Implement Gradio app for MCP tool listing and calling; add usage guide and JSON payload example

.github/copilot-instructions.md

@@ -0,0 +1,3 @@
+- All terminal commands should be in Windows PowerShell syntax.
+- Always activate the virtual environment before running any Python scripts with `.\.venv\Scripts\Activate.ps1;`
+- When reading files, always read the first 1000 lines of the file.

.gitignore

@@ -0,0 +1,50 @@
+# Python .gitignore
+__pycache__/
+*.py[cod]
+*.pyo
+*.pyd
+*.pdb
+*.egg
+*.egg-info/
+dist/
+build/
+.eggs/
+.env
+.venv
+env/
+venv/
+ENV/
+*.env
+*.venv
+*.log
+*.sqlite3
+.DS_Store
+.idea/
+.vscode/
+*.db
+*.bak
+*.swp
+*.swo
+*.~
+*.tmp
+coverage/
+htmlcov/
+.tox/
+.nox/
+.pytest_cache/
+.mypy_cache/
+.pyre/
+.pytype/
+.cache/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# VS Code settings
+.vscode/
+
+# macOS
+.DS_Store
+
+# Windows
+Thumbs.db

app.py

@@ -1,7 +1,56 @@
+
+# Gradio app to list available MCP tools from a server URL
+# Use githubRepo tool to fetch documentation on mcp from https://github.com/modelcontextprotocol/python-sdk
+# Docs for using curl to access gradio app:  https://github.com/gradio-app/gradio/blob/main/guides/09_gradio-clients-and-lite/03_querying-gradio-apps-with-curl.md
+
 import gradio as gr
+import asyncio
+import traceback
+import json
+from mcp import ClientSession
+from mcp.client.streamable_http import streamablehttp_client
+
+DEFAULT_MCP_URL = "https://mcp.composio.dev/composio/server/ca1477ae-dffb-404c-9021-ef25ff440221/mcp"
+DEFAULT_TOOL_NAME = "SALESFORCE_RETRIEVE_LEAD_BY_ID"
+DEFAULT_JSON_ARGS = '{"id": "00QgK0000005yjVUAQ"}'
+
+async def call_tool_from_server(url, tool_name, json_args):
+    try:
+        async with streamablehttp_client(url) as (read, write, _):
+            async with ClientSession(read, write) as session:
+                await session.initialize()
+                if not tool_name:
+                    # List tools if no tool name provided
+                    tools = await session.list_tools()
+                    return {"available_tools": [tool.name for tool in tools.tools]}
+                # Parse JSON args
+                try:
+                    args = json.loads(json_args) if json_args else {}
+                except Exception as e:
+                    return {"error": f"Invalid JSON arguments: {e}"}
+                # Call the tool
+                result = await session.call_tool(tool_name, args)
+                return result
+    except Exception as e:
+        tb = traceback.format_exc()
+        return {"error": str(e), "traceback": tb}
+
+def call_tool(url, tool_name, json_args):
+    return asyncio.run(call_tool_from_server(url, tool_name, json_args))
 
-def greet(name):
-    return "Hello " + name + "!!"
+def gradio_tool_lister_and_caller(url, tool_name, json_args):
+    """Gradio interface function for tool lister/caller."""
+    return call_tool(url, tool_name, json_args)
 
-demo = gr.Interface(fn=greet, inputs="text", outputs="text")
+demo = gr.Interface(
+    fn=gradio_tool_lister_and_caller,
+    inputs=[
+        gr.Textbox(label="MCP Server URL", value=DEFAULT_MCP_URL),
+        gr.Textbox(label="Tool Name", value=DEFAULT_TOOL_NAME),
+        gr.Textbox(label="Tool Arguments (JSON)", value=DEFAULT_JSON_ARGS, lines=4),
+    ],
+    outputs=gr.JSON(label="Result"),
+    title="MCP Tool Lister and Caller",
+    description="Enter an MCP server URL, tool name, and JSON arguments to call a tool. Leave tool name blank to list available tools.",
+)
 demo.launch()

