LLM Analytics Endpoints

🔒 Authentication

Endpoints in this group (10 total)

Common Query Parameters

All endpoints accept:

GET /api/v1/llm-analytics/summary

High-level KPIs for the LLM system performance.

200 OK example:

{
  "total_calls": 1842,
  "total_cost_usd": 48.32,
  "avg_confidence": 0.74,
  "avg_win_rate": 0.612,
  "total_pnl": 3250.00,
  "best_model": "claude-opus-4-7",
  "most_used_provider": "anthropic"
}

Pydantic Schema: backend/api/routes/llm_analytics.py :: LLMAnalyticsSummary

GET /api/v1/llm-analytics/model-performance

Per-model breakdown of call count, avg confidence, win rate, and total P&L attributed to LLM decisions.

200 OK example:

[
  {
    "provider": "anthropic",
    "model": "claude-opus-4-7",
    "calls": 850,
    "avg_confidence": 0.78,
    "win_rate": 0.64,
    "total_pnl": 2100.00,
    "avg_cost_per_call": 0.028
  }
]

Pydantic Schema: backend/api/routes/llm_analytics.py :: ModelPerformanceRow

GET /api/v1/llm-analytics/heatmap

Win rate matrix: rows = symbols, columns = models. Visualized as a color heatmap in Page - LLM Analytics.

200 OK example:

{
  "symbols": ["EURUSD", "GBPUSD", "XAUUSD"],
  "models": ["claude-opus-4-7", "gpt-4o"],
  "data": [
    [0.65, 0.58],
    [0.71, 0.62],
    [0.55, 0.49]
  ]
}

GET /api/v1/llm-analytics/learning/confidence-calibration

Compares stated LLM confidence vs. actual win rate, bucketed by confidence range. Reveals over/under-confidence.

200 OK example:

[
  { "bucket": "0.6-0.65", "avg_confidence": 0.62, "actual_win_rate": 0.55, "count": 45 },
  { "bucket": "0.65-0.70", "avg_confidence": 0.67, "actual_win_rate": 0.61, "count": 88 },
  { "bucket": "0.70-0.75", "avg_confidence": 0.72, "actual_win_rate": 0.68, "count": 120 }
]

GET /api/v1/llm-analytics/learning/lessons

Auto-extracted lessons from completed trades — patterns the LLM's post-trade analysis identified repeatedly.

200 OK example:

[
  {
    "lesson": "BUY signals on EURUSD during London open (07:00-09:00 UTC) outperform by 18%",
    "frequency": 12,
    "avg_pnl_impact": 45.20
  }
]

🗂️ Related Files

🗂️ Related