fix: return 0 completion tokens for empty content

[steebchen]

Summary

• estimateTokensFromContent("") was returning 1 instead of 0 due to Math.max(1, ...), causing error and content_filter responses to report 1 completion token instead of 0
• Added early return of 0 for empty content strings

Test plan

• Updated unit tests to verify empty content returns 0 tokens
• All existing unit tests pass

🤖 Generated with Claude Code

Summary by CodeRabbit

Bug Fixes

• Improved token estimation accuracy for empty content. Empty inputs now correctly estimate to 0 tokens instead of 1, preventing inflated token counts and providing more precise usage tracking.

[coderabbitai]

Walkthrough

This PR modifies the token estimation utility to return 0 tokens for empty content instead of the previous minimum of 1 token. The implementation adds an early guard clause, and corresponding test cases are updated to reflect this new behavior.

Changes

Cohort / File(s)	Summary
Token Estimation Implementation apps/gateway/src/chat/tools/estimate-tokens-from-content.ts	Adds early guard to return 0 tokens for falsy/empty content, replacing the prior behavior that enforced a minimum of 1 token.
Token Estimation Tests apps/gateway/src/lib/prompt-tokens.spec.ts	Updates test expectations for empty string from 1 to 0 tokens; replaces assertion of constant minimum with separate tests for empty (0) and non-empty (≥1) content.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~8 minutes

Possibly related PRs

• fix(chat): improve reasoning token estimation logic #1155: Adjusts estimateTokensFromContent behavior for reasoning token fallback logic that depends on empty-content handling.
• perf(gateway): optimize large response handling #1584: Uses estimateTokensFromContent within chat.ts for token heuristics and will be affected by the empty-content return value change.
• feat(token-usage): enhance token estimation for images #1202: Modifies token estimation behavior and call signatures, extending content-based estimation to handle edge cases like empty strings and image content.

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title 'fix: return 0 completion tokens for empty content' directly and clearly summarizes the main change in the pull request, which adds an early guard to return 0 tokens for empty strings instead of 1.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch fix-zero-completion-tokens

---

No actionable comments were generated in the recent review. 🎉

🧹 Recent nitpick comments

apps/gateway/src/lib/prompt-tokens.spec.ts (1)

11-11: Minor test duplication: empty string assertion appears twice.

Line 11 and lines 19–21 both assert estimateTokensFromContent("") === 0. This is harmless but redundant — consider removing one to keep the suite concise.

Also applies to: 19-21

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[Copilot]

Pull request overview

This PR fixes a bug in token estimation where empty content was incorrectly reporting 1 token instead of 0. The issue affected error responses and content_filter responses that have no content. The fix adds an early return for empty content before the Math.max(1, ...) calculation that was forcing a minimum of 1 token.

Changes:

• Modified estimateTokensFromContent to return 0 for empty content instead of 1
• Updated and added unit tests to verify empty content returns 0 tokens and non-empty content returns at least 1 token

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
apps/gateway/src/chat/tools/estimate-tokens-from-content.ts	Added early return to output 0 tokens for empty content
apps/gateway/src/lib/prompt-tokens.spec.ts	Updated tests to expect 0 tokens for empty content and added dedicated test cases

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.