Platform Market Analysis System

Platform Market Analysis Skill

Overview

The Market Analysis System is an AI-powered chat interface that allows users to ask questions about crypto markets, trends, and technical indicators. It uses a "ReAct" style agent approach to reason, fetch data, and formulate answers.

Key Features:

• Natural Language Parsing: Understands "Analyze BTC" or "Is ETH bullish?".
• Real-time Data: Fetches live prices, technical indicators (RSI, MAX), and news.
• Streaming Response: Shows the "thought process" (steps) in real-time before the final answer.
• Session Management: Multi-session chat history with pinning and persistence.

Key Files:

• Frontend: web/js/chat.js (Chat UI & Stream Handler), web/js/market.js (Data Visualization)
• Backend API: api/routers/analysis.py (Endpoints)
• Core Logic: interfaces/chat_interface.py (Bot & Parser), core/agents.py (Agent Logic)

Architecture

Data Flow

1. User Input: Typed in web/js/chat.js.
2. API Request: POST /api/analyze (Stream).
3. Query Parsing: CryptoQueryParser calls LLM to extract intent (Symbol: BTC, Action: Analyze).
4. Execution: CryptoAnalysisBot decides which tools to run (Price, News, Indicators).
5. Streaming:
   • Backend yields "Process Steps" ([PROCESS] Fetching data...)
   • Backend yields "Final Answer" ([RESULT] Bitcoin is currently...)
6. Rendering: Frontend parses the stream and updates the UI incrementally.

Components

1. CryptoAnalysisBot (interfaces/chat_interface.py)

the main orchestrator. It:

• Maintains conversation history.
• Manages the CryptoAgent or executes the "Standard Analysis Mode" (legacy).
• Generates the streaming response generator.

2. Streaming Protocol (NDJSON style)

The /api/analyze endpoint returns a stream of JSON objects or specially formatted text lines.

• data: {"content": "[PROCESS] Checking limits..."} -> Updates the "Thinking" UI.
• data: {"content": "[RESULT] **Analysis**: ..."} -> Renders the final markdown response.
• data: {"done": true} -> Closes the stream.

3. Frontend Chat (web/js/chat.js)

• Session Management: Loads/Creates/Deletes sessions via /api/chat/sessions.
• Stream Reader: Uses response.body.getReader() to process chunks.
• UI Rendering:
  • renderStoredBotMessage() parses the special tags ([PROCESS_START], [RESULT]) to create the collapsible "Process" section and the Markdown result.

API Endpoints

Analysis

POST /api/analyze

Streamed analysis response. Payload:

{
  "message": "Analyze BTC",
  "manual_selection": ["rsi", "macd"],
  "market_type": "spot",
  "session_id": "uuid..."
}

POST /api/analysis/backtest

Runs a quick backtest for a strategy.

Session Management

GET /api/chat/sessions

List user's chat sessions.

POST /api/chat/sessions

Create a new session.

GET /api/chat/history?session_id=...

Get message history for a session.

Frontend Integration (chat.js)

Key Function: sendMessage()

1. Checks for API Key.
2. Creates/Selects a session.
3. Sends POST /api/analyze.
4. Reads stream loop:

   while (true) {
       const { value, done } = await reader.read();
       // Decode and update UI
       botMsgDiv.innerHTML = renderStoredBotMessage(fullContent, true);
   }
   

UI States:

• Thinking: Shows a spinner and elapsed time.
• Process: Collapsible details tag showing step-by-step actions.
• Result: Final markdown content.

Modification Guidelines

✅ Safe Modifications

1. Adding New Tools:

   • Add tool function in core/tools.py.
   • Register tool in core/agents.py.
   • The Bot will automatically have access to it (if using Agent mode).

2. Customizing UI:

   • web/js/chat.js: renderStoredBotMessage controls how the message looks. You can change styling here.

⚠️ Risks

1. LLM Context Window:

   • Passing too much history or too large data (news articles) can hit token limits.
   • Fix: Use trim_history or summarization techniques in CryptoAnalysisBot.

2. Timeout:

   • Complex analysis takes time (>30s).
   • Fix: Ensure Nginx/frontend timeout settings allow for long-lived streams.

Related Skills

• platform-db-pattern: For session storage logic.
• pi-auth: For user identification in sessions.

Converted and distributed by TomeVault — claim your Tome and manage your conversions.