Skip to content

docs(observability): sync reasoning-aware TPS spec - #3

Closed
mastertyko wants to merge 1 commit into
mainfrom
docs/sync-tps-speed-spec
Closed

mastertyko wants to merge 1 commit into
mainfrom
docs/sync-tps-speed-spec

Conversation

@mastertyko

Copy link
Copy Markdown
Owner

Change

Sync the active expand-ttft-output-events delta into the main proxy-runtime-observability spec.

Details

  • Define TPS using non-reasoning output tokens.
  • Extend websocket first-token events to token-bearing deltas and custom/apply-patch tool items.
  • Exclude lifecycle, placeholder, metadata-only, and empty tool events.
  • Align source-routed TPS wording and preserve existing timing scenarios.

Verification

  • openspec validate proxy-runtime-observability --type spec --strict — valid

The active delta remains available for later archival.

@mastertyko

Copy link
Copy Markdown
Owner Author

@codex review

@mastertyko

Copy link
Copy Markdown
Owner Author

Superseded by upstream PR Soju06#1889.

@mastertyko mastertyko closed this Aug 22, 2026

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

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

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a6553642b5

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

### Requirement: Dashboard request logs show generation speed

The dashboard request-log table MUST show time to first token and output-token generation speed when the required latency and output-token fields are available. Generation speed MUST use output tokens divided by elapsed generation time after time to first token, not total input plus output tokens and not total request latency including TTFT.
The dashboard request-log table MUST show time to first token and output-token generation speed when the required latency and output-token fields are available. Generation speed MUST use non-reasoning output tokens divided by elapsed generation time after time to first token, not total input plus output tokens and not total request latency including TTFT. The displayed metric MUST remain named `TPS`.

Copy link
Copy Markdown

Choose a reason for hiding this comment

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

P2 Badge Define the fallback for unknown reasoning usage

For older rows, prematurely terminated streams, and custom model-source responses, reasoning_tokens can be null even when output tokens and timing are present. In that case the dashboard subtracts zero (frontend/src/features/dashboard/components/recent-requests-table.tsx:287) and Reports does the same via coalesce (app/modules/reports/repository.py:482), so both can include reasoning tokens despite this new unconditional non-reasoning-TPS requirement. Specify whether unknown reasoning is a missing speed input or an explicit compatibility fallback; otherwise the normative SSOT promises behavior the implementation cannot provide.

AGENTS.md reference: AGENTS.md:L44-L45

Useful? React with 👍 / 👎.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant