Feat/add litellm provider

[RheagalFire]

Summary

• Adds LiteLLM as a first-class provider in Maka, giving users access to 100+ LLM providers (AWS Bedrock, Azure, Vertex AI, Cohere, Replicate, etc.) through a single OpenAI-compatible proxy endpoint.
• Includes model factory integration, provider defaults, UI registration, and comprehensive tests.
• Additive only - no existing providers modified.

Changes

Core (packages/core/src/llm-connections.ts):

• Added 'litellm' to ProviderType union
• Added PROVIDER_DEFAULTS entry: protocol 'openai', category 'custom', base URL http://localhost:4000/v1, auth 'api_key', badge 'Gateway'
• Added 'litellm' to CATALOG_PROVIDER_TYPES array

Runtime (packages/runtime/src/model-factory.ts):

• Added case 'litellm' in getAIModel() using createOpenAICompatible() from @ai-sdk/openai-compatible with name 'litellm', apiKey, and baseURL fallback

UI (apps/desktop/src/renderer/settings/):

• provider-brand-marks.tsx - Added 'litellm' to the OpenAI brand mark case
• provider-display.tsx - Added display name "LiteLLM" with description and "Gateway" badge

UI (packages/ui/src/chat-model-helpers.ts):

• Added litellm: 'LiteLLM' to exhaustive PROVIDER_SHORT_LABEL map

Tests (packages/core/src/__tests__/llm-connections.test.ts, +26 lines):

• PROVIDER_DEFAULTS['litellm'] correctness: label, protocol, category, baseUrl, authKind, backendKind, status, catalogBadge, empty fallbackModels
• persistedBaseUrl('litellm', ...) collapses default URL to undefined, persists custom overrides

Tests (packages/runtime/src/__tests__/model-factory-thinking.test.ts, +43 lines):

• getAIModel with litellm: correct provider namespace 'litellm.chat' and modelId, default and custom baseUrl
• buildProviderOptions with litellm: falls through to default case, returns empty {} for all thinking levels

Tests

1. All 2293 tests pass (446 suites):

# tests 2293
# suites 446
# pass 2293
# fail 0
# duration_ms 9982

2. TypeScript build: clean across all 7 workspaces.

Example usage

import{typeLlmConnection}from'@maka/core';import{getAIModel}from'@maka/runtime';// Create a LiteLLM connectionconstconnection: LlmConnection={slug: 'my-litellm-proxy',name: 'LiteLLM Proxy',providerType: 'litellm',baseUrl: 'http://localhost:4000/v1',// default, pre-filled in UIdefaultModel: 'anthropic/claude-sonnet-4-6',enabled: true,createdAt: Date.now(),updatedAt: Date.now(),};// Get AI model - uses @ai-sdk/openai-compatible under the hoodconstmodel=getAIModel({
connection,apiKey: 'sk-your-litellm-key',modelId: 'anthropic/claude-sonnet-4-6',});

Risk / Compatibility

• Additive only. No existing provider types, model factory cases, or UI components modified.
• No new dependencies. LiteLLM proxy speaks the OpenAI protocol, so the existing @ai-sdk/openai-compatible package handles everything.
• Protocol: 'openai'. Model discovery automatically uses the standard /v1/models endpoint.
• Type-safe.satisfies Record<ProviderType, string> compile-time check ensures the new provider is registered in every exhaustive map.

[RheagalFire]

Summary

• Adds LiteLLM as a first-class provider in Maka, giving users access to 100+ LLM providers (AWS Bedrock, Azure, Vertex AI, Cohere, Replicate, etc.) through a single OpenAI-compatible proxy endpoint.
• Includes model factory integration, provider defaults, UI registration, and comprehensive tests.
• Additive only - no existing providers modified.

Changes

Core (packages/core/src/llm-connections.ts):

• Added 'litellm' to ProviderType union
• Added PROVIDER_DEFAULTS entry: protocol 'openai', category 'custom', base URL http://localhost:4000/v1, auth 'api_key', badge 'Gateway'
• Added 'litellm' to CATALOG_PROVIDER_TYPES array

Runtime (packages/runtime/src/model-factory.ts):

• Added case 'litellm' in getAIModel() using createOpenAICompatible() from @ai-sdk/openai-compatible with name 'litellm', apiKey, and baseURL fallback

UI (apps/desktop/src/renderer/settings/):

• provider-brand-marks.tsx - Added 'litellm' to the OpenAI brand mark case
• provider-display.tsx - Added display name "LiteLLM" with description and "Gateway" badge

UI (packages/ui/src/chat-model-helpers.ts):

• Added litellm: 'LiteLLM' to exhaustive PROVIDER_SHORT_LABEL map

Tests (packages/core/src/__tests__/llm-connections.test.ts, +26 lines):

• PROVIDER_DEFAULTS['litellm'] correctness: label, protocol, category, baseUrl, authKind, backendKind, status, catalogBadge, empty fallbackModels
• persistedBaseUrl('litellm', ...) collapses default URL to undefined, persists custom overrides

Tests (packages/runtime/src/__tests__/model-factory-thinking.test.ts, +43 lines):

• getAIModel with litellm: correct provider namespace 'litellm.chat' and modelId, default and custom baseUrl
• buildProviderOptions with litellm: falls through to default case, returns empty {} for all thinking levels

Tests

1. All 2293 tests pass (446 suites):

# tests 2293
# suites 446
# pass 2293
# fail 0
# duration_ms 9982

2. TypeScript build: clean across all 7 workspaces.

Example usage

import{typeLlmConnection}from'@maka/core';import{getAIModel}from'@maka/runtime';// Create a LiteLLM connectionconstconnection: LlmConnection={slug: 'my-litellm-proxy',name: 'LiteLLM Proxy',providerType: 'litellm',baseUrl: 'http://localhost:4000/v1',// default, pre-filled in UIdefaultModel: 'anthropic/claude-sonnet-4-6',enabled: true,createdAt: Date.now(),updatedAt: Date.now(),};// Get AI model - uses @ai-sdk/openai-compatible under the hoodconstmodel=getAIModel({
connection,apiKey: 'sk-your-litellm-key',modelId: 'anthropic/claude-sonnet-4-6',});

Risk / Compatibility

• Additive only. No existing provider types, model factory cases, or UI components modified.
• No new dependencies. LiteLLM proxy speaks the OpenAI protocol, so the existing @ai-sdk/openai-compatible package handles everything.
• Protocol: 'openai'. Model discovery automatically uses the standard /v1/models endpoint.
• Type-safe.satisfies Record<ProviderType, string> compile-time check ensures the new provider is registered in every exhaustive map.

[RheagalFire]

Summary

• Adds LiteLLM as a first-class provider in Maka, giving users access to 100+ LLM providers (AWS Bedrock, Azure, Vertex AI, Cohere, Replicate, etc.) through a single OpenAI-compatible proxy endpoint.
• Includes model factory integration, provider defaults, UI registration, and comprehensive tests.
• Additive only - no existing providers modified.

Changes

Core (packages/core/src/llm-connections.ts):

• Added 'litellm' to ProviderType union
• Added PROVIDER_DEFAULTS entry: protocol 'openai', category 'custom', base URL http://localhost:4000/v1, auth 'api_key', badge 'Gateway'
• Added 'litellm' to CATALOG_PROVIDER_TYPES array

Runtime (packages/runtime/src/model-factory.ts):

• Added case 'litellm' in getAIModel() using createOpenAICompatible() from @ai-sdk/openai-compatible with name 'litellm', apiKey, and baseURL fallback

UI (apps/desktop/src/renderer/settings/):

• provider-brand-marks.tsx - Added 'litellm' to the OpenAI brand mark case
• provider-display.tsx - Added display name "LiteLLM" with description and "Gateway" badge

UI (packages/ui/src/chat-model-helpers.ts):

• Added litellm: 'LiteLLM' to exhaustive PROVIDER_SHORT_LABEL map

Tests (packages/core/src/__tests__/llm-connections.test.ts, +26 lines):

• PROVIDER_DEFAULTS['litellm'] correctness: label, protocol, category, baseUrl, authKind, backendKind, status, catalogBadge, empty fallbackModels
• persistedBaseUrl('litellm', ...) collapses default URL to undefined, persists custom overrides

Tests (packages/runtime/src/__tests__/model-factory-thinking.test.ts, +43 lines):

• getAIModel with litellm: correct provider namespace 'litellm.chat' and modelId, default and custom baseUrl
• buildProviderOptions with litellm: falls through to default case, returns empty {} for all thinking levels

Tests

1. All 2293 tests pass (446 suites):

# tests 2293
# suites 446
# pass 2293
# fail 0
# duration_ms 9982

2. TypeScript build: clean across all 7 workspaces.

Example usage

import{typeLlmConnection}from'@maka/core';import{getAIModel}from'@maka/runtime';// Create a LiteLLM connectionconstconnection: LlmConnection={slug: 'my-litellm-proxy',name: 'LiteLLM Proxy',providerType: 'litellm',baseUrl: 'http://localhost:4000/v1',// default, pre-filled in UIdefaultModel: 'anthropic/claude-sonnet-4-6',enabled: true,createdAt: Date.now(),updatedAt: Date.now(),};// Get AI model - uses @ai-sdk/openai-compatible under the hoodconstmodel=getAIModel({
connection,apiKey: 'sk-your-litellm-key',modelId: 'anthropic/claude-sonnet-4-6',});

Risk / Compatibility

• Additive only. No existing provider types, model factory cases, or UI components modified.
• No new dependencies. LiteLLM proxy speaks the OpenAI protocol, so the existing @ai-sdk/openai-compatible package handles everything.
• Protocol: 'openai'. Model discovery automatically uses the standard /v1/models endpoint.
• Type-safe.satisfies Record<ProviderType, string> compile-time check ensures the new provider is registered in every exhaustive map.

[RheagalFire]

Summary

• Adds LiteLLM as a first-class provider in Maka, giving users access to 100+ LLM providers (AWS Bedrock, Azure, Vertex AI, Cohere, Replicate, etc.) through a single OpenAI-compatible proxy endpoint.
• Includes model factory integration, provider defaults, UI registration, and comprehensive tests.
• Additive only - no existing providers modified.

Changes

Core (packages/core/src/llm-connections.ts):

• Added 'litellm' to ProviderType union
• Added PROVIDER_DEFAULTS entry: protocol 'openai', category 'custom', base URL http://localhost:4000/v1, auth 'api_key', badge 'Gateway'
• Added 'litellm' to CATALOG_PROVIDER_TYPES array

Runtime (packages/runtime/src/model-factory.ts):

• Added case 'litellm' in getAIModel() using createOpenAICompatible() from @ai-sdk/openai-compatible with name 'litellm', apiKey, and baseURL fallback

UI (apps/desktop/src/renderer/settings/):

• provider-brand-marks.tsx - Added 'litellm' to the OpenAI brand mark case
• provider-display.tsx - Added display name "LiteLLM" with description and "Gateway" badge

UI (packages/ui/src/chat-model-helpers.ts):

• Added litellm: 'LiteLLM' to exhaustive PROVIDER_SHORT_LABEL map

Tests (packages/core/src/__tests__/llm-connections.test.ts, +26 lines):

• PROVIDER_DEFAULTS['litellm'] correctness: label, protocol, category, baseUrl, authKind, backendKind, status, catalogBadge, empty fallbackModels
• persistedBaseUrl('litellm', ...) collapses default URL to undefined, persists custom overrides

Tests (packages/runtime/src/__tests__/model-factory-thinking.test.ts, +43 lines):

• getAIModel with litellm: correct provider namespace 'litellm.chat' and modelId, default and custom baseUrl
• buildProviderOptions with litellm: falls through to default case, returns empty {} for all thinking levels

Tests

1. All 2293 tests pass (446 suites):

# tests 2293
# suites 446
# pass 2293
# fail 0
# duration_ms 9982

2. TypeScript build: clean across all 7 workspaces.

Example usage

import{typeLlmConnection}from'@maka/core';import{getAIModel}from'@maka/runtime';// Create a LiteLLM connectionconstconnection: LlmConnection={slug: 'my-litellm-proxy',name: 'LiteLLM Proxy',providerType: 'litellm',baseUrl: 'http://localhost:4000/v1',// default, pre-filled in UIdefaultModel: 'anthropic/claude-sonnet-4-6',enabled: true,createdAt: Date.now(),updatedAt: Date.now(),};// Get AI model - uses @ai-sdk/openai-compatible under the hoodconstmodel=getAIModel({
connection,apiKey: 'sk-your-litellm-key',modelId: 'anthropic/claude-sonnet-4-6',});

Risk / Compatibility

• Additive only. No existing provider types, model factory cases, or UI components modified.
• No new dependencies. LiteLLM proxy speaks the OpenAI protocol, so the existing @ai-sdk/openai-compatible package handles everything.
• Protocol: 'openai'. Model discovery automatically uses the standard /v1/models endpoint.
• Type-safe.satisfies Record<ProviderType, string> compile-time check ensures the new provider is registered in every exhaustive map.

[RheagalFire]

Summary

• Adds LiteLLM as a first-class provider in Maka, giving users access to 100+ LLM providers (AWS Bedrock, Azure, Vertex AI, Cohere, Replicate, etc.) through a single OpenAI-compatible proxy endpoint.
• Includes model factory integration, provider defaults, UI registration, and comprehensive tests.
• Additive only - no existing providers modified.

Changes

Core (packages/core/src/llm-connections.ts):

• Added 'litellm' to ProviderType union
• Added PROVIDER_DEFAULTS entry: protocol 'openai', category 'custom', base URL http://localhost:4000/v1, auth 'api_key', badge 'Gateway'
• Added 'litellm' to CATALOG_PROVIDER_TYPES array

Runtime (packages/runtime/src/model-factory.ts):

• Added case 'litellm' in getAIModel() using createOpenAICompatible() from @ai-sdk/openai-compatible with name 'litellm', apiKey, and baseURL fallback

UI (apps/desktop/src/renderer/settings/):

• provider-brand-marks.tsx - Added 'litellm' to the OpenAI brand mark case
• provider-display.tsx - Added display name "LiteLLM" with description and "Gateway" badge

UI (packages/ui/src/chat-model-helpers.ts):

• Added litellm: 'LiteLLM' to exhaustive PROVIDER_SHORT_LABEL map

Tests (packages/core/src/__tests__/llm-connections.test.ts, +26 lines):

• PROVIDER_DEFAULTS['litellm'] correctness: label, protocol, category, baseUrl, authKind, backendKind, status, catalogBadge, empty fallbackModels
• persistedBaseUrl('litellm', ...) collapses default URL to undefined, persists custom overrides

Tests (packages/runtime/src/__tests__/model-factory-thinking.test.ts, +43 lines):

• getAIModel with litellm: correct provider namespace 'litellm.chat' and modelId, default and custom baseUrl
• buildProviderOptions with litellm: falls through to default case, returns empty {} for all thinking levels

Tests

1. All 2293 tests pass (446 suites):

# tests 2293
# suites 446
# pass 2293
# fail 0
# duration_ms 9982

2. TypeScript build: clean across all 7 workspaces.

Example usage

import{typeLlmConnection}from'@maka/core';import{getAIModel}from'@maka/runtime';// Create a LiteLLM connectionconstconnection: LlmConnection={slug: 'my-litellm-proxy',name: 'LiteLLM Proxy',providerType: 'litellm',baseUrl: 'http://localhost:4000/v1',// default, pre-filled in UIdefaultModel: 'anthropic/claude-sonnet-4-6',enabled: true,createdAt: Date.now(),updatedAt: Date.now(),};// Get AI model - uses @ai-sdk/openai-compatible under the hoodconstmodel=getAIModel({
connection,apiKey: 'sk-your-litellm-key',modelId: 'anthropic/claude-sonnet-4-6',});

Risk / Compatibility

• Additive only. No existing provider types, model factory cases, or UI components modified.
• No new dependencies. LiteLLM proxy speaks the OpenAI protocol, so the existing @ai-sdk/openai-compatible package handles everything.
• Protocol: 'openai'. Model discovery automatically uses the standard /v1/models endpoint.
• Type-safe.satisfies Record<ProviderType, string> compile-time check ensures the new provider is registered in every exhaustive map.

[RheagalFire]

Summary

• Adds LiteLLM as a first-class provider in Maka, giving users access to 100+ LLM providers (AWS Bedrock, Azure, Vertex AI, Cohere, Replicate, etc.) through a single OpenAI-compatible proxy endpoint.
• Includes model factory integration, provider defaults, UI registration, and comprehensive tests.
• Additive only - no existing providers modified.

Changes

Core (packages/core/src/llm-connections.ts):

• Added 'litellm' to ProviderType union
• Added PROVIDER_DEFAULTS entry: protocol 'openai', category 'custom', base URL http://localhost:4000/v1, auth 'api_key', badge 'Gateway'
• Added 'litellm' to CATALOG_PROVIDER_TYPES array

Runtime (packages/runtime/src/model-factory.ts):

• Added case 'litellm' in getAIModel() using createOpenAICompatible() from @ai-sdk/openai-compatible with name 'litellm', apiKey, and baseURL fallback

UI (apps/desktop/src/renderer/settings/):

• provider-brand-marks.tsx - Added 'litellm' to the OpenAI brand mark case
• provider-display.tsx - Added display name "LiteLLM" with description and "Gateway" badge

UI (packages/ui/src/chat-model-helpers.ts):

• Added litellm: 'LiteLLM' to exhaustive PROVIDER_SHORT_LABEL map

Tests (packages/core/src/__tests__/llm-connections.test.ts, +26 lines):

• PROVIDER_DEFAULTS['litellm'] correctness: label, protocol, category, baseUrl, authKind, backendKind, status, catalogBadge, empty fallbackModels
• persistedBaseUrl('litellm', ...) collapses default URL to undefined, persists custom overrides

Tests (packages/runtime/src/__tests__/model-factory-thinking.test.ts, +43 lines):

• getAIModel with litellm: correct provider namespace 'litellm.chat' and modelId, default and custom baseUrl
• buildProviderOptions with litellm: falls through to default case, returns empty {} for all thinking levels

Tests

1. All 2293 tests pass (446 suites):

# tests 2293
# suites 446
# pass 2293
# fail 0
# duration_ms 9982

2. TypeScript build: clean across all 7 workspaces.

Example usage

import{typeLlmConnection}from'@maka/core';import{getAIModel}from'@maka/runtime';// Create a LiteLLM connectionconstconnection: LlmConnection={slug: 'my-litellm-proxy',name: 'LiteLLM Proxy',providerType: 'litellm',baseUrl: 'http://localhost:4000/v1',// default, pre-filled in UIdefaultModel: 'anthropic/claude-sonnet-4-6',enabled: true,createdAt: Date.now(),updatedAt: Date.now(),};// Get AI model - uses @ai-sdk/openai-compatible under the hoodconstmodel=getAIModel({
connection,apiKey: 'sk-your-litellm-key',modelId: 'anthropic/claude-sonnet-4-6',});

Risk / Compatibility

• Additive only. No existing provider types, model factory cases, or UI components modified.
• No new dependencies. LiteLLM proxy speaks the OpenAI protocol, so the existing @ai-sdk/openai-compatible package handles everything.
• Protocol: 'openai'. Model discovery automatically uses the standard /v1/models endpoint.
• Type-safe.satisfies Record<ProviderType, string> compile-time check ensures the new provider is registered in every exhaustive map.

[RheagalFire]

Summary

• Adds LiteLLM as a first-class provider in Maka, giving users access to 100+ LLM providers (AWS Bedrock, Azure, Vertex AI, Cohere, Replicate, etc.) through a single OpenAI-compatible proxy endpoint.
• Includes model factory integration, provider defaults, UI registration, and comprehensive tests.
• Additive only - no existing providers modified.

Changes

Core (packages/core/src/llm-connections.ts):

• Added 'litellm' to ProviderType union
• Added PROVIDER_DEFAULTS entry: protocol 'openai', category 'custom', base URL http://localhost:4000/v1, auth 'api_key', badge 'Gateway'
• Added 'litellm' to CATALOG_PROVIDER_TYPES array

Runtime (packages/runtime/src/model-factory.ts):

• Added case 'litellm' in getAIModel() using createOpenAICompatible() from @ai-sdk/openai-compatible with name 'litellm', apiKey, and baseURL fallback

UI (apps/desktop/src/renderer/settings/):

• provider-brand-marks.tsx - Added 'litellm' to the OpenAI brand mark case
• provider-display.tsx - Added display name "LiteLLM" with description and "Gateway" badge

UI (packages/ui/src/chat-model-helpers.ts):

• Added litellm: 'LiteLLM' to exhaustive PROVIDER_SHORT_LABEL map

Tests (packages/core/src/__tests__/llm-connections.test.ts, +26 lines):

• PROVIDER_DEFAULTS['litellm'] correctness: label, protocol, category, baseUrl, authKind, backendKind, status, catalogBadge, empty fallbackModels
• persistedBaseUrl('litellm', ...) collapses default URL to undefined, persists custom overrides

Tests (packages/runtime/src/__tests__/model-factory-thinking.test.ts, +43 lines):

• getAIModel with litellm: correct provider namespace 'litellm.chat' and modelId, default and custom baseUrl
• buildProviderOptions with litellm: falls through to default case, returns empty {} for all thinking levels

Tests

1. All 2293 tests pass (446 suites):

# tests 2293
# suites 446
# pass 2293
# fail 0
# duration_ms 9982

2. TypeScript build: clean across all 7 workspaces.

Example usage

import{typeLlmConnection}from'@maka/core';import{getAIModel}from'@maka/runtime';// Create a LiteLLM connectionconstconnection: LlmConnection={slug: 'my-litellm-proxy',name: 'LiteLLM Proxy',providerType: 'litellm',baseUrl: 'http://localhost:4000/v1',// default, pre-filled in UIdefaultModel: 'anthropic/claude-sonnet-4-6',enabled: true,createdAt: Date.now(),updatedAt: Date.now(),};// Get AI model - uses @ai-sdk/openai-compatible under the hoodconstmodel=getAIModel({
connection,apiKey: 'sk-your-litellm-key',modelId: 'anthropic/claude-sonnet-4-6',});

Risk / Compatibility

• Additive only. No existing provider types, model factory cases, or UI components modified.
• No new dependencies. LiteLLM proxy speaks the OpenAI protocol, so the existing @ai-sdk/openai-compatible package handles everything.
• Protocol: 'openai'. Model discovery automatically uses the standard /v1/models endpoint.
• Type-safe.satisfies Record<ProviderType, string> compile-time check ensures the new provider is registered in every exhaustive map.

[RheagalFire]

Summary

• Adds LiteLLM as a first-class provider in Maka, giving users access to 100+ LLM providers (AWS Bedrock, Azure, Vertex AI, Cohere, Replicate, etc.) through a single OpenAI-compatible proxy endpoint.
• Includes model factory integration, provider defaults, UI registration, and comprehensive tests.
• Additive only - no existing providers modified.

Changes

Core (packages/core/src/llm-connections.ts):

• Added 'litellm' to ProviderType union
• Added PROVIDER_DEFAULTS entry: protocol 'openai', category 'custom', base URL http://localhost:4000/v1, auth 'api_key', badge 'Gateway'
• Added 'litellm' to CATALOG_PROVIDER_TYPES array

Runtime (packages/runtime/src/model-factory.ts):

• Added case 'litellm' in getAIModel() using createOpenAICompatible() from @ai-sdk/openai-compatible with name 'litellm', apiKey, and baseURL fallback

UI (apps/desktop/src/renderer/settings/):

• provider-brand-marks.tsx - Added 'litellm' to the OpenAI brand mark case
• provider-display.tsx - Added display name "LiteLLM" with description and "Gateway" badge

UI (packages/ui/src/chat-model-helpers.ts):

• Added litellm: 'LiteLLM' to exhaustive PROVIDER_SHORT_LABEL map

Tests (packages/core/src/__tests__/llm-connections.test.ts, +26 lines):

• PROVIDER_DEFAULTS['litellm'] correctness: label, protocol, category, baseUrl, authKind, backendKind, status, catalogBadge, empty fallbackModels
• persistedBaseUrl('litellm', ...) collapses default URL to undefined, persists custom overrides

Tests (packages/runtime/src/__tests__/model-factory-thinking.test.ts, +43 lines):

• getAIModel with litellm: correct provider namespace 'litellm.chat' and modelId, default and custom baseUrl
• buildProviderOptions with litellm: falls through to default case, returns empty {} for all thinking levels

Tests

1. All 2293 tests pass (446 suites):

# tests 2293
# suites 446
# pass 2293
# fail 0
# duration_ms 9982

2. TypeScript build: clean across all 7 workspaces.

Example usage

import{typeLlmConnection}from'@maka/core';import{getAIModel}from'@maka/runtime';// Create a LiteLLM connectionconstconnection: LlmConnection={slug: 'my-litellm-proxy',name: 'LiteLLM Proxy',providerType: 'litellm',baseUrl: 'http://localhost:4000/v1',// default, pre-filled in UIdefaultModel: 'anthropic/claude-sonnet-4-6',enabled: true,createdAt: Date.now(),updatedAt: Date.now(),};// Get AI model - uses @ai-sdk/openai-compatible under the hoodconstmodel=getAIModel({
connection,apiKey: 'sk-your-litellm-key',modelId: 'anthropic/claude-sonnet-4-6',});

Risk / Compatibility

• Additive only. No existing provider types, model factory cases, or UI components modified.
• No new dependencies. LiteLLM proxy speaks the OpenAI protocol, so the existing @ai-sdk/openai-compatible package handles everything.
• Protocol: 'openai'. Model discovery automatically uses the standard /v1/models endpoint.
• Type-safe.satisfies Record<ProviderType, string> compile-time check ensures the new provider is registered in every exhaustive map.