I want to create a script that automatically fetch...

Below is a high-level, step-by-step blueprint you can hand to an AI agent (or follow yourself) to build the requested “daily best-node picker” around Telethon + v2rayN. Each stage tells the agent what to do, why it matters, and what libraries / APIs to use.

---

0 | Prerequisites & environment

Item	Purpose
Python ≥ 3.10	Core scripting language
Telethon	Telegram client SDK
ping3 (or asyncio + raw sockets)	ICMP / TCP ping tests
speedtest-cli (optional but useful)	Quick throughput check through the proxy
pywin32 or python-mss / pyautogui	UI automation fallback for v2rayN
v2rayN (portable ZIP)	User-side GUI client with bundled xray-core
git & VS Code	Dev workflow (optional)

Tip – Keep the script and v2rayN in the same top-level folder so the script can find v2rayN/core/xray.exe without hunting for paths.

---

1 | Log in to Telegram via Telethon

1. Obtain api_id and api_hash from https://my.telegram.org.
2. Start a Telethon TelegramClient('session_name', api_id, api_hash).
3. If the .session file doesn’t exist, prompt the user once for the telephone number and OTP; Telethon stores the session for future runs.
   docs.telethon.dev

---

2 | Fetch & parse fresh configs from the three channels

1. Store the target channel usernames or IDs in a list channels = ['@chanA', '@chanB', '@chanC'].

2. For each channel run:

   python
   async for msg in client.iter_messages(channel, limit=200):
       for proto in ('vmess://', 'vless://', 'trojan://', 'ss://'):
           if proto in msg.raw_text:
               extract each line that starts with proto
               add to raw_config_strings

3. Decode each URL:

   • For vmess:// links the part after vmess:// is Base64-encoded JSON.
   • For vless://, trojan://, etc., parse as standard URI (urllib.parse).

4. Build a Python dict list with at minimum:

   python
   {
       'host': 'example.com',
       'port': 443,
       'type': 'vmess',
       'uuid': '…',
       'raw_url': 'original link',
   }

---

3 | Latency & speed benchmarking

1. Ping test

   python
   from ping3 import ping
   rtt = ping(host, unit='ms', timeout=1)

   If ICMP is blocked, open a TCP socket to (host, port) and time the handshake (asyncio.open_connection).

2. Throughput test (optional / slow)
   Set environment variables ALL_PROXY=socks5://127.0.0.1:port after spinning up each node (see next step) and call speedtest-cli --simple.

3. Record results as {'rtt_ms': 123, 'dl_Mbps': 42, …}; keep failed pings as None so they can be filtered out.

---

4 | Choose the best node

python
viable = [n for n in nodes if n['rtt_ms'] is not None]
best = sorted(viable, key=lambda n: (n['rtt_ms'], -n.get('dl_Mbps', 0)))[0]

Break ties by download speed or simply lowest latency.

---

5 | Feed the winner into v2rayN

Problem: v2rayN itself offers no documented CLI switcher — confirmed by the project maintainers. github.com
Work-around: use the bundled xray-core directly or automate the GUI. Two clean options are below; pick whichever your environment tolerates:

Option	How it works	Effort
A. Drive xray.exe directly (recommended)	1. Dump the chosen node into a tiny config.json template.<br>2. Spawn v2rayN\\core\\xray.exe -config config.json as a detached process.<br>3. (Optional) use the gRPC API in xray to tear down / replace the node without killing the process. v2fly.org	Simple, fully headless
B. Automate v2rayN’s GUI	1. Start v2rayN if not running.<br>2. Put the raw URL into the clipboard (pyperclip).<br>3. Use pywin32 or pyautogui to send:<br> • Tray-icon right-click → Import from clipboard<br> • Right-click on the imported node → Set as active server.<br>4. Toggle System proxy → Set system proxy.	Works but brittle; UI changes break it

---

6 | Verify connectivity

1. Sleep 3–5 s.
2. requests.get('https://1.1.1.1/cdn-cgi/trace', timeout=5) through the proxy.
   Verify the public IP is not your ISP’s.

---

7 | Schedule & loop

• Windows Task Scheduler: create a daily task at, e.g., 06:00 with python fetch_and_connect.py.
• Inside the script wrap everything in asyncio.run(main()) and add proper try/except logging so the task never hangs.

---

8 | Hardening & polish (nice-to-have)

Concern	Mitigation
Channel message spike	Cache message IDs and stop once you hit an already-seen one.
Rate limits	One Telegram request ≈ 200 msgs; keep under 30/sec to avoid flood-waits.
Multiple configs per post	Regex can grab them all; they’re separated by line breaks.
JSON template	Keep a template.json with inbound (socks 10808) and routing pre-filled; only swap the outbound section.
Proxy for the speed test	Re-export ALL_PROXY before running speedtest-cli.
Log rotation	Write logs to %TEMP%/v2ray_picker.log, rotate weekly.

---

Deliverables for the AI

1. fetch_and_connect.py – the full script encapsulating steps 1-7.
2. template.json – minimal xray/v2rayN JSON with a placeholder outbound.
3. README.md – one-pager instructions + screenshots if you go the GUI-automation route.

Hand these specs to the agent and let it implement; every part is deterministic, so the AI can write & test the code autonomously.