You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
feat: add modelNameTemplate for discovered model display names - #55
Adds an optional modelNameTemplate setting per provider. Every discovered model's display name is built from it, so two providers backed by the same proxy and different keys stop showing identical entries in the picker.
Placeholders: {name} (today's formatted name, or the raw id with formatModelNames: false), {id}, {provider_id}, {provider_name}.
Default is "{name}", so behaviour is unchanged. It also normalizes to "unset", so existing cache keys stay warm instead of rediscovering on upgrade.
An explicit models.<id>.name still wins and is never templated.
Unknown placeholders are left as literal text; empty, whitespace-only and non-string values are ignored.
Substitution is single-pass, so a value that itself contains {id} is inserted literally rather than expanded again.
{provider_name} is the provider's configured name, or its id when none is set.
On OpenCode 1, Responses models are served by a <provider>-responses sibling, so {provider_id}/{provider_name} resolve to that sibling for those models (the sibling's generated label is <base name or id> Responses). The cached entries stay as discovered, because OpenCode 2 reads the same cache with a single provider.
Part of the cache identity alongside formatModelNames: the template and the resolved provider name go into the fingerprint, so a template change or a provider rename triggers a fresh discovery instead of serving stale names. Without a template the key is byte-identical to today's.
Works on OpenCode 2 and OpenCode 1, including background refresh.
Implementation
applyModelNameTemplate / ModelNaming in src/utils/format-model-name.ts; readModelNameTemplate next to readFormatModelNames.
Applied once in toConfigModel; the Responses sibling re-renders its own names at injection time without mutating the cache.
Naming threaded through discoverModels, refreshContexts and the v2 ProviderSource.
buildCacheKey takes ModelNaming and appends the template + resolved provider name to the fingerprint only when a template is set.
runtimeSettings strips the plugin-level modelNameTemplate option, so a plugin option never lands in the host's provider settings. A template already present in provider.settings is preserved — it is the host's own config and later reconciles re-derive the naming and cache key from it.
test/model-cache.test.ts: key changes on template or provider rename, and stays plain when no template is set.
test/plugin-v1.test.ts: templated injection, curated names preserved, cache keyed by the template, sibling placeholders for named and unnamed bases.
test/plugin-v2.test.ts: templated model registration from provider.settings, curated name preserved, setting preserved, plugin-level option kept out of info.settings, {provider_name} id fallback, cache keyed by the template.
npm run typecheck clean; npm test 186/186.
Summary by CodeRabbit
New Features
Customize discovered model display names with templates using {name}, {id}, {provider_id}, and {provider_name}. Templates are applied after existing name formatting; explicitly configured model names remain unchanged.
Unknown placeholders are left as written, and substitutions are applied once. For Responses API models in OpenCode 1, provider placeholders use the Responses sibling’s identity.
Bug Fixes
Model lists now reflect discovery changes while preserving customized entries. Naming settings are included in cache identity to prevent stale names from being reused.
Documentation
Added guidance on template placeholders, default behavior, provider-name fallback, and interaction with existing name formatting. Empty or non-string template settings are ignored.
No actionable comments were generated in the recent review.
Walkthrough
The change adds modelNameTemplate for discovered model names in both plugin versions. Templates support model and provider placeholders. Naming settings pass through discovery and refresh, affect cache identity, and apply to Responses models. The v2 plugin also tracks published model snapshots during provider replacement.
Changes: Discovered Model Name Templates
Layer / File(s)
Summary
Template definition and model conversion src/utils/format-model-name.ts, src/plugin/index.ts, test/to-config-model.test.ts
Adds shared naming options and placeholder substitution for discovered model names. Tests cover supported and unknown placeholders, single-pass substitution, formatting options, and invalid template values.
Both plugin versions pass naming settings through discovery and refresh. Responses models use the sibling provider's identity for template values. Cache keys include configured templates and provider names. Tests and documentation cover these behaviors.
