Skip to content

fix: cache token double counting at message_start #2241

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
xiaolin593 wants to merge 2 commits into
base: main
Choose a base branch
from
+32 −7
Open

fix: cache token double counting at message_start #2241

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
7b31cb4
fix cache token double counting
xiaolin593 Jun 16, 2026
9b22d15
add test
xiaolin593 Jun 16, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 0 additions & 7 deletions internal/translator/anthropic_helper.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -971,13 +971,6 @@ func (p *anthropicStreamParser) handleAnthropicStreamEvent(eventType []byte, dat
if input, ok := usage.InputTokens(); ok {
p.tokenUsage.SetInputTokens(input)
}
if cached, ok := usage.CachedInputTokens(); ok {

@PatilHrushikesh PatilHrushikesh Jun 18, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

i think SetInputTokens should also be moved to MessageDelta block, so all token related thing happens at one place

p.tokenUsage.SetCachedInputTokens(cached)
}
if cacheCreation, ok := usage.CacheCreationInputTokens(); ok {
p.tokenUsage.SetCacheCreationInputTokens(cacheCreation)
}

// reset the toolIndex for each message
p.toolIndex = -1
return nil, nil
Expand Down
32 changes: 32 additions & 0 deletions internal/translator/anthropic_helper_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1157,3 +1157,35 @@ func TestBuildAnthropicParamsWithReasoningEffort(t *testing.T) {
require.Equal(t, anthropic.OutputConfigEffort(""), params.OutputConfig.Effort)
})
}

// TestAnthropicStreamParser_CacheTokensNotSetFromMessageStart verifies that cache tokens are not
// written during message_start processing. Before the fix, message_start called SetCachedInputTokens
// and SetCacheCreationInputTokens; if message_delta then called AddCachedInputTokens with the same
// values, both would be counted, doubling the result.
func TestAnthropicStreamParser_CacheTokensNotSetFromMessageStart(t *testing.T) {
p := newAnthropicStreamParser("claude-3-5-sonnet")

// Simulate message_start with cache tokens.
messageStart := []byte(`{"type":"message_start","message":{"id":"msg_test","type":"message","role":"assistant","content":[],"model":"claude-3-5-sonnet","stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":10,"cache_read_input_tokens":5,"cache_creation_input_tokens":2,"output_tokens":0}}}`)
_, err := p.handleAnthropicStreamEvent([]byte("message_start"), messageStart)
require.NoError(t, err)

// Cache tokens must NOT be set after message_start; only input tokens are set.
_, cachedSet := p.tokenUsage.CachedInputTokens()
require.False(t, cachedSet, "cache_read tokens must not be set from message_start")
_, cacheCreationSet := p.tokenUsage.CacheCreationInputTokens()
require.False(t, cacheCreationSet, "cache_creation tokens must not be set from message_start")

// Simulate message_delta carrying the authoritative cache and output counts.
messageDelta := []byte(`{"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"output_tokens":8,"cache_read_input_tokens":5,"cache_creation_input_tokens":2}}`)
_, err = p.handleAnthropicStreamEvent([]byte("message_delta"), messageDelta)
require.NoError(t, err)

cachedTokens, cachedSet := p.tokenUsage.CachedInputTokens()
require.True(t, cachedSet)
require.Equal(t, uint32(5), cachedTokens, "cache_read tokens must not be double-counted")

cacheCreationTokens, cacheCreationSet := p.tokenUsage.CacheCreationInputTokens()
require.True(t, cacheCreationSet)
require.Equal(t, uint32(2), cacheCreationTokens, "cache_creation tokens must not be double-counted")
Comment on lines +1184 to +1190

@gemini-code-assist gemini-code-assist Bot Jun 16, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

To make the test more robust and ensure that the total input tokens (prompt tokens) and output tokens are also correctly calculated and not double-counted, consider adding assertions for the total input and output tokens.

	cachedTokens, cachedSet := p.tokenUsage.CachedInputTokens()
	require.True(t, cachedSet)
	require.Equal(t, uint32(5), cachedTokens, "cache_read tokens must not be double-counted")

	cacheCreationTokens, cacheCreationSet := p.tokenUsage.CacheCreationInputTokens()
	require.True(t, cacheCreationSet)
	require.Equal(t, uint32(2), cacheCreationTokens, "cache_creation tokens must not be double-counted")

	inputTokens, inputSet := p.tokenUsage.InputTokens()
	require.True(t, inputSet)
	require.Equal(t, uint32(17), inputTokens, "total input tokens must be correctly calculated")

	outputTokens, outputSet := p.tokenUsage.OutputTokens()
	require.True(t, outputSet)
	require.Equal(t, uint32(8), outputTokens, "output tokens must be correctly set")

}