Powerbi and documentation

[zic04]

Idea title

I have created a powerbi following the integration of log exports

Describe your idea

Possible directions or next steps

No response

Additional context

No response

[owndev]

Thanks for sharing. Clever idea and useful implementation. Good work!

[zic04]

Open WebUI Usage Overview

OpenWebuiMetrics_mvp_CL
| summarize 
    TotalTokens = sum(requestTokens_d + responseTokens_d),
    ModelCount = count_distinct(model_s),
    AvgTokensPerSecond = avg(tokensPerSecond_d),
    AvgResponseTime = avg(responseTime_d)
| project Models_Used = ModelCount, Total_Tokens = TotalTokens, Avg_Tokens_Per_Second = AvgTokensPerSecond, Avg_Response_Time = AvgResponseTime

Count users by days

OpenWebuiMetrics_mvp_CL
| where ['TimeGenerated'] >= datetime('2025-04-27T13:31:34Z') and ['TimeGenerated'] <= datetime('2025-04-27T16:31:34Z')
| summarize Users = dcount(userId_g) by Date_Only = bin(TimeGenerated, 1d)
| extend DateAndUsers = strcat(format_datetime(Date_Only, 'yyyy-MM-dd'), ' (', tostring(Users), ' users)')
| project DateAndUsers, Date_Only

Token Usage by Model

OpenWebuiMetrics_mvp_CL
| extend TotalTokens = requestTokens_d + responseTokens_d
| summarize TotalTokens = sum(TotalTokens) by model_s
| where TotalTokens > 0
| order by TotalTokens desc
| project model_s, TotalTokens

Model Evaluation Counts

OpenWebuiMetrics_mvp_CL
| where model_s != ""
| where model_s != "unknown"
| summarize EvalCount=count() by (TimeGenerated), model_s
| order by TimeGenerated asc

LLM Response Historical

OpenWebuiMetrics_mvp_CL
| summarize PromptTokens = sum(requestTokens_d) by bin(TimeGenerated, 1m)
| order by TimeGenerated asc

Prompt Tokens per Second

OpenWebuiMetrics_mvp_CL
| where requestTokens_d > 0
| project TimeGenerated, promptTokens_d=requestTokens_d/responseTime_d
| summarize AvgPromptTokensPerSecond=avg(promptTokens_d) by (TimeGenerated)
| order by TimeGenerated asc

Response Tokens per Second

OpenWebuiMetrics_mvp_CL
| where responseTokens_d > 0
| project TimeGenerated, responseTokens_d=responseTokens_d/responseTime_d
| summarize AvgResponseTokensPerSecond=avg(responseTokens_d) by (TimeGenerated)
| order by TimeGenerated asc

Average Duration Over Time

OpenWebuiMetrics_mvp_CL
| summarize Avg_Response_Time = avg(responseTime_d) by (TimeGenerated)
| order by TimeGenerated asc

Total Prompt Tokens

OpenWebuiMetrics_mvp_CL
| summarize 
    SumRequestTokens = sum(requestTokens_d),
    SumResponseTokens = sum(responseTokens_d),
    TotalTokens = sum(requestTokens_d + responseTokens_d)

[zic04]

I was inspired by the : https://grafana.com/grafana/dashboards/22867-grafana-dashboard-for-open-webui/

here's how it works with grafana