Fix provider-aware /model picker catalogs#1201
Conversation
|
Warning You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again! |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: a33d4351c4
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".
|
Addressed in Provider-native IDs returned by live catalogs now pass settings validation and survive restart, while DeepSeek aliases still normalize. The picker persistence path also surfaces Validated locally:
|
|
The provider-aware /model picker catalog fix was harvested into #1649 and merged to main with contributor credit. For v0.8.38, that narrow bug-fix slice is covered; leaving this PR open for maintainer judgment on any remaining scope. |
…ouping The v0.8.38 upgrade dramatically changed two user-visible behaviors that were not intended as regressions: - The /model picker was reworked (Hmbown#1201/Hmbown#1632) to make a blocking network fetch on open and replace the curated tier list with the raw provider catalog. Revert model_picker.rs and the OpenModelPicker handler to the v0.8.37 instant curated picker. The /models command still lists the live catalog. - Hmbown#1617 rekeyed the approval cache to an exact full-argument fingerprint, which also dropped the v0.8.37 arity-aware command-family grouping for "approve for session". Reintroduce build_approval_grouping_key (the lossy v0.8.37 logic) for approvals while keeping the exact key for denials, so denying one call no longer over-blocks later differing calls. https://claude.ai/code/session_01NDuRxM56o17SE7SDLcTFYT
…ouping The v0.8.38 upgrade dramatically changed two user-visible behaviors that were not intended as regressions: - The /model picker was reworked (Hmbown#1201/Hmbown#1632) to make a blocking network fetch on open and replace the curated tier list with the raw provider catalog. Revert model_picker.rs and the OpenModelPicker handler to the v0.8.37 instant curated picker. The /models command still lists the live catalog. - Hmbown#1617 rekeyed the approval cache to an exact full-argument fingerprint, which also dropped the v0.8.37 arity-aware command-family grouping for "approve for session". Reintroduce build_approval_grouping_key (the lossy v0.8.37 logic) for approvals while keeping the exact key for denials, so denying one call no longer over-blocks later differing calls. https://claude.ai/code/session_01NDuRxM56o17SE7SDLcTFYT
|
Thanks for the provider/model picker work. We harvested the narrow provider-selected-model preservation pieces into #1649, but the live catalog behavior in /model caused a v0.8.38 regression: a common picker path could block on provider network calls and lost the curated DeepSeek tier list. v0.8.39 reverted that path. Closing this PR so it is not accidentally revived; future provider catalog work should target explicit /models behavior or an async/cached picker design with regression tests. |
Summary
Fixes the
/modelpicker so it reflects the currently selected provider instead of always showing the static DeepSeek Pro/Flash list.Changes:
/model/model <id>through the active provider, so short DeepSeek IDs map to provider IDs such asdeepseek-ai/deepseek-v4-flashfor NVIDIA NIMhigh (thinking enabled)readable with one trailing cell before the border-D warningsTesting
cargo test --workspace --all-features --lockedcargo fmt --all -- --checkcargo clippy --workspace --all-targets --all-features --locked -- -D warningscargo buildLocal environment note:
cargo run -p deepseek-tui --all-features -- evalfails on Windows because the offline eval harness runsprintf, which is not available undercmd.exe.Checklist