Provider replacement and model provenance src/plugin/v2.ts, test/plugin-v2.test.ts
Provider sources retain snapshots of published models and snapshots inherited during replacement. Merging compares live entries with those snapshots to determine whether entries are replaced, removed, or preserved.
yuseferi/opencode-litellm#53: Changes credential inputs to cache identity and discovery refresh, which share code paths with this PR’s naming settings.
Suggested reviewers:alexanderwillner, mkowalski
Merge Risk: Moderate · up to b4e8f
A transient discovery failure during provider takeover can leave the provider without models until refresh succeeds. Preserve the last-good inventory and replacement provenance before merging.
Pre-merge checks: 4 · 1
Failed checks: 1 warning
Check name
Status
Explanation
Resolution
Docstring Coverage
Docstring coverage is 78.95% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff with cyclomatic complexity of at least 2. Documenta…
Write docstrings for the functions missing them to satisfy the coverage threshold.
Passed checks: 4 passed
Check name
Status
Explanation
Description Check
Check skipped - CodeRabbit’s high-level summary is enabled.
Title check
The title clearly and concisely describes the main change: adding the modelNameTemplate feature for discovered model display names.
Linked Issues check
Issue #52 is the only active direct target. The change adds provider-level modelNameTemplate handling and applies it after formatted or raw model naming. applyModelNameTemplate supports {name}, …
Out of Scope Changes check
The changes stay within Issue #52. The OpenCode 1 Responses-sibling handling, background-refresh threading, cache updates, documentation, and tests support provider-specific discovered names. The Open…
Full details: Linked Issues check
Explanation
Issue #52 is the only active direct target. The change adds provider-level modelNameTemplate handling and applies it after formatted or raw model naming. applyModelNameTemplate supports {name}, {id}, {provider_id}, and {provider_name}, performs single-pass replacement, preserves unknown placeholders, and ignores invalid or blank settings. Explicit curated names remain unchanged. The provider name falls back to the provider ID. Cache identity includes the template and resolved provider name when a template is set, while the default remains compatible with existing keys. OpenCode 1, OpenCode 2, discovery, refresh, and source replacement paths carry the naming data. Tests cover the required naming, cache, override, fallback, and refresh behavior.
Full details: Out of Scope Changes check
Explanation
The changes stay within Issue #52. The OpenCode 1 Responses-sibling handling, background-refresh threading, cache updates, documentation, and tests support provider-specific discovered names. The OpenCode 2 provenance and replacement logic re-renders stale generated names when a provider source changes and preserves curated entries during that update. No unrelated change is established.
Full details: Docstring Coverage
Explanation
Docstring coverage is 78.95% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff with cyclomatic complexity of at least 2. Documentation-only edits and deletions within surviving functions are included; unchanged diff context and fully deleted functions are excluded. Analyzed 19 functions across 8 files.
Resolution
Write docstrings for the functions missing them to satisfy the coverage threshold.
Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Tip
🎯 Customize your risk assessments
Guide how CodeRabbit assesses merge risk and security architecture risk. Highlight sensitive changes and explain your security boundaries using Merge Risk Instructions and Security Risk Instructions in your review settings. Security architecture risk assessments require the Advanced plan or above.
✅ Automatic approvals, with your rules
Let CodeRabbit approve pull requests once required threads are resolved, pre-merge checks pass, and your risk limits are met. Enable Automatic approvals in review settings, with overrides for teams or file paths.
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.
Run CodeRabbit Verify to test the changes in this pull request.
Recent review info
Run configuration
Source mode: Auto
Repository: Unavailable
Branch: Unavailable
File: No configuration file from this repository applied
Commit: Unavailable
Layers: defaults
Review profile: CHILL
Plan: Advanced
Run ID: 9b764a77-26de-414f-b1e4-ddbe05cb2bbc
Commits
Reviewing files that changed from the base of the PR and between 52687d0 and 1ca824d.
Files selected for processing: 2
src/plugin/v2.ts
test/plugin-v2.test.ts
Included review availability: This review used your included allowance. Your plan provides up to 2 included reviews per hour; 1 remain after this review.
Adds a per-provider template that wraps every discovered model's display
name with that provider's context, so several providers backed by the
same LiteLLM proxy stay distinguishable in the picker.
