What's your AI tech stack in 2026?

Curious what everyone's full AI stack looks like in 2026. Here's ours:

LLM: GPT-4o (primary), GPT-4o-mini (high-volume), o1 (complex reasoning) Embeddings: text-embedding-3-large (1536 dims) Vector DB: Pinecone Serverless Framework: LangChain (reluctantly, considering switching) Orchestration: Custom Python + FastAPI Monitoring: LangSmith + custom dashboards Eval: Custom eval suite + promptfoo Cache: Redis for semantic caching Cost: ~$2K/month API spend

What does your stack look like? Especially curious about alternatives to LangChain — it feels like too much abstraction for our needs.