gradio-api.md

@@ -0,0 +1,72 @@
+# Using Gradio API with curl.exe on Windows
+
+This guide explains how to interact with any Gradio app as an API using `curl.exe` from Windows PowerShell. It covers both public Hugging Face Spaces and local Gradio apps.
+
+## Prerequisites
+- Windows with PowerShell
+- `curl.exe` (included in Windows 10+)
+- The URL of your Gradio app (public or local, e.g., http://127.0.0.1:7860/)
+
+## Step 1: Prepare Your Input
+Create a file named `payload.json` in your working directory. For a local MCP Gradio app, the payload might look like:
+
+```json
+{
+  "data": [
+    "https://mcp.composio.dev/composio/server/ca1477ae-dffb-404c-9021-ef25ff440221/mcp",
+    "SALESFORCE_RETRIEVE_LEAD_BY_ID",
+    "{\"id\": \"00QgK0000005yjVUAQ\"}"
+  ]
+}
+```
+
+## Step 2: Make the POST Request (Local Gradio)
+Send your input to the local Gradio API endpoint using this command:
+
+```powershell
+curl.exe -X POST http://127.0.0.1:7860/gradio_api/call/predict -s -H "Content-Type: application/json" -d '@payload.json'
+```
+
+This returns a JSON response with an `event_id`, e.g.:
+
+```json
+{"event_id":"YOUR_EVENT_ID"}
+```
+
+## Step 3: Fetch the Result (Local Gradio)
+Use the `event_id` from the previous step to fetch the result:
+
+```powershell
+curl.exe -N http://127.0.0.1:7860/gradio_api/call/predict/YOUR_EVENT_ID
+```
+
+You will see output like:
+
+```
+event: complete
+data: [ ... ]
+```
+
+## Example for Public Gradio Apps
+For public Spaces, use the `/call/predict` endpoint and adjust the payload as needed:
+
+```powershell
+curl.exe -X POST "https://abidlabs-en2fr.hf.space/call/predict" -H "Content-Type: application/json" -d '@payload.json'
+```
+
+Then fetch the result as above, replacing the URL and event ID.
+
+## Notes
+- For Gradio apps with multiple inputs, adjust the `data` array in `payload.json` accordingly.
+- For private Spaces, add your Hugging Face token with `-H "Authorization: Bearer YOUR_TOKEN"` to both requests.
+- For file inputs, use a dictionary with a `path` key pointing to a file URL.
+- If authentication is required, first POST to `/login` and use the returned cookies in subsequent requests.
+
+## Troubleshooting
+- Always use single quotes around `@payload.json` in PowerShell to avoid variable expansion errors.
+- If you get JSON errors, check your `payload.json` formatting.
+- If you get a 404 or "Not Found" error, check the endpoint path (e.g., `/gradio_api/call/predict` for local apps).
+
+---
+
+This guide is based on the official Gradio documentation and adapted for both public and local Windows PowerShell use cases.

payload.json

@@ -0,0 +1,7 @@
+{
+    "data": [
+        "https://mcp.composio.dev/composio/server/ca1477ae-dffb-404c-9021-ef25ff440221/mcp",
+        "SALESFORCE_RETRIEVE_LEAD_BY_ID",
+        "{\"id\": \"00QgK0000005yjVUAQ\"}"
+    ]
+}

utputFormat

@@ -0,0 +1,7 @@
+HTTP/1.1 422 Unprocessable Content
+date: Sun, 10 Aug 2025 03:02:52 GMT
+server: uvicorn
+content-length: 158
+content-type: application/json
+
+{"detail":[{"type":"json_invalid","loc":["body",1],"msg":"JSON decode error","input":{},"ctx":{"error":"Expecting property name enclosed in double quotes"}}]}