Upload README.md
Browse files
README.md
CHANGED
@@ -10,4 +10,23 @@ license: mit
|
|
10 |
short_description: uses playwright to scrape a fake app hosted on vercel
|
11 |
tag: "mcp-server-track"
|
12 |
---
|
13 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
short_description: uses playwright to scrape a fake app hosted on vercel
|
11 |
tag: "mcp-server-track"
|
12 |
---
|
13 |
+
## Fake app scraper
|
14 |
+
<div>
|
15 |
+
<p>
|
16 |
+
This app uses Playwright to log in and scrape the content of the dashboard of the fake app
|
17 |
+
<a href="https://fake-app-omega.vercel.app" target="_blank">fake-app-omega.vercel.app</a>.
|
18 |
+
</p>
|
19 |
+
<p>
|
20 |
+
The starting point was to provide an LLM-friendly, API-fied version of a real app that does not provide any API. Used as an MCP server, any user of the app could simply ask their AI assistant to fetch information from their dashboard.
|
21 |
+
</p>
|
22 |
+
<h3>Problem</h3>
|
23 |
+
<p>
|
24 |
+
When run locally, the app can take the credentials as environment variables to log into the user's dashboard. However, when hosted as a Hugging Face Space, I couldn't find a way to send the credentials securely—i.e., without explicitly providing them to the LLM.
|
25 |
+
</p>
|
26 |
+
<p>
|
27 |
+
In conclusion, as it stands, this app together with the fake Next.js app only serves demonstration or educational purposes and does not solve a real-life problem.
|
28 |
+
</p>
|
29 |
+
<p><strong>I’d be happy to get any suggestions on how to send credentials in the context of a Gradio HF Space–hosted app.</strong></p>
|
30 |
+
<img src="login.png" alt="login" style="max-width: 35%; margin-right: 10px;" />
|
31 |
+
<img src="dashboard.png" alt="dashboard" style="max-width: 35%;" />
|
32 |
+
</div>
|