feat(gateway): add opt-in token usage footer to responses - #11394
Open
Floris-Jan wants to merge 1 commit into
Open
feat(gateway): add opt-in token usage footer to responses#11394Floris-Jan wants to merge 1 commit into
Floris-Jan wants to merge 1 commit into
Conversation
Token data is tracked internally (session_prompt_tokens, etc.) but only visible via explicit /usage command. Users report 'token cost anxiety' — they want per-response visibility without asking. - New config: display.token_usage_footer (default: false) - Appends compact footer: '📊 1.2k in · 456 out · ~$0.0234 · claude-sonnet-4' - Gated by per-platform display config (same system as show_reasoning) - Skipped on failed responses
alt-glitch
added
type/feature
teknium1
reviewed
Jul 12, 2026
teknium1
left a comment
teknium1
left a comment
Contributor
There was a problem hiding this comment.
Thanks for the focused opt-in footer contribution. The token/cost capability is not yet present on current main, but this implementation predates the merged canonical footer system and needs a rewrite rather than a direct salvage.
Problems
gateway/run.py:18818-18819populatesagent_result["input_tokens"]andagent_result["output_tokens"]from cumulative session counters. The proposed reads atgateway/run.py:4032-4033would display and price session totals after each reply, not per-response usage.- Current
gateway/runtime_footer.py:106-118is the canonical configurable footer renderer, merged ine123f4ecf(#17026). It supportsmodel,context_pct, andcwd; the proposed separatedisplay.token_usage_footerpath would bypass that system and its streaming trailing-footer handling ingateway/run.py:12065-12071. tests/gateway/test_token_usage_footer.py:6copies formatter logic into a test helper, so it does not exercise the production config, result, or delivery paths.
Suggested changes
- Extend
display.runtime_footer.fieldsandgateway/runtime_footer.pywith usage/cost fields. - Pass a per-turn usage delta to the renderer, then cover the real renderer and both streamed and non-streamed gateway delivery paths.
This is an automated hermes-sweeper review.
| _show_usage_footer = False | ||
|
|
||
| if _show_usage_footer and response and not agent_result.get("failed"): | ||
| _input_toks = agent_result.get("input_tokens", 0) or 0 |
Contributor
There was a problem hiding this comment.
agent_result["input_tokens"] is populated from the cumulative AIAgent.session_prompt_tokens counter on current main (gateway/run.py:18818-18819), so this would show a growing session total on every reply rather than the claimed per-response usage. Please plumb a per-turn usage delta before rendering or pricing it.
| import pytest | ||
|
|
||
|
|
||
| def _format_usage_footer(input_tokens, output_tokens, model, cost_amount=None, cost_status="estimated"): |
Contributor
There was a problem hiding this comment.
This helper duplicates production formatting rather than testing the gateway path. Please test the canonical footer renderer/delivery integration so config resolution, failed responses, and streamed trailing-footers are covered.
teknium1
added
sweeper:risk-message-delivery
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What does this PR do?
Token data is tracked internally (
session_prompt_tokens,session_completion_tokens) but only visible via explicit/usagecommand. Users report "token cost anxiety" — they want per-response visibility without asking.Changes:
display.token_usage_footer(default:false)📊 1.2k in · 456 out · ~$0.0234 · claude-sonnet-4show_reasoning)Type of Change
Changes Made
gateway/run.py— after reasoning display block, beforeagent:endhook: readsdisplay.token_usage_footerconfig (viaresolve_display_setting), formats compact footer with token counts + estimated cost + model name, appends to responsehermes_cli/config.py— added"token_usage_footer": Falsetodisplaydefaultstests/gateway/test_token_usage_footer.py(new) — 8 tests covering formatting (small/large counts, cost/no-cost, model name extraction) and config default verificationHow to Test
pytest tests/gateway/test_token_usage_footer.py -v— all 8 tests passdisplay.token_usage_footer: truein~/.hermes/config.yaml📊 X in · Y out · $Z · model-nameChecklist
feat(gateway): add opt-in token usage footer to responsespytest tests/gateway/test_token_usage_footer.py -v— 8 passedcli-config.yaml.example— N/A (no example file for gateway display keys)