Skip to content

LLM Analytics Dashboard Missing Model + Cost Data #338

@mottych

Description

@mottych

Description

LLM Analytics Dashboard Not Displaying Model Names or Token Cost Calculations

Priority

Low (non‑blocking, does not impact core user workflows)

Environment

  • Purpose Path Admin Dashboard
  • Section: LLM Analytics → Usage & Invocations
  • Date observed: May 7, 2026
  • QA: Daily validation workflow

Issue Summary

The LLM analytics dashboard is missing two critical pieces of information:

  1. Model column is blank
    • The “Model” field shows a dash () instead of the actual model used for each invocation.
  2. Cost column always shows 0.00
    • Even when token usage is present (e.g., 24,758 total tokens), the cost field remains zero.
    • No breakdown of input vs output token cost is displayed.
      These fields are required for accurate accounting, cost tracking, and model‑level usage analysis.

Expected Behavior

  • The Model column should display the correct model name (e.g., GPT‑5.4, GPT‑5.5, etc.).
  • The Cost column should calculate and display:
    • Input token cost
    • Output token cost
    • Total cost
  • The summary panel should reflect accurate cost totals based on token usage.

Actual Behavior

  • Model column displays for all rows.
  • Cost column displays 0.00 for all rows.
  • Summary section shows:
    • Total tokens: 24,758
    • In/Out tokens: 19,220 / 4,773
    • Cost (USD): 0.00 (incorrect)

Impact

  • Prevents accurate AI cost accounting.
  • Blocks financial forecasting and usage auditing.
  • Limits visibility into which models are being used across tenants.
  • Not urgent, but important for long‑term operational clarity.

Steps to Reproduce

  1. Open Purpose Path Admin → LLM Analytics.
  2. Go to Usage & Invocations.
  3. Apply any filter or leave default.
  4. Observe the Model and Cost columns.
  5. Observe the Summary panel.

Suspected Cause

  • Missing or unlinked model metadata.
  • Pricing table not applied to token usage rows.
  • Cost calculation logic not implemented or returning null.
  • Backend not returning model identifiers.

Recommendation

  • Add model metadata to invocation records.
  • Implement cost calculation using the correct pricing table.
  • Update UI to display:
    • Model name
    • Input cost
    • Output cost
    • Total cost

Priority

normal

Attachments

Metadata

Metadata

Assignees

No one assigned

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions