Repository navigation
fix: group model tiers and fix provider sign-in - #1030
Open
small-angel wants to merge 2 commits into
Open
small-angel wants to merge 2 commits into
small-angel wants to merge 2 commits into
Conversation
OpenCode lists speed tiers as their own models (claude-fable-5-2-fast, gpt-5-flex, gpt-6-astra-ultrafast), so the picker showed each twice. A trailing fast, flex, ultrafast or thinking word after the version now marks a tier of the plain model, and the Effort menu offers speed as one choice. A latest alias (mistral-large-latest) is left out when its catalog lists the versions. OpenCode 2 lists GitHub Copilot with only an API key until its plugins activate, so orx now waits for them before reading sign-ins and models. Listed providers link to their key pages, an agent install waits for another install instead of failing on the lock, and the Name step no longer flashes "no models" before it re-checks. Assisted-by: claude-opus-5.5
Contributor
[Medium impact] The PR changes model selection and setup flows across the app. The alias filter should reach Settings and Add agent before merging to satisfy the repository’s cross-view requirement. Findings
Fix with agent prompt### Issue 1
ui/src/components/pickerAgents.tsx:90
The new filter hides aliases only in `agentOffers`. Settings and Add agent still pass every model to `agentModelLines`. A catalog containing `mistral-large-latest` and its versions therefore advertises extra models that users cannot find in the picker.
Apply the same alias filter to those counts and lists before merging. The repository requires that “a feature that works in one view works wherever that thing appears.”
---
For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.SummaryGroups model speed and thinking tiers under one picker entry and adds speed choices to the Effort menu.
Reviews (1) · Last reviewed commit: "fix: group model tiers and fix provider ..." · Reviewed by Greptile |
Settings and the add-agent dialog counted and listed aliases the picker already leaves out. Assisted-by: claude-opus-5.5
This branch has not been deployed
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
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.
OpenCode lists speed tiers as models of their own (
claude-fable-5-2-fast,gpt-5-flex,gpt-6-astra-ultrafast), so the picker showed Claude Fable 5.2 twice. A trailingfast,flex,ultrafastorthinkingafter the version, in any agent's id or name, now marks a tier of the plain model, and the Effort menu offers speed as one icon choice (Flex · Standard · Fast · Ultrafast; a single tier stays the Fast switch). Alatestalias likemistral-large-latestis left out when its catalog lists the versions it points between.OpenCode 2 lists GitHub Copilot with only a generic API key until its plugins activate, which is the Copilot "API key" field people hit. orx now waits for
/api/pluginbefore reading integrations and models, so Copilot, ChatGPT, SuperGrok and GitLab sign-ins show. Also: listed providers link "Get a key" to their key pages, an agent install waits for another install or a background update instead of failing on the lock, and the Name step no longer flashes "no models" for a frame before it re-checks.Load-bearing: the wire
variantgainsspeedand model names gainsuperseded;variant.fastis still sent for dashboards from 0.2.27.Proved live in the dev app with real Anthropic, OpenAI, Gemini, Mistral, OpenRouter, Together, Cohere, DeepSeek and Perplexity keys (one row per model, Speed choice switching gpt-5 ↔ gpt-5-flex, aliases hidden, Copilot device sign-in reaching the code), plus cargo test, clippy,
pnpm test, andnode e2e/run.ts. Not checked on Windows or Linux.Reviewed with sol-loop: 2 rounds with gpt-6.1-sol. Sol verdict: approved
Assisted-by: claude-opus-5.5