Secretary

Secretary is an open-source automated social media analysis tool that monitors accounts on Truth Social and Twitter, fetches their latest posts, analyzes content using any LLM compatible with langchain, and delivers AI-generated reports via customizable messengers such as WeChat Work bots and Bark push notifications.

What is Secretary?

Secretary is an AI-powered social media monitoring and analysis tool created by GitHub user zkd8907 (Bobby). It takes a list of social media accounts (Truth Social and Twitter), fetches their latest posts, runs each post through a user-configured analysis prompt using any LLM provider that works with langchain, and sends the resulting analysis (optionally with translation) to configured messenger endpoints. The tool runs as a Python 3.11 script, deployable via Docker or directly with uv/venv.

Key Features

• Multi-platform support — Monitor accounts on Truth Social (token-based login) and Twitter (session-based or username/password login).
• Configurable AI analysis — Each monitored account gets its own prompt using the $content placeholder; compatible with any langchain-compatible LLM.
• Flexible message delivery — Configure messengers with custom HTTP request bodies (YAML or JSON string) supporting variables like $poster_name, $post_time, $content, $translation:zh-cn, $ai_result, $post_url.
• Scheduled execution — Built-in scheduler (scheduler.py) runs at configurable intervals (default 5 minutes) with optional run-on-start; Docker and env-var configurable.
• Translation — Uses googletrans to translate original content to target language (ISO 639-1 codes) before analysis or delivery.
• Redis integration — Stores last processed message IDs to avoid duplicate analysis.
• Open source MIT license — Full source on GitHub with 360 stars, 50 forks, 2 contributors.

Who is it for?

• Financial analysts — Monitor tweets from accounts like @realDonaldTrump, @myfxtrader, get AI-driven bullish/bearish summaries on US stocks, bonds, tech stocks, etc.
• Tech trend watchers — Set prompts to analyze posts from tech influencers for insights on semiconductors, AI, etc.
• News aggregators — Automatically fetch and translate posts from multiple accounts, then push digested analysis to WeChat groups via bots.

What can you do with Secretary?

• Monitor political figures — Configure Truth Social for realDonaldTrump with a financial analysis prompt; results delivered via WeCom robot or Bark app.
• Track multiple financial bloggers — Set up Twitter accounts (myfxtrader, HAOHONG_CFA) with shared prompts; receive unified bullish/bearish reports.
• Customize delivery channels — Use environment variables to define webhook URLs (e.g., $WECOM_TRUMP_ROBOT_URL) per account, with different message formats for each channel.

How does Secretary work?

1. Install dependencies via uv sync or pip install -e . (Python 3.11 required).
2. Configure environment variables in .env including Redis URL, LLM API credentials, social media login tokens/sessions.
3. Edit config/social-networks.yml to define monitored accounts: each entry has type (truthsocial/twitter), account ID(s), analysis prompt, and messenger configuration (HTTP method, headers, body with built-in variables).
4. Run python main.py for a single fetch, or python scheduler.py for scheduled runs.
5. The tool fetches newest posts, runs through LLM with the prompt, translates if configured, then POSTs the result to each messenger.

Pros and cons

• Pros: Fully open-source (MIT), supports multiple LLMs via langchain, configurable prompts per account, built-in scheduler, Docker support, no vendor lock-in.
• Cons: Requires Redis for deduplication; Twitter and Truth Social login tokens/sessions must be obtained manually; not accessible in mainland China due to platform blocks.

Pricing

Free and open-source under the MIT license.

FAQ

Is Secretary free?

Yes, Secretary is open-source under the MIT license. You only need to provide your own LLM API key and Redis instance.

What social media platforms does Secretary support?

Currently Truth Social and Twitter. Support is implemented via modular plugins in the modules directory.

Can I use Secretary without Redis?

No, Redis is required to store the last processed message ID and avoid re-processing the same post on each run.

What LLMs are supported?

Any LLM that is compatible with the langchain library. You configure model name, API key, and base URL in .env.

How do I get a Twitter session?

Use username/password login, which saves a .tw_session file. Alternatively, provide session content directly. Ensure no spaces in the token.