Walkthrough

The pull request introduces alias resolution functionality for models by adding an Aliases field to ModelEntry, a new ProviderModelID field to ProviderModelEntry, and refactoring the Resolve function to handle alias-to-model lookups with candidate scoring. A reverse-index (aliasTargetsByID) is built during model list initialization to map aliases to their canonical models.

Changes

Sequence Diagram

sequenceDiagram
    participant Caller as Caller
    participant Resolve as Resolve()
    participant Direct as Direct Lookup
    participant Reverse as Reverse Lookup
    participant Alias as Alias Resolution<br/>(aliasTargetsByID)
    participant Score as Score Candidates
    participant ModelRef as Resolve ModelRef

    Caller->>Resolve: Resolve(list, providerType, modelID)
    Resolve->>Direct: Try providerType + modelID
    Direct-->>Resolve: Found or Not
    alt Not Found
        Resolve->>Reverse: Try reverse lookup<br/>via providerModelByActualID
        Reverse-->>Resolve: Found or Not
        alt Still Not Found
            Resolve->>Alias: Check aliasTargetsByID[modelID]
            Alias-->>Resolve: List of alias targets
            alt Has Targets
                Resolve->>Score: Score candidates by<br/>ProviderType match
                Score-->>Resolve: Scored targets
                Resolve->>ModelRef: Select highest-scoring<br/>ModelRef
                ModelRef-->>Resolve: ModelRef + ProviderType
            end
        end
    end
    Resolve-->>Caller: ModelRef or nil

Estimated code review effort

🎯 4 (Complex) | ⏱️ ~60 minutes

Possibly related PRs

• PR #82: Extends internal/modeldata (types.go, merger.go, fetcher_test.go) with alias handling and provider_model_id preference, directly modifying the same ModelList/Resolve implementation and reverse-index logic.
• PR #160: Touches provider-qualified model lookup and registry enrichment behavior, including tests in internal/providers/registry_test.go that validate alias-based model registration.

Poem

🐰 Aliases bloom in the model meadow bright,
Reverse indices guide us through the night,
With scoring and logic we pick just the best,
No ambiguity clouds our quest!
Provider-qualified models now dance with glee,
A cleaner resolution, wild and free! 🎯

❌ Failed checks (1 warning)

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

• X
• Mastodon
• Reddit
• LinkedIn