Skip to content

prompt.md

Latest commit

 

History

History
654 lines (493 loc) · 12.2 KB

prompt.md

File metadata and controls

654 lines (493 loc) · 12.2 KB

You are a senior backend and infrastructure engineer.

Build a production-grade V2Ray/VLESS/Trojan/ShadowSocks subscription aggregator platform using Python.

The goal is NOT to build a giant config dump repository. The goal is to build a CURATED, lightweight, high-quality subscription generator.

The system must:

  • Collect configs from public Telegram channels
  • Collect configs from GitHub repositories and raw URLs
  • Parse and normalize configs
  • Deduplicate intelligently
  • Score and rank configs
  • Generate categorized lightweight subscriptions
  • Publish subscriptions automatically every 10 minutes

==================================================== PROJECT GOALS

Main goals:

  1. Lightweight outputs Avoid massive subscription files. Output only high-quality configs.

  2. Curated subscriptions Publish only the best configs.

  3. Multiple categorized outputs Example:

    • mix
    • cloudflare-only
    • reality-only
    • mobile-friendly
    • fast
    • clean

  4. Stable architecture Async, modular, scalable.

  5. Easy deployment Support:

    • GitHub Actions
    • VPS cron jobs
    • Docker

==================================================== TECH STACK

Use:

  • Python 3.11+
  • asyncio
  • httpx
  • telethon
  • pydantic
  • orjson
  • uvloop (optional)
  • PyYAML

Do NOT use:

  • synchronous requests
  • monolithic code
  • unnecessary databases

==================================================== ARCHITECTURE

Design modular architecture:

project/ │ ├── app/ │ ├── collectors/ │ │ ├── telegram.py │ │ ├── github.py │ │ └── subscriptions.py │ │ │ ├── extractors/ │ │ └── links.py │ │ │ ├── parsers/ │ │ ├── vmess.py │ │ ├── vless.py │ │ ├── trojan.py │ │ └── shadowsocks.py │ │ │ ├── normalizers/ │ │ └── normalize.py │ │ │ ├── scoring/ │ │ └── engine.py │ │ │ ├── filters/ │ │ ├── cloudflare.py │ │ ├── reality.py │ │ ├── mobile.py │ │ ├── fast.py │ │ └── clean.py │ │ │ ├── health/ │ │ └── checker.py │ │ │ ├── outputs/ │ │ ├── generator.py │ │ ├── base64.py │ │ └── clash.py │ │ │ ├── utils/ │ │ ├── logger.py │ │ ├── retry.py │ │ └── cache.py │ │ │ └── models/ │ └── config.py │ ├── config/ │ ├── telegram_channels.txt │ ├── github_sources.json │ ├── subscription_sources.txt │ └── settings.yaml │ ├── generated/ │ ├── mix.txt │ ├── cloudflare.txt │ ├── reality.txt │ ├── mobile.txt │ ├── fast.txt │ ├── clean.txt │ └── base64/ │ ├── cache/ ├── tests/ ├── requirements.txt ├── Dockerfile ├── docker-compose.yml ├── README.md ├── .env.example └── main.py

==================================================== CONFIG TYPES

Support parsing:

  • vmess://
  • vless://
  • trojan://
  • ss://

==================================================== TELEGRAM COLLECTION

Use Telethon.

Requirements:

  • Read public Telegram channels
  • Async message fetching
  • Parse recent messages
  • Extract config links from messages
  • Handle Telegram rate limits
  • Use retry logic
  • Use environment variables:
    • TG_API_ID
    • TG_API_HASH
    • TG_SESSION

Create:

config/telegram_channels.txt

Example:

@freev2ray @proxy_channel @config_collector

==================================================== GITHUB COLLECTION

Collect from:

  • raw GitHub URLs
  • GitHub repositories
  • subscription text files

Create:

config/github_sources.json

Format:

[ { "name": "source1", "url": "https://raw.githubusercontent.com/...", "type": "raw" } ]

Requirements:

  • async downloads
  • timeout handling
  • retry handling
  • cache responses

==================================================== SUBSCRIPTION URL COLLECTION

Support collecting from existing subscription URLs.

Create:

config/subscription_sources.txt

Requirements:

  • auto detect base64 subscriptions
  • decode automatically
  • extract configs

==================================================== EXTRACTION

Extract links using regex.

Supported patterns:

  • vmess://
  • vless://
  • trojan://
  • ss://

Handle malformed inputs safely.

==================================================== NORMALIZATION

Normalize all configs.

Requirements:

  • trim whitespace
  • fix malformed URLs
  • decode vmess JSON
  • validate required fields
  • standardize transport names
  • normalize security values
  • normalize hostnames

==================================================== PYDANTIC MODEL

Create a unified config model.

Fields:

  • protocol
  • host
  • port
  • uuid
  • password
  • security
  • transport
  • tls
  • sni
  • path
  • network
  • fp
  • reality
  • source
  • raw
  • score

==================================================== DEDUPLICATION

Do NOT deduplicate using raw string compare.

