26 lines
987 B
Markdown
26 lines
987 B
Markdown
# PR#43: the statistics show an unknown model cost as a lower bound
|
|
|
|
> **WARNING:** This document describes only the change applied in this PR.
|
|
> It may already be outdated once the next PR is merged.
|
|
> Historic PR documentation is not maintained along with new PRs; treat it as a snapshot.
|
|
|
|
## The Problem
|
|
|
|
When the model call of a review fails, e.g. after a timeout, PR-Agent reports no usage, so its cost is unknown.
|
|
The list of recent reviews on the statistics page then shows the cost as `≈ 0.0001 USD + LLM unknown`.
|
|
The text in the cost column is no amount and widens the column.
|
|
|
|
## Non-Goals
|
|
|
|
- No change of the sums: they keep the count of reviews with an unknown cost.
|
|
|
|
## The Solution
|
|
|
|
- The cost column shows the known part as a lower bound, e.g. `≥ 0.0001 USD`.
|
|
- A tooltip on the amount says that the cost of the model call is unknown.
|
|
- werkrevisor 0.21.3.
|
|
|
|
## Prerequisite PRs
|
|
|
|
- PR#40: werkrevisor 0.21.1, a failed advising model leaves the verdict to Jev, merged.
|