feat: prompt for LLM provider and key instead of Coval proxy#7
Conversation
Remove the /v1/wizard/complete proxy endpoint (which doesn't exist) and instead prompt users to select a provider (Anthropic/OpenAI/Gemini) and paste their own API key. WIZARD_LLM_KEY/WIZARD_LLM_PROVIDER env vars still bypass the prompt for CI/power users.
|
Caution Review failedPull request was closed or merged during review WalkthroughThe changes refactor LLM configuration from environment-variable-based routing to explicit llmConfig objects. A new getLLMConfig() function prompts for provider and API key selection when WIZARD_LLM_KEY is absent, returning a structured LLMConfig type. callWizardLLM signature updated to accept llmConfig instead of apiKey. Provider adapters (Anthropic, OpenAI, Gemini) now consume LLMConfig directly. The Coval proxy fallback path removed. Tests converted to use explicit configuration objects, and the public COVAL_WIZARD_ENDPOINT constant eliminated. 🚥 Pre-merge checks | ✅ 1✅ Passed checks (1 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches
🧪 Generate unit tests (beta)
Comment |
Summary
POST /v1/wizard/complete(Coval proxy endpoint that doesn't exist)WIZARD_LLM_KEY+WIZARD_LLM_PROVIDERenv vars still bypass the prompt for CI/scripted usellmConfigexplicitly instead of routing through env vars (72 tests, all passing)New flow
Summary
Breaking Changes:
callWizardLLMnow acceptsllmConfig: LLMConfiginstead ofapiKeyparameterCOVAL_WIZARD_ENDPOINTconstant and Coval proxy endpoint dependencyFeatures:
WIZARD_LLM_KEYandWIZARD_LLM_PROVIDERenvironment variables for CI/scripted useRefactors:
LLMConfigexported type across all providersllmConfigobjects instead of environment variables (72 tests updated)