feat: add multi-provider AI support (Groq, OpenAI, Anthropic, Google)

[dhairyashiil]

feat: add multi-provider AI support (Groq, OpenAI, Anthropic, Google)

Summary

Replaces the hardcoded Groq provider in apps/chat with a dynamic, env-driven provider selector. Two env vars now control AI behavior:

• AI_PROVIDER — groq | openai | anthropic | google (default: groq)
• AI_MODEL — optional model override (each provider has a sensible default)

The Vercel AI SDK's provider-agnostic LanguageModel interface means only lib/ai-provider.ts needs to know which provider is active — the rest of the codebase (streamText, tools, etc.) is unchanged.

Files changed (6):

File	What changed
apps/chat/package.json	Added @ai-sdk/anthropic, @ai-sdk/google
apps/chat/lib/ai-provider.ts	Rewrote with PROVIDER_CONFIG registry + getModel() exhaustive switch
apps/chat/lib/agent.ts	Broadened isAIToolCallError / isAIRateLimitError for all providers
apps/chat/lib/env.ts	Validates AI_PROVIDER value and that the selected provider's API key is present
apps/chat/.env.example	Updated AI/LLM section with all provider options
apps/chat/README.md	Updated env var table and architecture description

Review & Testing Checklist for Human

• Default model changed: The Groq default model changed from openai/gpt-oss-120b → llama-3.3-70b-versatile. Verify this is the intended default for existing deployments, or whether existing deploys should set AI_MODEL=openai/gpt-oss-120b explicitly.
• "function_call" pattern in isAIToolCallError: This substring is fairly generic. Confirm it won't false-positive on non-error messages that mention "function_call" in passing.
• Rate limit detection loosened: isAIRateLimitError now returns true for any 429 status code, whereas previously it also required "retry" in the error message. Verify this won't cause unintended error handling behavior.
• Test with at least one non-Groq provider (e.g. set AI_PROVIDER=openai with a real API key) to confirm end-to-end streaming works through the agent.
• Version pinning: New deps use ^ ranges (^3.0.58, ^3.0.43) while existing @ai-sdk/groq and @ai-sdk/openai are pinned exactly. Confirm this inconsistency is acceptable.

Notes

• Typecheck passes (bun run typecheck:chat). Pre-existing biome lint errors in packages/cli/ are unrelated to this PR.
• No runtime testing was performed (no API keys available in the dev environment).
• Link to Devin session
• Requested by: @dhairyashiil

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR that start with 'DevinAI' or '@devin'.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring

[vercel]

Deployment failed with the following error:

You don't have permission to create a Preview Deployment for this Vercel project: companion-chat.

View Documentation: https://vercel.com/docs/accounts/team-members-and-roles

[cubic-dev-ai]

1 issue found across 7 files

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="apps/chat/lib/agent.ts">

<violation number="1" location="apps/chat/lib/agent.ts:888">
P2: Treating any 429 as an AI token limit can misclassify unrelated 429 errors and return an incorrect user message.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}

[cubic-dev-ai]

P2: Treating any 429 as an AI token limit can misclassify unrelated 429 errors and return an incorrect user message.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At apps/chat/lib/agent.ts, line 888:

<comment>Treating any 429 as an AI token limit can misclassify unrelated 429 errors and return an incorrect user message.</comment>

<file context>
@@ -858,30 +858,34 @@ export function isAIToolCallError(err: unknown): boolean {
     (err as { statusCode?: number }).statusCode === 429 ||
     (cause as { statusCode?: number } | undefined)?.statusCode === 429;
-  return hasRateLimit || (status429 && (msg.includes("retry") || causeMsg.includes("retry")));
+  return hasRateLimit || status429;
 }
 
</file context>