Skip to content

armgit / E2B

Public Imported from GitHub
0

Open-source, secure environment with real-world tools for enterprise-grade agents.

Add tool for testing server [WIP]

Tomas Valenta2e9103c9de43
diff --git a/api-service/codegen/tools/playground/playground.py b/api-service/codegen/tools/playground/playground.py
index a03127f3a..8fcc7437f 100644
--- a/api-service/codegen/tools/playground/playground.py
+++ b/api-service/codegen/tools/playground/playground.py
@@ -1,4 +1,5 @@
import os
+import time
import playground_client
@@ -14,6 +15,9 @@ configuration.cert_file = None
class Playground:
+ port_check_interval = 1 # 1s
+ max_port_checks = 10
+
def __init__(self, env_id: str):
self.client = playground_client.ApiClient(configuration)
self.api = playground_client.DefaultApi(self.client)
@@ -31,9 +35,36 @@ class Playground:
self.api.delete_session(self.session.id)
self.client.close()
+ def get_open_ports(self):
+ response = self.api.get_session(
+ id=self.session.id,
+ )
+ return response.ports
+
+ def is_port_open(self, port: float):
+ open_ports = self.get_open_ports()
+ return any(open_port.port == port and open_port.state == "LISTEN" for open_port in open_ports)
+
def run_command(self, cmd: str, rootdir: str = "/", env_vars = {}):
- return self.api.run_process(
- self.session.id, playground_client.RunProcessParams(cmd=cmd, envVars=env_vars, rootdir=rootdir)
+ return self.api.start_process(
+ self.session.id,
+ playground_client.StartProcessParams(cmd=cmd, envVars=env_vars, rootdir=rootdir),
+ wait=True
+ )
+
+ def start_process(self, cmd: str, rootdir: str = "/", env_vars = {}):
+ """Start process and return the process ID."""
+ response = self.api.start_process(
+ self.session.id,
+ playground_client.StartProcessParams(cmd=cmd, envVars=env_vars, rootdir=rootdir),
+ )
+ return response.process_id
+
+ def stop_process(self, process_id: str):
+ return self.api.stop_process(
+ self.session.id,
+ process_id=process_id,
+ results=True,
)
def read_file(self, path: str):
@@ -56,6 +87,18 @@ class Playground:
def make_dir(self, path: str):
self.api.make_filesystem_dir(self.session.id, path)
+ def test_server_code(self, server_cmd: str, test_cmd: str, port: float, rootdir: str):
+ server_process_id = self.start_process(cmd=server_cmd, rootdir=rootdir)
+
+ for _ in range(self.max_port_checks):
+ if self.is_port_open(port):
+ break
+ time.sleep(self.port_check_interval)
+
+ test_result = self.run_command(cmd=test_cmd, rootdir=rootdir)
+ server_result = self.stop_process(server_process_id)
+ return test_result, server_result
+
class NodeJSPlayground(Playground):
node_js_env_id = "QfqlXmCo5iKl"
@@ -85,6 +128,7 @@ class NodeJSPlayground(Playground):
def install_dependencies(self, dependencies: str):
return self.run_command(f"npm install {dependencies}", rootdir=self.rootdir)
- # TODO: How to handle endless run_commands?
- # If we actually run express server, the process will never end.
- # Can we somehow start server and make a request so we check if it works? This is tests, right?
+
+ def test_javascript_server_code(self, code: str, test_cmd: str, port: float):
+ self.write_file(self.default_javascript_code_file, code)
+ return self.test_server_code(server_cmd=f"node {self.default_javascript_code_file}", test_cmd=test_cmd, port=port, rootdir=self.rootdir)
diff --git a/api-service/codegen/tools/playground/tools/code.py b/api-service/codegen/tools/playground/tools/code.py
index 8ed37b903..0addeed8c 100644
--- a/api-service/codegen/tools/playground/tools/code.py
+++ b/api-service/codegen/tools/playground/tools/code.py
@@ -61,6 +61,28 @@ def create_code_tools(playground: NodeJSPlayground):
yield run_javascript_code
+ @tool("TestJavaScriptServerCode")
+ def test_server_code(input: str) -> str:
+ """
+ Execute JavaScript code that should start a server and then executes a testing command that should test if the server correctly handles needed requests.
+ The server should run on http://localhost:3000 and the request should be made on http://localhost:3000 too.
+ The testing command should be a `curl` command.
+ The input to this tool should be the testing command followed by a newline then the server code wrapped in three backtics.
+ The returned result is a the testing command output followed by the server code output and errors.
+ """
+
+ test_cmd, server_code = input.split("\n", 1)
+ code = extract_code(server_code)
+
+ port = 3000
+ test_output, server_output = playground.test_javascript_server_code(code=code, test_cmd=test_cmd, port=port)
+
+ test_result = encode_command_output(test_output)
+ server_result = encode_command_output(server_output)
+
+ return f"Test command result:\n{test_result}\n\nServer result:\n{server_result}"
+
+
+ # yield test_server_code
+
# @tool("JavaScriptREPL/Shell")
- # @tool("ServerRequestTest")
- # @tool("ServerRequestTest")
diff --git a/api-service/codegen/tools/playground/tools/process.py b/api-service/codegen/tools/playground/tools/process.py
index a7e1824cd..1f96a4eef 100644
--- a/api-service/codegen/tools/playground/tools/process.py
+++ b/api-service/codegen/tools/playground/tools/process.py
@@ -1,10 +1,10 @@
from langchain.agents import tool
from codegen.tools.playground.playground import NodeJSPlayground
-from playground_client.models.run_process_response import RunProcessResponse
+from playground_client.models.process_response import ProcessResponse
-def encode_command_output(response: RunProcessResponse, only_errors: bool = False) -> str:
+def encode_command_output(response: ProcessResponse, only_errors: bool = False) -> str:
result = ""
err_lines = [err.line for err in response.stderr]