Placeholders are {name}, {id}, {provider_id} and {provider_name}. The
default is "{name}", normalized to "unset" so existing cache keys stay
warm. An explicit models.<id>.name still wins and is never templated.
The template is applied once in toConfigModel, so the chat provider and
the Responses sibling inherit it, and it is threaded through cold
discovery, background refresh and the OpenCode 2 provider source. The
template and the resolved provider name are part of the cache identity,
so changing either starts a fresh discovery instead of serving stale
names.
Closes#52
- Re-render Responses-sibling names on OpenCode 1 so `{provider_id}` and
`{provider_name}` resolve to the provider the model is selected through,
without mutating the cached discovery entries OpenCode 2 shares.
- Derive the sibling's generated label from the base provider's name, or
its id when unnamed, so its label and `{provider_name}` agree.
- Use the provider id as the `{provider_name}` fallback when no name is
configured, instead of the shared `LiteLLM (proxy)` label.
- Keep the host's own `modelNameTemplate` setting intact (later
reconciles re-derive naming from it); plugin-level options still never
reach provider settings.
- Widen `formatModelName` to `Pick<LiteLLMModel, 'id'>` and share one
base-name helper between discovery and the sibling rename.
- Tests for the sibling placeholders (named and unnamed base), the v2
provider-settings path, and the plugin-option stripping path.
Rebase of the SSO credential keying (main) onto the sibling placeholder
resolution (#55) crossed exactly here: a sign-in's Responses sibling must
both re-render its models with the sibling's own provider identity and
keep the sign-in runtime fetch with the placeholder key its runtime
package insists on. Extends the existing auth-v1 sibling test to hold
both behaviors at once, so neither side of the merge can regress alone.
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
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.
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Closes #52.
Adds an optional
modelNameTemplatesetting per provider. Every discovered model's display name is built from it, so two providers backed by the same proxy and different keys stop showing identical entries in the picker.Behaviour
{name}(today's formatted name, or the raw id withformatModelNames: false),{id},{provider_id},{provider_name}."{name}", so behaviour is unchanged. It also normalizes to "unset", so existing cache keys stay warm instead of rediscovering on upgrade.models.<id>.namestill wins and is never templated.{id}is inserted literally rather than expanded again.{provider_name}is the provider's configured name, or its id when none is set.<provider>-responsessibling, so{provider_id}/{provider_name}resolve to that sibling for those models (the sibling's generated label is<base name or id> Responses). The cached entries stay as discovered, because OpenCode 2 reads the same cache with a single provider.formatModelNames: the template and the resolved provider name go into the fingerprint, so a template change or a provider rename triggers a fresh discovery instead of serving stale names. Without a template the key is byte-identical to today's.Implementation
applyModelNameTemplate/ModelNaminginsrc/utils/format-model-name.ts;readModelNameTemplatenext toreadFormatModelNames.toConfigModel; the Responses sibling re-renders its own names at injection time without mutating the cache.discoverModels,refreshContextsand the v2ProviderSource.buildCacheKeytakesModelNamingand appends the template + resolved provider name to the fingerprint only when a template is set.runtimeSettingsstrips the plugin-levelmodelNameTemplateoption, so a plugin option never lands in the host's provider settings. A template already present inprovider.settingsis preserved — it is the host's own config and later reconciles re-derive the naming and cache key from it.Tests
test/to-config-model.test.ts: every placeholder, raw-id composition, unknown placeholders, single-pass substitution, provider-name fallback, empty template, reader normalization.test/model-cache.test.ts: key changes on template or provider rename, and stays plain when no template is set.test/plugin-v1.test.ts: templated injection, curated names preserved, cache keyed by the template, sibling placeholders for named and unnamed bases.test/plugin-v2.test.ts: templated model registration fromprovider.settings, curated name preserved, setting preserved, plugin-level option kept out ofinfo.settings,{provider_name}id fallback, cache keyed by the template.npm run typecheckclean;npm test186/186.Summary by CodeRabbit
{name},{id},{provider_id}, and{provider_name}. Templates are applied after existing name formatting; explicitly configured model names remain unchanged.