update the code and script.

Author: gali-leilei

README.md

@@ -186,9 +186,11 @@ cp .env.template .env
 vim .env
 ```
 
-**Step 3:** Prepare E2B Sandbox:
+**Step 3:** Prepare E2B Sandbox (Optional)
 
 > [!TIP]
+> We provide a public E2B sandbox template. Follow this step if you want to reproduce.
+>
 > For the E2B sandbox service, we recommend setting up a Linux Docker image with a comprehensive set of apt and Python packages pre-installed. Without these pre-installed packages, the agent will need to spend extra steps and context installing them, resulting in reduced token efficiency.
 >
 > you need to have `npm` install and `docker` running locally.

apps/run-agent/scripts/claude-sonnet-3.7/run_evaluate_multiple_runs_gaia-validation.sh

@@ -35,7 +35,7 @@ for i in $(seq 1 $NUM_RUNS); do
             benchmark.execution.max_concurrent=5 \
             benchmark.execution.pass_at_k=1 \
             agent=$AGENT_SET \
-            output_dir=${RESULTS_DIR}/$RUN_ID \
+            output_dir="$RESULTS_DIR/$RUN_ID" \
             > "$RESULTS_DIR/${RUN_ID}_output.log" 2>&1
 
         if [ $? -eq 0 ]; then

libs/miroflow-tool/src/miroflow/tool/mcp_servers/python_server.py

@@ -1,8 +1,8 @@
+import asyncio
 import os
 
 from e2b_code_interpreter import Sandbox
 from fastmcp import FastMCP
-import asyncio
 
 # Initialize FastMCP server
 mcp = FastMCP("e2b-python-interpreter")
@@ -11,6 +11,10 @@
 E2B_API_KEY = os.environ.get("E2B_API_KEY")
 LOGS_DIR = os.environ.get("LOGS_DIR")  # Directory where benchmark logs are stored
 
+# DEFAULT TEMPLATE ID
+# see README.md on how to build this
+DEFAULT_TEMPLATE_ID = "1av7fdjfvcparqo8efq6"
+
 # DEFAULT CONFS
 DEFAULT_TIMEOUT = 1200  # seconds
 
@@ -138,7 +142,9 @@ async def create_sandbox() -> str:
         sandbox = None
         try:
             sandbox = Sandbox(
-                template="all_pip_apt_pkg", timeout=DEFAULT_TIMEOUT, api_key=E2B_API_KEY
+                template=DEFAULT_TEMPLATE_ID,
+                timeout=DEFAULT_TIMEOUT,
+                api_key=E2B_API_KEY,
             )
             info = sandbox.get_info()