LLM CallsiTotal number of LLM API calls in the selected time range. Input TokensiSum of user_prompt_token across all LLM calls. Output TokensiSum of response_token across all LLM calls. Est. CostiEstimated cost using a flat rate of ~$3 per 1M tokens. This is a rough average — does not differentiate by model or input/output. LLM Calls TrendiHourly count of LLM API calls.
Token Usage TrendiHourly token usage broken down by input (user_prompt), output (response), and system (system_prompt).
Est. Cost Trend ($)iHourly estimated LLM cost. Rough rates: ~$1/1M input tokens, ~$5/1M output tokens.
Est. Cost by Model ($)iEstimated cost per model. Rough rates: ~$1/1M input tokens, ~$5/1M output tokens.
Model Calls Over TimeiHourly call count broken down by provider and model.
Avg Latency by Model (ms)iHourly average response latency per model group. Only successful calls.
Model LatencyiAvg TTFT = average time to first token (response_start_latency). Avg Total = average full response time. Only successful calls.
Model Success / Error RateiSuccess and error counts per model. Error Rate = error count / total calls per model.
Usage by FeatureiDistribution of LLM calls grouped by feature field. Top 10 features.
Error TrendiHourly count of failed LLM calls over time.
Token Usage by ModeliTotal token usage (user_prompt + response) broken down by model. Top 10 models.
Model Call DistributioniPie chart showing the share of LLM calls per model. Top 10 models.
Monitor