Skip to main content

​
Overview

CreditNexus agent workflows have access to a comprehensive set of tools for data fetching, analysis, and research. These tools enable agents to perform complex quantitative analysis, web research, and business intelligence tasks. Code Reference: app/agents/langalpha_tools.py, app/services/web_search_service.py

​
LangAlpha Tools

​
Market Data Tools

​
Get Market Data

Fetch real-time and historical market data from Polygon.io. Tool: get_market_data Parameters:
  • ticker: Stock ticker symbol (e.g., “AAPL”)
  • start_date: Start date for historical data
  • end_date: End date for historical data
  • timespan: Time span (minute, hour, day, week, month, quarter, year)
Returns:
  • Market data (OHLCV)
  • Volume data
  • Price movements
  • Technical indicators
Configuration Required:
  • POLYGON_API_KEY: Polygon.io API key
Code Reference: app/agents/langalpha_tools.py (get_market_data function)

​
Get Ticker Snapshot

Get current snapshot of ticker data. Tool: get_ticker_snapshot Parameters:
  • ticker: Stock ticker symbol
Returns:
  • Current price
  • Volume
  • Market cap
  • Recent price movements
Code Reference: app/agents/langalpha_tools.py (get_ticker_snapshot function)

​
Fundamental Data Tools

​
Get Fundamental Data

Fetch company fundamental data from Alpha Vantage. Tool: get_fundamental_data Parameters:
  • ticker: Stock ticker symbol
  • data_type: Type of data (income_statement, balance_sheet, cash_flow, earnings)
Returns:
  • Financial statements
  • Earnings data
  • Company fundamentals
  • Financial ratios
Configuration Required:
  • ALPHA_VANTAGE_API_KEY: Alpha Vantage API key
Code Reference: app/agents/langalpha_tools.py (get_fundamental_data function)

​
Trading Signals Tools

​
Get Trading Signals

Generate trading signals and technical indicators. Tool: get_trading_signals Parameters:
  • ticker: Stock ticker symbol
  • timeframe: Timeframe for analysis
  • indicators: List of indicators to calculate
Returns:
  • Trading signals (buy, sell, hold)
  • Technical indicators (RSI, MACD, moving averages)
  • Signal strength
  • Confidence scores
Code Reference: app/agents/langalpha_tools.py (get_trading_signals function)

​
Web Research Tools

Perform web search using Serper API or WebSearchService. Tool: web_search Parameters:
  • query: Search query
  • num_results: Number of results to return
  • search_type: Type of search (web, news, images)
Returns:
  • Search results with titles, snippets, URLs
  • Relevance scores
  • Source metadata
Configuration:
  • SERPER_API_KEY: Serper API key (optional, uses WebSearchService fallback)
Code Reference: app/agents/langalpha_tools.py (web_search function), app/services/web_search_service.py

​
Get Tickertick News

Fetch financial news from Tickertick. Tool: get_tickertick_news Parameters:
  • ticker: Stock ticker symbol
  • limit: Number of news items to return
Returns:
  • Financial news articles
  • News headlines
  • Publication dates
  • Source information
Configuration:
  • TICKERTICK_API_KEY: Tickertick API key (optional, service may not be publicly available)
Code Reference: app/agents/langalpha_tools.py (get_tickertick_news function)

​
Browser Automation Tools

​
Browser Tool

Perform web browsing and content extraction using Playwright. Tool: browser_tool Parameters:
  • url: URL to browse
  • action: Action to perform (navigate, click, extract, screenshot)
  • selector: CSS selector for element interaction
Returns:
  • Page content
  • Extracted data
  • Screenshots (if requested)
  • Page metadata
Code Reference: app/agents/langalpha_tools.py (browser_tool function)

​
Python REPL Tool

​
Execute Python Code

Execute Python code for calculations and data analysis. Tool: python_repl_tool Parameters:
  • code: Python code to execute
  • variables: Input variables (optional)
Returns:
  • Execution result
  • Output values
  • Error messages (if any)
Capabilities:
  • Financial calculations
  • Data analysis with pandas/numpy
  • Statistical analysis
  • Custom algorithms
Security: Code execution is sandboxed for safety. Code Reference: app/agents/langalpha_tools.py (python_repl_tool function)

​
DeepResearch Tools

​
Web Search Service

DeepResearch uses the WebSearchService for comprehensive web research: Tool: WebSearchService.search Parameters:
  • query: Search query
  • num_results: Number of results (default: 10)
  • search_type: Type of search (web, news)
Returns:
  • Search results with titles, snippets, URLs
  • Relevance scores
  • Source metadata
Configuration:
  • SERPER_API_KEY: Serper API key (optional, uses fallback search)
Code Reference: app/services/web_search_service.py

​
Extract Content

Tool: WebSearchService.extract_content Parameters:
  • url: URL to extract content from
Returns:
  • Extracted text content
  • Metadata (title, author, date)
  • Structured data
Uses: trafilatura for content extraction Code Reference: app/services/web_search_service.py

