mirror of
https://github.com/MindWorkAI/AI-Studio.git
synced 2026-10-14 01:53:48 +00:00
Fixed reasoning tokens inflating the exact token count
This commit is contained in:
12 files changed
+172
-104
No files matched your search
@@ -61,8 +61,11 @@ public readonly record struct ConversationTokens
|
||||
/// </summary>
|
||||
/// <remarks>
|
||||
/// True once a provider has stated what a request of this conversation cost, which makes
|
||||
/// everything but the draft an exact number. It says nothing about the draft, which stays an
|
||||
/// estimate either way -- nobody has charged for that one yet.
|
||||
/// everything up to the last answer an exact number. That answer is counted by this app, since
|
||||
/// the provider's number for it includes reasoning which no request carries; next to the rest
|
||||
/// of the conversation, its share of the error is small enough to still call the history exact.
|
||||
/// It says nothing about the draft, which stays an estimate either way -- nobody has charged
|
||||
/// for that one yet.
|
||||
/// </remarks>
|
||||
public bool HistoryIsReported { get; init; }
|
||||
|
||||
|
||||
Reference in new issue
Block a user