v0.9.0

🎉 Release v0.9.0

Changes since v0.8.0.

Features

• feat: enable model info discovery by default (0329004)
• feat: enrich LiteLLM discovered models (20dc482)

Fixes

• fix: make release flow branch-protection compatible (b58a8a2)
• fix: skip embedding model discovery (952d18f)

Maintenance

• Merge pull request #4 from yuhp/release/v0.9.0 (0210508)
• Merge pull request #3 from xz-dev/litellm-model-info-discovery (ae3cee5)
• refactor: make model info enrichment provider-specific (0b3ec10)

Installation

npm install opencode-models-discovery@0.9.0
# or
bun add opencode-models-discovery@0.9.0

v0.8.0

Release v0.8.0

Changes since v0.7.0.

Highlights

• Add provider-specific model discovery endpoints
• Support non-/v1 discovery APIs such as DeepSeek /models
• Allow provider-level discovery overrides to bypass default provider detection
• Recommend provider-level discovery configuration for new setups

Features

• Support provider.<name>.options.modelsDiscovery.endpoint
• Keep /v1/models as the default discovery endpoint when no custom endpoint is configured
• Allow discovery when a provider defines an explicit models endpoint
• Allow provider.<name>.options.modelsDiscovery.enabled: true to force discovery even when the provider does not match the default detection signals
• Reduce discovery overhead by using a single request per provider instead of separate probe and fetch requests

Documentation

• Update usage examples to include DeepSeek
• Clarify that provider-level discovery configuration is the recommended approach for new setups
• Document the default discovery endpoint behavior
• Add provider examples for /v1/models and non-/v1 endpoints
• Clarify provider detection versus forced discovery override behavior

Upcoming 1.0 Change

This release does not yet change the current global discovery behavior.
However, v1.0 is planned to introduce a strict provider-first discovery model:

• If any provider explicitly sets provider.<name>.options.modelsDiscovery.enabled
• Discovery enablement will be treated as provider-first
• Users should move discovery enablement decisions from global plugin config into each provider
  Recommended preparation:
• Configure modelsDiscovery.enabled: true only on providers you want to discover
• Configure modelsDiscovery.enabled: false on providers you want to skip
• Treat global discovery config as a compatibility layer during migration

Included Commits

Features

• feat: support provider-specific models endpoints (a4d55c4)
• feat: allow discovery overrides to bypass provider detection (09e6d19)

Documentation

• docs: update discovery configuration examples (f305b01)

Installation

npm install opencode-models-discovery@0.8.0
# or
bun add opencode-models-discovery@0.8.0

v0.7.0

🎉 Release v0.7.0

Changes since v0.6.1.

Features

• feat: add provider-level discovery overrides (8fe6297)

Fixes

• fix: improve release note generation (9344ea8)
• fix: prefer provider model filters over global config (00fee02)
• fix: apply provider discovery overrides consistently (eab00c7)

Documentation

• docs: document provider discovery overrides (4f151f0)
• docs: add repository badges (9033ebd)

Maintenance

• Merge branch 'config-by-provider-options' (3a2482e)

Installation

npm install opencode-models-discovery@0.7.0
# or
bun add opencode-models-discovery@0.7.0

v0.6.1

🎉 Release v0.6.1

Changes

• 75c68e3 chore: bump version to 0.6.1
• ed80c86 chore: align plugin naming and logging
• 523859a chore: bump version to 0.6.0
• 8736901 docs: clarify smart model name behavior
• e61f92f fix: default discovered models to raw ids
• b0473df docs: add project disclaimer
• eef0c70 chore: bump version to 0.5.3
• 62f8f19 chore: update release workflow for trusted publishing
• 50e7ef2 chore: bump version to 0.5.2
• f71e597 fix: use package metadata in release script

Installation

npm install opencode-models-discovery@0.6.1
# or
bun add opencode-models-discovery@0.6.1

Features

• Dynamic Model Discovery: Queries OpenAI-compatible providers through the standard /v1/models endpoint
• Smart Model Formatting: Automatically formats model names for better readability
• Organization Owner Extraction: Extracts and sets organizationOwner field
• Health Check Monitoring: Verifies providers are accessible before discovery
• Flexible Configuration: Supports provider filtering, model filtering, and configurable discovery behavior
• Model Merging: Intelligently merges discovered models with existing configuration
• Comprehensive Caching: Reduces API calls with intelligent caching
• Error Handling: Smart error categorization with auto-fix suggestions

v0.6.0

🎉 Release v0.6.0

Changes

• 523859a chore: bump version to 0.6.0
• 8736901 docs: clarify smart model name behavior
• e61f92f fix: default discovered models to raw ids
• b0473df docs: add project disclaimer
• eef0c70 chore: bump version to 0.5.3
• 62f8f19 chore: update release workflow for trusted publishing
• 50e7ef2 chore: bump version to 0.5.2
• f71e597 fix: use package metadata in release script
• cc4d4a0 chore: bump version to 0.5.1
• c3d7d40 fix: configure git identity for release workflow

Installation

npm install opencode-models-discovery@0.6.0
# or
bun add opencode-models-discovery@0.6.0

Features

• Auto-detection: Automatically detects LM Studio running on common ports
• Dynamic Model Discovery: Queries LM Studio's /v1/models endpoint
• Smart Model Formatting: Automatically formats model names for better readability
• Organization Owner Extraction: Extracts and sets organizationOwner field
• Health Check Monitoring: Verifies LM Studio is accessible
• Automatic Configuration: Auto-creates lmstudio provider if detected
• Model Merging: Intelligently merges discovered models with existing configuration
• Comprehensive Caching: Reduces API calls with intelligent caching
• Error Handling: Smart error categorization with auto-fix suggestions

v0.5.3

🎉 Release v0.5.3

Changes

• eef0c70 chore: bump version to 0.5.3
• 62f8f19 chore: update release workflow for trusted publishing
• 50e7ef2 chore: bump version to 0.5.2
• f71e597 fix: use package metadata in release script
• cc4d4a0 chore: bump version to 0.5.1
• c3d7d40 fix: configure git identity for release workflow
• 7180de9 chore: switch npm release to trusted publishing
• a9ccf72 fix: route plugin logs through structured app logger
• 047df57 merge: release v0.5.0 regex model filtering
• ab5c42c feat: add regex model filtering

Installation

npm install opencode-models-discovery@0.5.3
# or
bun add opencode-models-discovery@0.5.3

Features

• Auto-detection: Automatically detects LM Studio running on common ports
• Dynamic Model Discovery: Queries LM Studio's /v1/models endpoint
• Smart Model Formatting: Automatically formats model names for better readability
• Organization Owner Extraction: Extracts and sets organizationOwner field
• Health Check Monitoring: Verifies LM Studio is accessible
• Automatic Configuration: Auto-creates lmstudio provider if detected
• Model Merging: Intelligently merges discovered models with existing configuration
• Comprehensive Caching: Reduces API calls with intelligent caching
• Error Handling: Smart error categorization with auto-fix suggestions

v0.5.2

🎉 Release v0.5.2

Changes

• 50e7ef2 chore: bump version to 0.5.2
• f71e597 fix: use package metadata in release script
• cc4d4a0 chore: bump version to 0.5.1
• c3d7d40 fix: configure git identity for release workflow
• 7180de9 chore: switch npm release to trusted publishing
• a9ccf72 fix: route plugin logs through structured app logger
• 047df57 merge: release v0.5.0 regex model filtering
• ab5c42c feat: add regex model filtering
• 44b381e chore: remove .tgz packaging artifact from repository
• 7ee78bb docs: add npm badges and update provider table with CLIProxyAPI

Installation

npm install opencode-models-discovery@0.5.2
# or
bun add opencode-models-discovery@0.5.2