​
Rerank Results

Tool: WebSearchService.rerank Parameters:
  • query: Original search query
  • results: Search results to rerank
Returns:
  • Reranked results with improved relevance
  • Relevance scores
Configuration:
  • RERANKING_USE_LOCAL: Use local reranking model (default: false)
  • RERANKING_MODEL: Reranking model (default: “BAAI/bge-reranker-base”)
  • RERANKING_API_URL: Remote reranking API URL (optional)
Code Reference: app/services/web_search_service.py

​
PeopleHub Tools

​
LinkedIn Integration

Tool: LinkedIn profile fetching and analysis Capabilities:
  • Profile data extraction
  • Professional history analysis
  • Skills and endorsements
  • Education and certifications
Code Reference: app/workflows/peoplehub_research_graph.py

​
Web Research

PeopleHub uses the same WebSearchService as DeepResearch for:
  • Web search
  • Content extraction
  • News analysis
  • Reputation research
Code Reference: app/services/web_search_service.py

​
Psychometric Analysis

Tool: Psychometric profiling Capabilities:
  • Big Five personality traits
  • Risk tolerance assessment
  • Decision-making style analysis
  • Buying and savings behavior
Code Reference: app/workflows/peoplehub_research_graph.py

​
Tool Configuration

​
Required API Keys

LangAlpha:
  • POLYGON_API_KEY: Polygon.io API key (required for market data)
  • ALPHA_VANTAGE_API_KEY: Alpha Vantage API key (required for fundamentals)
  • SERPER_API_KEY: Serper API key (optional, for web search)
  • TICKERTICK_API_KEY: Tickertick API key (optional, for financial news)
DeepResearch:
  • SERPER_API_KEY: Serper API key (optional, uses WebSearchService fallback)
PeopleHub:
  • No additional API keys required (uses existing WebSearchService and LLM)

​
Rate Limiting

Web Search:
  • Default: 360 requests/hour
  • Configurable via WEB_SEARCH_RATE_LIMIT
Market Data:
  • Respects Polygon.io rate limits
  • Configurable via API provider settings
Fundamental Data:
  • Respects Alpha Vantage rate limits
  • Free tier: 5 requests/minute, 500 requests/day

​
Caching

Web Search:
  • Caching enabled by default
  • TTL: 24 hours (configurable via WEB_SEARCH_CACHE_TTL_HOURS)
Market Data:
  • Caching handled by API provider
  • Historical data cached locally
Fundamental Data:
  • Caching handled by API provider
  • Financial statements cached locally

​
Audit Logging

All tool usage is logged for audit compliance:

​
Logged Information

  • Tool Name: Name of tool used
  • Parameters: Input parameters (sanitized for security)
  • Results: Tool results (summary, not full data)
  • User ID: User who triggered the tool
  • Deal ID: Deal context (if applicable)
  • Analysis ID: Analysis context (if applicable)
  • Timestamp: When tool was used
  • Duration: Time taken to execute

​
Audit Context

Tool usage is linked to:
  • Agent workflows (LangAlpha, DeepResearch, PeopleHub)
  • Deals (if deal_id provided)
  • Documents (if document_id provided)
  • Users (via user_id)
Code Reference: app/utils/audit.py, app/agents/langalpha_tools.py (_log_tool_usage function)

​
Error Handling

​
Retry Logic

Tools implement retry logic for transient failures:
  • Market Data: 3 retries with exponential backoff
  • Web Search: 3 retries with exponential backoff
  • Fundamental Data: 3 retries with exponential backoff

​
Error Responses

Tools return structured error responses: 
{
  "status": "error",
  "tool": "get_market_data",
  "error": "API rate limit exceeded",
  "retry_after": 60
}

​
Fallback Mechanisms

  • Web Search: Falls back to WebSearchService if Serper API unavailable
  • News: Falls back to web_search tool if Tickertick unavailable
  • Market Data: Returns cached data if API unavailable

​
Performance Optimization

​
Parallel Execution

  • Market Data: Parallel requests for multiple tickers
  • Web Search: Parallel searches for multiple queries
  • Content Extraction: Parallel extraction for multiple URLs

​
Caching Strategy

  • Web Search: Cache results for 24 hours
  • Market Data: Cache historical data locally
  • Fundamental Data: Cache financial statements locally

​
Rate Limit Management

  • Automatic Throttling: Respects API rate limits
  • Queue Management: Queues requests when rate limits exceeded
  • Priority Handling: Prioritizes critical requests

​
Security Considerations

​
API Key Security

  • All API keys stored as SecretStr in Pydantic settings
  • Never logged or exposed in error messages
  • Rotated regularly

​
Code Execution Safety

  • Python REPL: Sandboxed execution environment
  • Input Validation: All inputs validated before execution
  • Output Sanitization: Outputs sanitized before returning

​
Data Privacy

  • PII Handling: Personal information handled according to GDPR
  • Data Retention: Tool results retained per data retention policy
  • Access Control: Tool access controlled via user permissions

​
Next Steps

Last Updated: 2025-01-14
Code Reference: app/agents/langalpha_tools.py, app/services/web_search_service.py