Feature: Multi-LLM Provider Support

[bossfelfo]

Description

Refactors the application to support multiple LLM providers (OpenAI, Anthropic, Gemini) and removes the .env requirement by migrating to client-side API key management.

Changes

• Abstracts LLM Providers: Implemented a generic LLMProvider interface (lib/llm/types.ts) with distinct strategy adapters for OpenAI, Anthropic, and Gemini.
• Client-Side Key Management: Removed .env.example and hardcoded key references. API keys are now collected in the UI, stored in localStorage, and passed dynamically in the request payload.
• Error Normalization: Added normalizeProviderError to intercept and safely extract human-readable messages from deeply nested JSON payloads across all three vendor SDKs before displaying them in UI Toasts.
• Gemini Stabilisation:
  • Bound the provider to the explicitly defined v1beta endpoint to support the 2.x models.
  • Stripped out the systemInstruction config field (which is notoriously unstable across versions) and manually prepends system instructions to user messages to guarantee cross-model compatibility.
• Model Registry Updates: Bumped default models to current standards (gpt-4o-mini, claude-3-5-sonnet-latest, gemini-2.5-flash).

Developer Evidence

Testing Notes

• Verified end-to-end classification and generation flows for all three providers.
• Confirmed localStorage properly encrypts/stores/clears keys from the settings UI.
• Validated that raw SDK errors (e.g., Anthropic "Insufficient Credits") strictly map to readable frontend toasts without leaking raw HTTP payloads.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
assertify-io	Ready	Preview, Comment	Mar 14, 2026 0:10am
assertify.io	Ready	Preview, Comment	Mar 14, 2026 0:10am

[Copilot]

Pull request overview

Refactors the app’s LLM integration to support multiple providers (OpenAI/Anthropic/Gemini), shifts API key entry to the UI (persisted client-side), and normalizes provider errors surfaced to the frontend.

Changes:

• Introduces an LLMProvider abstraction with provider adapters (OpenAI, Anthropic, Gemini) and shared error normalization.
• Adds provider/model selection to settings + request payloads; updates API routes to instantiate the selected provider.
• Removes .env-based API key setup and updates docs accordingly.

Reviewed changes

Copilot reviewed 16 out of 17 changed files in this pull request and generated 9 comments.

Show a summary per file

File	Description
package.json	Adds Anthropic/Gemini SDK deps; updates Next + eslint config versions.
package-lock.json	Lockfile updates for new deps and version resolution changes.
lib/settings.ts	Extends persisted settings with provider + model, with sanitization defaults.
lib/llm/types.ts	Adds provider types/metadata registry and common interfaces.
lib/llm/index.ts	Exposes provider factory + llm exports.
lib/llm/openai.ts	OpenAI provider adapter.
lib/llm/anthropic.ts	Anthropic provider adapter.
lib/llm/gemini.ts	Gemini provider adapter with v1beta binding and system-prompt handling.
lib/llm/errors.ts	Adds cross-provider error normalization for UI-safe messages.
app/page.tsx	Adds provider/model selector + per-provider API key storage and request payload updates.
app/settings/page.tsx	Adds provider and model selection UI.
app/questions/page.tsx	Uses per-provider API key storage + includes provider/model in API requests.
app/api/classify/route.ts	Uses provider factory + normalized errors.
app/api/generate/route.ts	Uses provider factory + normalized errors.
app/api/generate-questions/route.ts	Uses provider factory + normalized errors.
README.md	Updates setup instructions to UI-based key entry and multi-LLM support.
.env.example	Removed env-based API key workflow.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[ShirleyRex]

Looks good