     logger.info("Returning agent with sandbox for thread %s", thread_id)
-    return create_server_agent(
-        model=None,
-        assistant_id="agent",
-        tools=tools,
-        sandbox=sandbox_backend,
-        sandbox_type="langsmith",
-        auto_approve=True,
-        working_dir=repo_dir,
+    return create_deep_agent(
+        model=None,  # TODO: Actually pass a model here
+        system_prompt=construct_system_prompt(repo_dir),
+        tools=[http_request, fetch_url],
+        backend=sandbox_backend,
         middleware=[
             check_message_queue_before_model,
             post_to_linear_after_model,
             open_pr_if_needed,
         ],
-    )
+    ).with_config(config)

apps/agent/agent/tools/__init__.py

@@ -0,0 +1,4 @@
+from .fetch_url import fetch_url
+from .http_request import http_request
+
+__all__ = ["fetch_url", "http_request"]

apps/agent/agent/tools/fetch_url.py

@@ -0,0 +1,50 @@
+from typing import Any
+
+import requests
+from markdownify import markdownify
+
+
+def fetch_url(url: str, timeout: int = 30) -> dict[str, Any]:
+    """Fetch content from a URL and convert HTML to markdown format.
+
+    This tool fetches web page content and converts it to clean markdown text,
+    making it easy to read and process HTML content. After receiving the markdown,
+    you MUST synthesize the information into a natural, helpful response for the user.
+
+    Args:
+        url: The URL to fetch (must be a valid HTTP/HTTPS URL)
+        timeout: Request timeout in seconds (default: 30)
+
+    Returns:
+        Dictionary containing:
+        - success: Whether the request succeeded
+        - url: The final URL after redirects
+        - markdown_content: The page content converted to markdown
+        - status_code: HTTP status code
+        - content_length: Length of the markdown content in characters
+
+    IMPORTANT: After using this tool:
+    1. Read through the markdown content
+    2. Extract relevant information that answers the user's question
+    3. Synthesize this into a clear, natural language response
+    4. NEVER show the raw markdown to the user unless specifically requested
+    """
+    try:
+        response = requests.get(
+            url,
+            timeout=timeout,
+            headers={"User-Agent": "Mozilla/5.0 (compatible; DeepAgents/1.0)"},
+        )
+        response.raise_for_status()
+
+        # Convert HTML content to markdown
+        markdown_content = markdownify(response.text)
+
+        return {
+            "url": str(response.url),
+            "markdown_content": markdown_content,
+            "status_code": response.status_code,
+            "content_length": len(markdown_content),
+        }
+    except requests.exceptions.RequestException as e:
+        return {"error": f"Fetch URL error: {e!s}", "url": url}

apps/agent/agent/tools/http_request.py

@@ -0,0 +1,70 @@
+from typing import Any
+
+import requests
+
+
+def http_request(
+    url: str,
+    method: str = "GET",
+    headers: dict[str, str] | None = None,
+    data: str | dict | None = None,
+    params: dict[str, str] | None = None,
+    timeout: int = 30,
+) -> dict[str, Any]:
+    """Make HTTP requests to APIs and web services.
+
+    Args:
+        url: Target URL
+        method: HTTP method (GET, POST, PUT, DELETE, etc.)
+        headers: HTTP headers to include
+        data: Request body data (string or dict)
+        params: URL query parameters
+        timeout: Request timeout in seconds
+
+    Returns:
+        Dictionary with response data including status, headers, and content
+    """
+    try:
+        kwargs: dict[str, Any] = {}
+
+        if headers:
+            kwargs["headers"] = headers
+        if params:
+            kwargs["params"] = params
+        if data:
+            if isinstance(data, dict):
+                kwargs["json"] = data
+            else:
+                kwargs["data"] = data
+
+        response = requests.request(method.upper(), url, timeout=timeout, **kwargs)
+
+        try:
+            content = response.json()
+        except (ValueError, requests.exceptions.JSONDecodeError):
+            content = response.text
+
+        return {
+            "success": response.status_code < 400,
+            "status_code": response.status_code,
+            "headers": dict(response.headers),
+            "content": content,
+            "url": response.url,
+        }
+
+    except requests.exceptions.Timeout:
+        return {
+            "success": False,
+            "status_code": 0,
+            "headers": {},
+            "content": f"Request timed out after {timeout} seconds",
+            "url": url,
+        }
+    except requests.exceptions.RequestException as e:
+        return {
+            "success": False,
+            "status_code": 0,
+            "headers": {},
+            "content": f"Request error: {e!s}",
+            "url": url,
+        }

apps/agent/pyproject.toml

@@ -8,26 +8,21 @@ license = { text = "MIT" }
 dependencies = [
     # Core deepagents CLI - points to branch with LangSmith sandbox and working_dir support
     "deepagents-cli @ git+https://github.com/langchain-ai/deepagents.git@yogesh/working_dir_and_template_config#subdirectory=libs/cli",
-    
     # FastAPI for webhook handling
     "fastapi>=0.104.0",
     "uvicorn>=0.24.0",
-    
     # HTTP client
     "httpx>=0.25.0",
-    
     # JWT for service authentication
     "PyJWT>=2.8.0",
-    
     # Encryption
     "cryptography>=41.0.0",
-    
     # LangGraph SDK for thread management
     "langgraph-sdk>=0.1.0",
-    
     # LangChain dependencies (will be pulled in by deepagents-cli but listing for clarity)
     "langchain>=0.2.0",
     "langgraph>=0.2.0",
+    "markdownify>=1.2.2",
 ]
 
 [project.optional-dependencies]