tools-only Metrics Dashboard Guide
The Local Deep Research metrics dashboard provides insights into your research activities, costs, and system performance.
Note: This documentation is maintained by the community and may contain inaccuracies. While we strive to keep it up-to-date, please verify critical information and report any errors via GitHub Issues.
Overview
The metrics dashboard tracks:
- Token usage and costs across LLM providers
- Search engine performance and health
- Research session statistics
- User satisfaction ratings
- Rate limiting status
Accessing Metrics
- Web Interface: Navigate to
/metricsin your browser - Direct URL:
http://localhost:5000/metrics
Dashboard Components
System Overview Cards
The main dashboard displays key metrics:
- Total Tokens Used: Shows total token consumption with expandable breakdown by model
- Total Researches: Number of research sessions conducted
- Average Response Time: System performance metric
- Success Rate: Percentage of successful operations
- User Satisfaction: Average star rating from user feedback
- Estimated Costs: Token-based cost estimation with provider breakdown
Time-based Filtering
Filter analytics by time period:
- Last 7 days
- Last 30 days
- Last 3 months
- Last year
- All time
Additional filters:
- Research mode (Quick Summary / Detailed / All)
Detailed Analytics Pages
Star Reviews Analytics
Access via: /metrics/star-reviews
Features:
- 5-star rating distribution
- Average ratings by time period
- Rating trends visualization
- Breakdown by model and search engine
- User feedback analysis
Cost Analytics
Access via: /metrics/costs
Tracked Metrics:
- Cost breakdown by provider (OpenAI, Anthropic, etc.)
- Token usage details (input/output/total)
- Cost trends over time
- Model-specific cost analysis
- Research type cost comparison
Rate Limiting Dashboard
Real-time monitoring of:
- Search engine rate limit status
- Success/failure rates per engine
- Wait time tracking
- Engine health indicators:
- 🟢 Healthy: >95% success rate
- 🟡 Degraded: 70-95% success rate
- 🔴 Poor: <70% success rate
Metrics Tracked
Token Metrics
- Total tokens (input + output)
- Token usage by model
- Average tokens per research
- Token consumption trends
Search Metrics
- Search engine usage frequency
- Response times per engine
- Success/failure rates
- Results count statistics
Research Metrics
- Total research sessions
- Research duration
- Completion status
- Strategy usage
- Query complexity
Performance Metrics
- API response times
- System latency
- Error rates
- Throughput statistics
Data Export
Research Reports
Export individual research results as:
- PDF: Formatted reports with citations
- Markdown: Raw markdown with formatting
- JSON: Structured data via API
Analytics Data
Access analytics data via API:
# Get overall metrics
curl http://localhost:5000/api/metrics
# Get specific research metrics
curl http://localhost:5000/api/metrics/research/<research_id>
# Get enhanced tracking data
curl http://localhost:5000/api/metrics/enhanced
# Get rating analytics
curl http://localhost:5000/api/star-reviews
# Get cost analytics
curl http://localhost:5000/api/cost-analytics
# Get rate limiting status
curl http://localhost:5000/api/rate-limiting
Visualizations
The dashboard uses Chart.js to display:
- Line Charts: Token usage and search activity over time
- Bar Charts: Model usage comparison, cost breakdown
- Pie Charts: Provider distribution, search engine usage
- Progress Indicators: Success rates, health status
Cost Tracking
Automatic Cost Calculation
Costs are calculated based on:
- Provider pricing (OpenAI, Anthropic, etc.)
- Actual token usage
- Model-specific rates
- Both input and output tokens
Supported Providers
- OpenAI (GPT-3.5, GPT-4)
- Anthropic (Claude models)
- Google (Gemini models)
- Local models (shown as $0)
Rate Limiting Analytics
Monitoring Features
- Real-time rate limit status
- Historical rate limit events
- Automatic wait time optimization
- Per-engine performance tracking
Managing Rate Limits
Via command line:
# Check status
python -m local_deep_research.web_search_engines.rate_limiting status
# Reset rate limits
python -m local_deep_research.web_search_engines.rate_limiting reset
Privacy and Data Storage
- All analytics data is stored locally
- No external analytics services used
- Data stored in SQLite databases
- Configurable data retention
Using Analytics for Optimization
Identify Cost Drivers
- Review high-token queries in cost analytics
- Compare model costs vs. quality ratings
- Optimize model selection for different tasks
Improve Search Performance
- Monitor search engine health status
- Identify frequently rate-limited engines
- Adjust search strategy based on success rates
Enhance Research Quality
- Analyze user ratings by research type
- Review low-rated sessions for patterns
- Adjust parameters based on feedback
Benchmarking Integration
For advanced users, analytics integrates with the benchmarking system:
- Track performance across different configurations
- Visualize optimization results (when matplotlib available)
- Compare quality vs. speed trade-offs
- Export benchmark data for analysis
API Reference
Metrics Endpoints
# General metrics with optional time filter
curl 'http://localhost:5000/api/metrics?days=30&mode=quick'
# Research-specific metrics
curl http://localhost:5000/api/metrics/research/<id>
# Enhanced metrics (detailed tracking)
curl http://localhost:5000/api/metrics/enhanced
# Star ratings data
curl 'http://localhost:5000/api/star-reviews?days=30'
# Cost breakdown
curl 'http://localhost:5000/api/cost-analytics?provider=openai'
# Rate limit status
curl http://localhost:5000/api/rate-limiting
Related Documentation
- Features Documentation
- Configuration Guide
- API Documentation
- Benchmarking Guide