Features

• Auto-detection: Automatically detects LM Studio running on common ports
• Dynamic Model Discovery: Queries LM Studio's /v1/models endpoint
• Smart Model Formatting: Automatically formats model names for better readability
• Organization Owner Extraction: Extracts and sets organizationOwner field
• Health Check Monitoring: Verifies LM Studio is accessible
• Automatic Configuration: Auto-creates lmstudio provider if detected
• Model Merging: Intelligently merges discovered models with existing configuration
• Comprehensive Caching: Reduces API calls with intelligent caching
• Error Handling: Smart error categorization with auto-fix suggestions

v0.5.1

🎉 Release v0.5.1

Changes

• cc4d4a0 chore: bump version to 0.5.1
• c3d7d40 fix: configure git identity for release workflow
• 7180de9 chore: switch npm release to trusted publishing
• a9ccf72 fix: route plugin logs through structured app logger
• 047df57 merge: release v0.5.0 regex model filtering
• ab5c42c feat: add regex model filtering
• 44b381e chore: remove .tgz packaging artifact from repository
• 7ee78bb docs: add npm badges and update provider table with CLIProxyAPI
• ab88214 chore: rename package to opencode-models-discovery (plural)
• a98c36a docs: remove unimplemented auto-detection feature from README

Installation

npm install opencode-lmstudio@0.5.1
# or
bun add opencode-lmstudio@0.5.1

Features

• Auto-detection: Automatically detects LM Studio running on common ports
• Dynamic Model Discovery: Queries LM Studio's /v1/models endpoint
• Smart Model Formatting: Automatically formats model names for better readability
• Organization Owner Extraction: Extracts and sets organizationOwner field
• Health Check Monitoring: Verifies LM Studio is accessible
• Automatic Configuration: Auto-creates lmstudio provider if detected
• Model Merging: Intelligently merges discovered models with existing configuration
• Comprehensive Caching: Reduces API calls with intelligent caching
• Error Handling: Smart error categorization with auto-fix suggestions

v0.5.0 Release

Adds regex-based model filtering for automatic model discovery, so only the models you want are injected into provider configuration.

Highlights

• Added models.includeRegex
  • When non-empty, only discovered model IDs matching at least one regex are auto-injected
• Added models.excludeRegex
  • When includeRegex is empty, discovered model IDs matching any regex are skipped
• Preserves explicit user configuration
  • Filtering only applies to auto-discovered models
  • Models already defined in provider.models are not removed or overwritten
• Invalid regex handling
  • Invalid patterns are ignored with a warning instead of breaking plugin startup
• Documentation and tests updated
  • Added coverage for include/exclude behavior and explicit model preservation

Configuration

{
  "plugin": [
    ["opencode-models-discovery", {
      "models": {
        "includeRegex": ["^qwen/", "gpt-4"],
        "excludeRegex": ["embedding", "test"]
      }
    }]
  ]
}

Filtering Priority

• If includeRegex is non-empty, excludeRegex is ignored
• If includeRegex is empty, excludeRegex is applied
• If both are empty, no model filtering is applied

Examples

Keep only selected models:

{
  "models": {
    "includeRegex": ["^qwen/", "^deepseek/"]
  }
}

Exclude unwanted models:

{
  "models": {
    "excludeRegex": ["embedding", "rerank", "test"]
  }
}

Notes

• Regex patterns are configured as strings and compiled with new RegExp(pattern)
• Regex flags such as /.../i are not supported in this release
• Flag support can be added in a future version if needed

Validation

• TypeScript typecheck passed
• Test suite passed with 21 tests

v0.4.9 Release

Breaking Changes

• None

New Features

• Universal OpenAI-Compatible Provider Support: Plugin now auto-discovers models from ANY provider with OpenAI-compatible endpoints, not just LM Studio
• Provider Detection: Supports both strict detection (npm === "@ai-sdk/openai-compatible") and URL-based detection (baseURL contains /v1/)
• Provider Filtering: Configure which providers to discover with providers.include and providers.exclude
• Configurable Discovery TTL: Set cache TTL via discovery.ttl (default 15000ms)
• API Key Support: Health checks and model discovery now support API key authentication

Refactoring

• Removed Validation Feature: Plugin now focuses exclusively on automatic model discovery. Validation (checking if a model is loaded/available) is handled by the AI SDK or provider
• Simplified PluginConfig: Removed validation-related configuration fields
• Moved Type Guards: canDiscoverModels, isOpenAICompatibleProvider, etc. moved to openai-compatible-api.ts

Documentation

• Added original author credit (forked from opencode-lmstudio)
• Clarified plugin focuses on discovery only, not validation
• Added engines.opencode: ">=1.4.0" requirement