Repository navigation
Fix Ringside native history and duplicate model labels - #145
Merged
Merged
Conversation
justfinethanku
changed the title
boydwold
added a commit
to boydwold/ringer
that referenced
this pull request
Oct 8, 2026
…Jones-Projects#145, reuse open tab NateBJones-Projects#149) Conflicts: - ringer.py: kept both hud_listen (fork) and reuse_open_tab (upstream). - dashboard/ringside.html: took upstream's redesign; the fork's iframe height fix is moot because the new page has no iframe. Co-Authored-By: Claude Opus 5.5 <[email protected]>
Zebra-rancher
added a commit
to Zebra-rancher/ringer
that referenced
this pull request
Oct 11, 2026
…#3) * Preserve fix-swarm patches across retries * Document check-writing failure modes * Warn early for unresolved engine binaries * Fix demo checks to enforce exact output bytes * registry: add claude + gemini engines; MODEL-NOTES audition entries Register two worker engines in the model-identity taxonomy: - claude (Claude Code CLI): Anthropic sonnet/haiku/opus/fable via `--model {model}`, default sonnet. Proven working 2026-09-03 (haiku probe, PASS attempt 1). MODEL-NOTES documents the setup-token auth path and its two footguns (silent expiry with no refresh token; paste truncation -> 401 invalid). - gemini (Gemini CLI): gemini-3.5-flash, verified 2026-09-02. MODEL-NOTES also adds dated audition entries for codex, glm-5.2, and gemini-3.5-flash from 2026-09-02/03 runs. Co-Authored-By: Claude Opus 4.8 <[email protected]> * registry: key claude models by reported slug; sonnet/opus bakeoff Fix the claude engine entries to key on the harness self-reported slug (claude-sonnet-5, claude-opus-5, claude-haiku-4-5-20251001) instead of the CLI aliases, per docs/TAXONOMY.md — alias keys left every claude model showing as [unregistered] on the scoreboard. Note that `--model opus` now self-reports claude-opus-5 (not 4.8). fable stays unregistered until a run self-reports its slug. Adds a code-feature bakeoff MODEL-NOTES entry: sonnet and opus both PASS attempt 1 on roman_to_int; opus ~5x tokens and slower for an identical correct result, so sonnet is the default lane. Co-Authored-By: Claude Opus 4.8 <[email protected]> * registry: add claude-fable-5-1 from identity probe fable probe PASS attempt 1, self-reports claude-fable-5-1 (1M context, provider firstParty). Now registered, completing the claude engine's four models (sonnet-5, opus-5, haiku-4-5, fable-5-1). MODEL-NOTES flags fable's ~$0.67 trivial-probe cost — orchestrator-class, not a worker lane. Co-Authored-By: Claude Opus 4.8 <[email protected]> * docs: credit mlava for check-writing guidance * docs: clarify demo literal and credit contributor Co-authored-by: Jonathan Edwards <[email protected]> * Credit mlava for early worker command warnings (NateBJones-Projects#59) * docs(model-notes): hermes-dave-brain-memory run — codex 2/2, glm sandbox lesson, check-quoting lesson Co-Authored-By: Claude Fable 5.1 <[email protected]> * MODEL-NOTES: codex lessons from the daily-desk-increment-1 run Co-Authored-By: Claude Fable 5.1 <[email protected]> * Credit mlava for preserving retry patches (NateBJones-Projects#56) * MODEL-NOTES: finapp-vehicles run; brittle quote-style check cost 173k tokens Co-Authored-By: Claude Opus 5 <[email protected]> * Strip per-token API keys from every worker env Codex and Claude Code prefer OPENAI_API_KEY / ANTHROPIC_API_KEY over their ChatGPT and Max logins, so an inherited key silently turned a subscription run into a metered API run. worker_env() removes ANTHROPIC_API_KEY, OPENAI_API_KEY, and CODEX_API_KEY at spawn. OpenCode keeps its OpenRouter key in its own auth store and is unaffected. Co-Authored-By: Claude Opus 5.5 <[email protected]> * Redesign Ringside around runs, outputs and model evidence (NateBJones-Projects#144) * Fix Ringside native history and duplicate model labels (NateBJones-Projects#145) * Keep native Ringside run history stable across data feeds * Use the model display label without duplicating reasoning effort * README: credit fork contributor (contributor-credit test) Co-Authored-By: Claude Opus 5.5 <[email protected]> * ci: trigger fork test run Co-Authored-By: Claude Opus 5.5 <[email protected]> --------- Co-authored-by: Mark Lavercombe <[email protected]> Co-authored-by: Fiddlehead-MB <[email protected]> Co-authored-by: Zebra-rancher <[email protected]> Co-authored-by: Claude Opus 4.8 <[email protected]> Co-authored-by: Jonathan Edwards <[email protected]>
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
1 participant
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.
The native tray poller drops completed runs after 60 seconds, while the browser HUD keeps longer history. After #144 both feeds could replace the same screen, making completed history disappear between HTTP refreshes.
Prefer the full HUD snapshot while its server is reachable. Keep native events as a fallback when the HUD is offline, and return to the HUD snapshot when it recovers. Regression checks cover empty tray events, retained completed history, offline fallback, and recovery. All 269 Python tests passed, the Node behavior checks passed, and the native app rebuilt with source/bundle byte checks.
Also use the API's model display label directly, since it already includes reasoning effort. This avoids labels such as “GPT-5.6 Sol · high high”; a regression check verifies effort appears once.