Add export to CSV/JSON for further analysis

[rchojn]

Problem

Users want to analyze eval results in external tools (Excel, Jupyter, R, Tableau) but GoEvals only shows data in the dashboard.

Proposed Solution

Add export functionality:

• Export to CSV (for Excel, Google Sheets)
• Export to JSON (for programmatic analysis)
• Filter before export (date range, model, score threshold)
• Batch export (all results or selection)

Competitive Context

• Langfuse: API export with filters
• Phoenix: Dataset export for fine-tuning
• Helicone: CSV/JSON export built-in

GoEvals approach: Client-side export using browser APIs, no server-side processing needed.

Implementation

1. Add export buttons to dashboard:

   • "Export to CSV" button
   • "Export to JSON" button
   • Filter controls (date range, model selection)

2. Client-side export logic:

// Server: Provide filtered results as JSON
func handleExportJSON(w http.ResponseWriter, r *http.Request) {
    // Parse filters from query params
    results := filterResults(allResults, filters)
    json.NewEncoder(w).Encode(results)
}

// Client: Browser downloads as file
<button onclick="downloadCSV()">Export to CSV</button>

3. CSV format:

timestamp,model,test_id,combined_score,accuracy,fluency,response_time_ms
2025-10-26T14:30:00Z,gemma2:2b,eval_001,0.85,0.90,0.88,1234

Value for SafeReader

• Statistical analysis: Run t-tests comparing models in R/Python
• Reporting: Create monthly performance reports in Excel
• Sharing: Send eval results to stakeholders as CSV
• Archiving: Backup results outside JSONL format

Philosophy

• Client-side export - no server processing overhead
• Stdlib only - generate CSV/JSON in Go
• Local-first - export happens in browser
• No external services - no cloud export APIs

Success Metrics

• Export 1000+ results to CSV in < 2 seconds
• CSV opens correctly in Excel/Google Sheets
• JSON is valid and parseable by jq/Python