fix(known-provider-models): refresh provider catalog metadata and fallbacks#101
Merged
Conversation
…lbacks
- deepseek: advertise reasoning (thinking mode) for deepseek-v4-flash/pro
via the supports_reasoning flag, mirroring Anthropic's reasoning-capable
handling; no OpenAI-style effort presets since DeepSeek uses a
provider-specific thinking toggle
- cerebras: add gemma-4-31b preview fallback + metadata (vision, no
reasoning presets), keeping gpt-oss-120b as default
- qwen: surface the orphaned qwen3.6-flash metadata entry in FALLBACK_MODELS
- anthropic: add claude-sonnet-5 fallback + metadata and correct stale
"latest Sonnet" wording on claude-sonnet-4-6
- nvidia: fix supports_tools for deepseek-v4-flash/pro (outlier vs the
direct and OpenRouter DeepSeek catalogs, which both allow tools)
- openrouter: add anthropic/claude-{fable-5,opus-4-8,sonnet-5} catalog
metadata (metadata-only, matching existing non-fallback entries)
- tests: cover each of the above
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 subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Provider catalog refresh closing todos 0bd8a614, 48b75f1d, cf7e3b63, 8ac5e7a8, 3fb80ba8, 066d32a3 plus the NVIDIA/qwen audit items. All changes preserve the repo's real Hasna-branded model slugs; only genuinely-missing entries were added and incorrect metadata flags fixed. Verified current state first and skipped items already correct (GLM-5.2 and OpenRouter DeepSeek text-only metadata were already landed and left untouched).
Changes
deepseek-v4-flash/deepseek-v4-pronow advertisesupports_reasoning = true(thinking mode). DeepSeek toggles thinking via a provider-specific parameter rather than the OpenAI-style reasoning-effort scale, so no effort presets are wired — the same treatment Anthropic already gets (supports_reasoningtrue,no_reasoning_levels()). This avoids faking generic effort support. OpenRouter DeepSeek text-only metadata is deliberately left unchanged.gemma-4-31bpreview fallback + metadata (context 131_072, tools, vision via default modalities, no reasoning presets).gpt-oss-120bstays the default.qwen3.6-flashwas orphaned inmetadata()but absent fromFALLBACK_MODELS; it is now exposed as a selectable fallback.qwen3.7-plus/qwen3.7-maxwere already present. Default kept atqwen3.5-flash(default-move was out of the narrowed scope and covered by an existing test).claude-sonnet-5fallback + metadata (1M context, adaptive thinking → reasoning) and corrected the stale "latest Sonnet" wording onclaude-sonnet-4-6.claude-fable-5stays the default.deepseek-ai/deepseek-v4-flash/-prohadsupports_tools = false, an outlier versus the direct DeepSeek and OpenRouter catalogs (bothtrue) and DeepSeek's documented tool-call support; corrected totrue.anthropic/claude-fable-5,anthropic/claude-opus-4-8,anthropic/claude-sonnet-5catalog metadata (metadata-only, matching the existing pattern of non-fallback OpenRouter entries). FALLBACK/default and theopenrouter/fusionquestion were left untouched since the source task is tagged blocked/no-auto.Tests
Added unit tests in
known-provider-models/src/lib.rsfor each item; updated the Anthropic fallback-ordering test.cargo test -p codex-known-provider-models= 16 passed.Verification
cargo fmt -- --config imports_granularity=Item --check— exit 0cargo build -p codex-known-provider-models -p codex-model-provider-info— okcargo test -p codex-known-provider-models -p codex-model-provider-info— 16 + 66 passedcargo clippy -p codex-known-provider-models— clean