~95% on SimpleQA (e.g. Qwen3.6-27B on a 3090). Supports all local and cloud LLMs (llama.cpp, Ollama, Google, ...). 10+ search engines - arXiv, PubMed, your private documents. Everything Local & Encrypted.
Metrics & Analytics Dashboard
This guide covers the Metrics Dashboard for monitoring research performance, costs, and usage analytics.
Table of Contents
Overview
The Metrics Dashboard provides insights into your research activity:
- Token usage and LLM costs
- Search performance and rate limiting
- Research quality ratings
- Link analytics and domain classification
Access the dashboard at: http://localhost:5000/metrics
Research Metrics
Overview Page
The main metrics page shows:
| Metric | Description |
|---|---|
| Total Researches | Number of research sessions |
| Total Tokens | Tokens consumed across all research |
| Total Cost | Estimated LLM API costs |
| Average Duration | Mean research completion time |
Incomplete Metrics Warning
If token, search, or rate-limit updates cannot be recorded, the main metrics page shows a warning for the signed-in user. It counts final write failures after retries, missing database credentials, and search or token updates skipped to preserve pending caller work or when no matching owner session is available. Research can continue while recording is unavailable. The warning is a snapshot taken when the page is rendered: changing the time range or research mode reloads the metrics data but does not refresh the warning. Reload the full page to see the latest count and reason categories; the server logs provide the existing diagnostic details.
The count describes failed or skipped recording attempts, not a verified number of missing rows: an error during session cleanup can occur after a write.
These counters are held in memory by the current server process, separately for each user. They reset on restart and retain at most 512 recent writers; older entries can be evicted. Counts are therefore lower bounds, not a durable audit or a count restricted to the dashboard's selected date range. An absent warning does not prove historical metrics are complete. Calls without an identified user and intentional programmatic mode do not create a user's warning. The counter stores no passwords, queries, or exception messages.
Per-Research Metrics
Click on any research to see detailed metrics:
Timeline
View the research execution timeline:
- Start and end times
- Iteration progress
- Question generation timing
- Search execution timing
Search Metrics
Details about search operations:
- Queries executed
- Results retrieved
- Search engine usage
- Response times
Links
All links discovered during research:
- Source URLs
- Domain classification
- Relevance scores
Cost Analytics
Token Tracking
The system tracks token usage for cost estimation:
| Metric | Description |
|---|---|
| Input Tokens | Tokens sent to LLM |
| Output Tokens | Tokens received from LLM |
| Total Tokens | Combined usage |
| Estimated Cost | Based on model pricing |
Cost Breakdown
View costs broken down by:
- Model - Which LLM models were used
- Provider - OpenAI, Anthropic, etc.
- Time Period - Daily, weekly, monthly
- Research - Per-research costs
Accessing Cost Analytics
- Navigate to Metrics → Costs
- View summary statistics
- Filter by date range
- Export data as needed
Pricing Information
The system includes pricing data for common models:
GET /metrics/api/pricing
GET /metrics/api/pricing/<model_name>
Rate Limiting
Monitoring Rate Limits
View current rate limiting status for search engines:
- Navigate to Metrics → Rate Limiting
- See per-engine statistics
| Column | Description |
|---|---|
| Engine | Search engine name |
| Current Wait | Active wait time |
| Success Rate | Percentage of successful requests |
| Total Requests | Request count |
| Last Updated | Most recent activity |
Rate Limit Management
From the dashboard you can:
- View current rate limit estimates
- See learned optimal wait times
- Monitor success rates
Link Analytics
Domain Classification
The system classifies domains found in research:
| Category | Examples |
|---|---|
| Academic | arxiv.org, pubmed.gov |
| News | bbc.com, reuters.com |
| Government | .gov, .edu |
| Commercial | company websites |
| Reference | wikipedia.org |
Link Statistics
View aggregate statistics:
- Total links discovered
- Unique domains
- Category distribution
- Top domains by frequency
Accessing Link Analytics
- Navigate to Metrics → Links
- View domain breakdown
- Filter by category
- See classification summary
Classification API
Classify domains programmatically:
POST /metrics/api/domain-classifications/classify
{
"domains": ["example.com", "arxiv.org"]
}
Star Reviews
Rating Research Quality
Rate your research results to track quality over time:
- Open a completed research
- Click the star rating (1-5)
- Optionally add feedback text
Viewing Ratings
Navigate to Metrics → Star Reviews to see:
- Average rating across all research
- Rating distribution
- Recent ratings with feedback
- Trends over time
Rating API
# Get ratings for a research
GET /metrics/api/ratings/<research_id>
# Submit a rating
POST /metrics/api/ratings/<research_id>
{
"rating": 5,
"feedback": "Excellent results"
}
API Reference
Metrics Endpoints
| Endpoint | Method | Description |
|---|---|---|
/metrics/api/metrics |
GET | Overview metrics |
/metrics/api/metrics/enhanced |
GET | Detailed metrics |
/metrics/api/metrics/research/<id> |
GET | Per-research metrics |
/metrics/api/metrics/research/<id>/timeline |
GET | Research timeline |
/metrics/api/metrics/research/<id>/search |
GET | Search metrics |
/metrics/api/metrics/research/<id>/links |
GET | Research links |
Cost Endpoints
| Endpoint | Method | Description |
|---|---|---|
/metrics/api/pricing |
GET | Model pricing data |
/metrics/api/pricing/<model> |
GET | Specific model pricing |
/metrics/api/cost-calculation |
POST | Calculate costs |
/metrics/api/research-costs/<id> |
GET | Research cost breakdown |
/metrics/api/cost-analytics |
GET | Cost analytics summary |
Rate Limiting Endpoints
| Endpoint | Method | Description |
|---|---|---|
/metrics/api/rate-limiting |
GET | Rate limit statistics |
/metrics/api/rate-limiting/current |
GET | Current rate limits |
Rating Endpoints
| Endpoint | Method | Description |
|---|---|---|
/metrics/api/ratings/<id> |
GET | Get research rating |
/metrics/api/ratings/<id> |
POST | Submit rating |
/metrics/api/star-reviews |
GET | All ratings summary |
Domain Classification Endpoints
| Endpoint | Method | Description |
|---|---|---|
/metrics/api/domain-classifications |
GET | All classifications |
/metrics/api/domain-classifications/summary |
GET | Classification summary |
/metrics/api/domain-classifications/classify |
POST | Classify domains |
/metrics/api/domain-classifications/progress |
GET | Classification progress |
/metrics/api/link-analytics |
GET | Link analytics data |
Troubleshooting
No Metrics Showing
- Ensure you have completed at least one research
- Check that token tracking is enabled
- Refresh the page
Cost Estimates Incorrect
- Verify model pricing is up to date
- Check provider is correctly identified
- Token counts may vary by model
Rate Limiting Data Missing
- Run some searches to generate data
- Verify rate limiting is enabled in Settings, then refresh the dashboard
See Also
- CLI Tools - Benchmarking and MCP server CLIs
- Full Configuration Reference - All settings and environment variables
- Troubleshooting - Common issues
- Architecture Overview - System design