Deduplicate using:

  • protocol
  • host
  • port
  • uuid/password
  • transport
  • security

Generate stable hash.

==================================================== SCORING ENGINE

Implement scoring system.

Example scoring:

+25 => Reality +20 => Cloudflare CDN +15 => TLS enabled +10 => clean hostname +10 => ws transport +8 => grpc transport +5 => IPv4 valid +5 => low latency -10 => suspicious port -10 => duplicate hostname spam -20 => malformed params -100 => failed health check

Configs with higher score should be prioritized.

==================================================== CLOUDFLARE DETECTION

Detect Cloudflare-based configs.

Methods:

  • known Cloudflare IP ranges
  • Cloudflare domains
  • CDN indicators
  • ws + tls patterns

==================================================== REALITY DETECTION

Detect:

  • reality
  • xtls
  • vision

==================================================== HEALTH CHECKS

Implement async health checks.

Requirements:

  • async TCP connect
  • configurable timeout
  • optional TLS handshake
  • lightweight checks only
  • avoid long blocking operations

Health checks must NOT slow down the whole pipeline.

==================================================== FILTERS

Implement categorized filters.

====================================================

  1. MIX FILTER ====================================================

Best overall configs.

Limit:

  • maximum 300 configs

==================================================== 2. CLOUDFLARE FILTER

Only:

  • Cloudflare CDN
  • ws/grpc
  • tls

Limit:

  • maximum 150 configs

==================================================== 3. REALITY FILTER

Only:

  • reality
  • xtls
  • vision

Limit:

  • maximum 100 configs

==================================================== 4. MOBILE FILTER

Prioritize:

  • lightweight configs
  • ws
  • tls
  • cloudflare
  • stable latency

==================================================== 5. FAST FILTER

Sort by:

  • lowest latency
  • successful checks

==================================================== 6. CLEAN FILTER

Only configs with:

  • readable hostnames
  • valid params
  • stable checks
  • no garbage fields

==================================================== OUTPUT GENERATION

Generate:

  • raw txt subscriptions
  • base64 subscriptions
  • optional Clash YAML

Output directory:

generated/

Examples:

  • generated/mix.txt
  • generated/cloudflare.txt
  • generated/reality.txt
  • generated/mobile.txt
  • generated/fast.txt
  • generated/clean.txt

Also generate:

  • generated/base64/mix.txt
  • generated/base64/cloudflare.txt

==================================================== BASE64 OUTPUTS

Encode subscriptions correctly.

Example:

base64.b64encode(content.encode()).decode()

==================================================== CLASH GENERATION

Optional.

Generate Clash-compatible YAML.

Requirements:

  • clean formatting
  • grouped proxies
  • proxy-groups
  • provider support

==================================================== ASYNC REQUIREMENTS

Everything must be async.

Use:

  • asyncio.gather
  • async context managers
  • async HTTP clients

Avoid blocking calls.

==================================================== LOGGING

Implement structured logging.

Log:

  • collection stats
  • parsing failures
  • dead configs
  • output counts
  • execution duration

==================================================== RETRY STRATEGY

Implement retry handling.

Requirements:

  • exponential backoff
  • configurable retries
  • safe failure handling

==================================================== CACHE

Implement lightweight cache.

Use:

  • JSON files
  • memory cache

Avoid databases initially.

==================================================== CONFIGURATION

Create settings.yaml.

Include:

  • output limits
  • timeout values
  • health check settings
  • source toggles
  • logging settings
  • retry settings

==================================================== DOCKER SUPPORT

Create:

  • Dockerfile
  • docker-compose.yml

Requirements:

  • lightweight image
  • production-ready
  • environment variables support

==================================================== GITHUB ACTIONS

Create GitHub Actions workflow.

Requirements:

  • run every 10 minutes
  • install dependencies
  • execute pipeline
  • commit generated files
  • push automatically

Workflow file:

.github/workflows/update.yml

==================================================== README

Create professional README.md.

Include:

  • project overview
  • architecture
  • setup instructions
  • Telegram setup
  • GitHub Actions setup
  • Docker setup
  • environment variables
  • deployment guide
  • subscription examples

==================================================== CODE QUALITY

Requirements:

  • typed Python
  • clean architecture
  • modular code
  • docstrings
  • maintainable structure
  • separation of concerns

==================================================== IMPORTANT REQUIREMENTS

DO NOT:

  • publish giant garbage dumps
  • keep dead configs
  • create monolithic files
  • use blocking requests
  • overcomplicate the system

PRIORITIZE:

  • lightweight outputs
  • clean architecture
  • curated configs
  • maintainability
  • scalability
  • async performance
  • code readability

==================================================== FINAL GOAL

Build a high-quality curated subscription platform that:

  • feels lightweight
  • has clean subscriptions
  • updates automatically
  • has categorized outputs
  • is scalable
  • is production-ready
  • is easy to maintain
  • provides significantly better quality than giant public dump repositories