PNArchives/Memoh
1
0
Fork 0
mirror of https://github.com/memohai/Memoh.git synced 2026-04-27 07:16:19 +09:00
Code Issues Packages Projects Releases Wiki Activity

fix(conversation): correct token trimming edge cases (#207)

Browse Source 
- Treat maxTokens=0 as "unconfigured/unlimited" instead of disabling
  trimming for any non-positive value (which masked exhausted budgets)
- Set historyBudget=1 when maxTokens>0 but overhead exceeds the limit,
  ensuring aggressive trimming instead of no trimming
- Estimate token cost for messages without usage data (len/4 fallback)
  so user/tool messages are not free-passed during budget accounting
This commit is contained in:
Menci
2026-03-09 13:06:19 +08:00
committed by GitHub
parent 90104c49dc
commit c741f2410b
2 changed files with 61 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
internal/conversation/flow/resolver.go
+9 -6
View File
@@ -342,7 +342,9 @@ func (r *Resolver) resolve(ctx context.Context, req conversation.ChatRequest) (r
overhead += systemPromptReserve
historyBudget := maxTokens - overhead
if historyBudget < 0 {
if maxTokens > 0 && historyBudget <= 0 {
historyBudget = 1
} else if historyBudget < 0 {
historyBudget = 0
}
@@ -1279,7 +1281,7 @@ func estimateMessageTokens(msg conversation.ModelMessage) int {
}
func trimMessagesByTokens(log *slog.Logger, messages []messageWithUsage, maxTokens int) []conversation.ModelMessage {
if maxTokens <= 0 || len(messages) == 0 {
if maxTokens == 0 || len(messages) == 0 {
result := make([]conversation.ModelMessage, len(messages))
for i, m := range messages {
result[i] = m.Message
@@ -1287,10 +1289,9 @@ func trimMessagesByTokens(log *slog.Logger, messages []messageWithUsage, maxToke
return result
}
// Scan from newest to oldest, accumulating per-message outputTokens from
// stored usage data. Messages without usage (user / tool) are included for
// free — the outputTokens of surrounding assistant turns already account
// for the context they consumed.
// Scan from newest to oldest, accumulating per-message token costs.
// Messages with stored usage data use that value; others fall back to a
// character-based estimate so that user/tool messages are not free-passed.
totalTokens := 0
cutoff := 0
messagesWithUsage := 0
@@ -1298,6 +1299,8 @@ func trimMessagesByTokens(log *slog.Logger, messages []messageWithUsage, maxToke
if messages[i].UsageOutputTokens != nil {
totalTokens += *messages[i].UsageOutputTokens
messagesWithUsage++
} else {
totalTokens += estimateMessageTokens(messages[i].Message)
}
if totalTokens > maxTokens {
cutoff = i + 1