TG Vibe Check

While you're analyzing charts and tokenomics, the real alpha might be hiding in the community vibes. We built an AI to read the room so you don't have to.

Community Health = Hidden Alpha

Every whale knows the drill: find the gem before it moons. But most DD stops at fundamentals and technicals. The community psychology? That's where the real signal lives.

Think about it - a project can have solid tech and good tokenomics, but if the community is full of bots, cope, and rugpull anxiety, you're stepping into a minefield. Conversely, a strong, engaged community can carry a project through bear markets.

💡

Alpha Insight The most successful whale moves often come from understanding crowd psychology before the crowd understands itself.

What We Built

TG Vibe Check is our AI-powered community health diagnostic tool. Point it at any Telegram channel, and it gives you a comprehensive psychological breakdown of the community - quantified, not just vibes.

Instead of manually lurking for hours trying to gauge sentiment, you get mathematical scores across multiple dimensions: FUD levels, cope detection, bot probability, and genuine engagement metrics.

The Intelligence Layer

Our AI analyzes 11 different message types across three critical dimensions:

Sentiment & Psychology Metrics

FUD Coefficient: Measures fear, uncertainty, and doubt vs. confidence levels

Cope Level: Detects unrealistic optimism during negative events

Community Cohesion: Unity vs. internal conflict ratio

Engagement Quality Indicators

Moon Boy Density: Ratio of low-effort hype to genuine questions

Helpfulness Ratio: How well the community answers substantive questions

Signal-to-Noise Ratio: Balance between meaningful discussion and mindless hype

Red Flag Detection

Rugpull Anxiety Index: Specific fears about developer abandonment or project failure

Bot/Shill Probability: Detection of inauthentic promotional content and coordinated messaging

Price Desperation Score: Unhealthy obsession with short-term price movements over fundamentals

Under the Hood

The magic happens in our prompt engineering. Here's a sample of what guides the AI analysis:

AI Prompt Sample

You are an AI assistant specialized in analyzing cryptocurrency community sentiment 
and health from Telegram chat messages. Your task is to provide a comprehensive 
analysis of the community's psychological state, engagement quality, and potential 
red flags.

## Classification Categories

### Sentiment & Psychology (5 categories):
1. **FUD**: Fear, uncertainty, doubt about the project
2. **HODL**: Confidence, diamond hands, bullish sentiment  
3. **COPE**: Unrealistic optimism during negative events
4. **CONFLICT**: Internal community disputes, arguments
5. **SUPPORT**: Community helping/supporting each other

### Engagement Quality (4 categories):
6. **MOON_BOY**: Low-effort hype, price predictions, emojis
7. **GENUINE_QUESTION**: Substantive questions about project
8. **COMMUNITY_HELP**: Helpful responses to questions
9. **TECHNICAL_DISCOURSE**: Deep technical discussions

### Red Flags (3 categories):
10. **RUGPULL_ANXIETY**: Specific fears about developers leaving
11. **BOT_SHILL**: Obvious promotional content, repetitive messaging
12. **PRICE_DESPERATION**: Unhealthy obsession with short-term price

## Metrics to Calculate

1. **FUD Coefficient** = FUD_count / (FUD_count + HODL_count)
2. **Cope Level** = COPE_count / total_sentiment_messages
3. **Community Cohesion** = SUPPORT_count / (SUPPORT_count + CONFLICT_count)
4. **Moon Boy Density** = MOON_BOY_count / (MOON_BOY_count + GENUINE_QUESTION_count)
5. **Helpfulness Ratio** = COMMUNITY_HELP_count / GENUINE_QUESTION_count
6. **Signal-to-Noise Ratio** = TECHNICAL_DISCOURSE_count / MOON_BOY_count
7. **Rugpull Anxiety Index** = RUGPULL_ANXIETY_count / total_messages
8. **Bot/Shill Probability** = BOT_SHILL_count / total_messages
9. **Price Desperation Score** = PRICE_DESPERATION_count / total_messages

For each metric, provide supporting evidence from the actual messages.

⚠️

Prompt Engineering This is just a snippet. The full prompt is 180+ lines with detailed examples and edge cases. Precision matters when analyzing psychology.

Simple Stack, Big Results

We kept the architecture lean but effective. No over-engineering, just what works:

4hrs

Build Time

~400

Lines of Code

5

Core Files

100%

Vibe Accuracy

# Core Stack
- Python 3.12 (clean, async architecture)
- LiteLLM (Claude Sonnet 4 integration)
- RapidAPI (Telegram Channel data source)
- Streamlit (UI so clean even degens can use it)
- Zero databases (stateless analysis)

ℹ️

RapidAPI Integration We use the Telegram Channel API via RapidAPI to fetch messages. Clean, reliable, and handles rate limiting for us.

Live Analysis

Test drive the tool on curated channels. See exactly how we quantify community vibes.

Launch Demo →

Get the code for free on GitHub →

ℹ️

Demo Note Currently features curated channels for demonstration. Full version with custom channel analysis available for serious inquiries.