Cryptocurrency Consultant Agents

This document covers the specialized AI agents that provide cryptocurrency trading consultation services within the agent swarm system. These agents are cryptocurrency-specific consultants that analyze market data, provide trading recommendations, and offer investment strategies for Bitcoin, Ethereum, Binance Coin, Solana, and Ripple.

For information about the overall agent architecture and setup, see Agent Architecture and Setup. For details about agent routing and navigation between agents, see Agent Navigation and Routing.

Agent Overview

The system implements five specialized cryptocurrency consultant agents, each dedicated to providing expert analysis and trading advice for a specific digital asset. These agents integrate real-time market data, technical indicators, and social sentiment analysis to deliver comprehensive investment recommendations.

Consultant Agent Architecture

Agent Implementation Details

Agent Registration and Configuration

All cryptocurrency consultant agents follow a consistent implementation pattern using the addAgent function from agent-swarm-kit. Each agent is configured with specific parameters for their cryptocurrency specialization.

Agent	AgentName Enum	Cryptocurrency	Symbol
Bitcoin Consultant	AgentName.BtcConsultantAgent	Bitcoin	BTC
Ethereum Consultant	AgentName.EthConsultantAgent	Ethereum	ETH
Binance Coin Consultant	AgentName.BnbConsultantAgent	Binance Coin	BNB
Solana Consultant	AgentName.SolConsultantAgent	Solana	SOL
Ripple Consultant	AgentName.XrpConsultantAgent	Ripple	XRP

Each agent uses CompletionName.GrokMiniStreamCompletion for AI processing and maintains conversation history with keepMessages: 10. Tool calls are unlimited with maxToolCalls: Number.POSITIVE_INFINITY.

Dynamic System Context

Each agent generates dynamic system context that includes real-time market information:

The systemDynamic function provides each agent with:

• Current date and time in DD/MM/YYYY HH:mm Z format
• Real-time cryptocurrency price from Binance API
• Balance report for web-based clients (not Telegram)

Trading Strategy Framework

Core Trading Parameters

All consultant agents implement a unified trading strategy based on the CC_LADDER_TRADE_REVENUE configuration parameter. The strategy focuses on trend reversal identification and risk management.

Trading Strategy Rules:

• Recommend purchases when price trends reverse from decline to growth
• Target growth of 1% over 56 hours with 8/10 confidence
• Maximum acceptable risk level: 4/10 (presented as "probability of success" 6/10)
• Risk-to-profit ratio requirement: minimum 1:1, preferably 1:2 or higher
• Stop-loss positioning to minimize immediate losses

Analysis Time Horizons

Each agent analyzes market conditions across three time perspectives:

Time Horizon	Duration	Technical Indicators	Data Sources
Short-term	8 hours	EMA, SMA	Price data, Mastodon posts
Medium-term	2-3 days	MACD, SMA	Price data, Mastodon posts
Long-term	1 week	MACD, RSI	Price data, Mastodon posts

Analysis Tools and Data Sources

Technical Analysis Tools

Each consultant agent has access to the same comprehensive set of analysis tools:

Technical Indicators:

• FetchShortRangeEmaSignals: Exponential Moving Average for short-term analysis
• FetchSwingRangeMacdSignals: MACD for swing trading signals
• FetchLongRangeRsiSignals: RSI for long-term momentum analysis
• FetchVolumeDataSmaSignals: Simple Moving Average of trading volume

Market Data:

• FetchPriceDataCandles: Historical price data (last 100 hourly candles)
• FetchMastodonNewsSignals: Social media sentiment from Mastodon posts (7-day window)

Navigation:

• NavigateToTriageTool: Seamless handoff to triage agent when topic changes

Agent Behavior and Communication Constraints

Response Format Limitations

All consultant agents operate under strict communication constraints to ensure compatibility with the chat interface:

Supported Formats:

• Text-only Markdown messages
• Hyperlinks
• Inline code formatting

Prohibited Formats:

• Files or file attachments
• Images or graphics
• Tables
• HTML markup
• Code blocks with triple backticks
• Line breaks within responses

Agent Navigation Logic

Each consultant agent monitors conversation context to determine when to redirect users to other agents:

Navigation Triggers:

• User expresses disinterest in the agent's specific cryptocurrency
• User mentions other cryptocurrencies (BTC, ETH, BNB, SOL, XRP)
• Trading discussions about non-specialized assets
• Purchase/sale requests for other cryptocurrencies

The navigation is implemented silently using NavigateToTriageTool without explicit user notification about the transfer.

Error Handling and Rate Limiting

Callback Implementation

Each consultant agent implements two critical callbacks for operational reliability:

onResurrect Callback:

• Triggered when rate limits are exceeded
• Sends user notification via Telegram API
• Only applies to Telegram-based clients (clientId starts with "telegram-")
• Message: "Привышено количество запросов. Пожалуйста, повторите попытку"

onToolError Callback:

• Captures tool execution failures
• Logs comprehensive error information to tool_error.txt
• Records: error message, error data, client ID, agent name, tool name

Agent Dependencies

All cryptocurrency consultant agents depend on the TriageAgent for proper routing and conversation management. This dependency ensures that users can be seamlessly transferred between specialized agents based on their queries and interests.

The dependency relationship is declared in each agent configuration using dependsOn: [AgentName.TriageAgent], establishing the hierarchical agent structure within the swarm system.