Lobster Insight

Binance Square Sentiment Stripping & Contrarian Alpha Engine

Lobster Insight is a lightweight, open-source demo kit for a Binance Square-native AI agent. It is designed to turn noisy, hype-heavy social posts into a structured risk report:

• strip emotional language from facts
• detect overheated consensus
• trace minority high-quality viewpoints
• generate calm action guidance

This repository is intentionally lightweight and demo-friendly. It is not a trading bot and does not place orders.

What this repo includes

• a reusable system prompt for OpenClaw or other LLM agent shells
• sample Binance Square-style hype inputs
• a Python mock pipeline that converts post batches into a structured report
• a CLI demo runner
• example outputs you can use for screenshots or demos

Core modules

1. Noise Filtration
   Separates hype, slogans, and manipulative language from verifiable facts.

2. Contrarian Index
   Detects one-sided crowd behavior, euphoric language, and FOMO extremes.

3. Narrative Saturation Monitor
   Measures repetition, low novelty, and overcrowded consensus.

4. Smart Consensus Tracer
   Surfaces minority posts that contain mechanism analysis, data reasoning, or explicit risk discussion.

5. Actionable Insight Composer
   Produces calm, explainable guidance such as “avoid chasing”, “wait for confirmation”, or “keep size disciplined”.

Quick start

1) Use the Python demo

python -m src.demo --input examples/hype_wave_posts.json

2) Or paste the system prompt into OpenClaw

The reusable prompt lives here:

prompts/system_prompt.md

Recommended first demo input:

prompts/demo_input_1.md

Example output

The CLI prints a structured JSON report and a human-readable markdown report. A sample markdown report is included at:

examples/sample_report.md

Repository structure

Lobster-Insight/
├─ README.md
├─ LICENSE
├─ requirements.txt
├─ prompts/
│  ├─ system_prompt.md
│  ├─ demo_input_1.md
│  └─ demo_input_2.md
├─ examples/
│  ├─ hype_wave_posts.json
│  ├─ product_card.json
│  └─ sample_report.md
├─ lobster_insight/
│  ├─ __init__.py
│  ├─ models.py
│  ├─ heuristics.py
│  ├─ engine.py
│  └─ render.py
└─ src/
   └─ demo.py

Intended use

This project is built for:

• demo videos
• hackathon submissions
• prompt prototyping
• reproducible agent workflows
• future extensions into richer NLP pipelines

Not included

This repository does not include:

• exchange connectivity
• order execution
• portfolio management
• live scraping of Binance Square
• financial advice

Future extensions

• real post ingestion layer
• embedding-based viewpoint clustering
• stronger sentiment labeling
• event extraction from multilingual post streams
• score calibration from live community data

License

MIT