configs/env.server

# ============================================================================
# Auralynq — Device profile: Linux server (yourIP)
# Hardware: 3× RTX 2080 Ti (44 GB VRAM total) · 188 GB RAM · CUDA 12.9
# Stack:    Ollama (primary LLM + embeddings) · Qdrant (vector store)
# Copy to .env at repo root before starting.
# ============================================================================

# ── Network ─────────────────────────────────────────────────────────────────
# Browser calls the API directly at the server's LAN IP.
NEXT_PUBLIC_API_BASE=http://yourIP:8000/api
AURALYNQ_SERVE__CORS_ORIGINS=["http://yourIP:3000","http://localhost:3000"]
AURALYNQ_SERVE__API_KEY=

# ── LLM (auto → Ollama first; SLM if Ollama down) ───────────────────────────
AURALYNQ_LLM__PROVIDER=auto
AURALYNQ_LLM__MODEL=llama3.1:8b
# SLM fallback: Qwen2.5-0.5B GGUF — ~350 MB, all layers on GPU when CUDA present
AURALYNQ_LLM__SLM_REPO=Qwen/Qwen2.5-0.5B-Instruct-GGUF
AURALYNQ_LLM__SLM_FILENAME=qwen2.5-0.5b-instruct-q4_k_m.gguf
AURALYNQ_LLM__SLM_N_GPU_LAYERS=-1   # -1 = all layers on GPU

# ── Embeddings (auto → Ollama nomic-embed-text first) ───────────────────────
AURALYNQ_EMBEDDING__PROVIDER=auto

# ── Vector store: Qdrant running on :6333 ────────────────────────────────────
# Start: ~/.local/bin/qdrant --config-path ~/.config/qdrant.yaml
AURALYNQ_VECTOR__BACKEND=qdrant
AURALYNQ_VECTOR__URL=http://localhost:6333

# ── Visual grounding ────────────────────────────────────────────────────────
AURALYNQ_VISUAL__ENABLED=true
AURALYNQ_VISUAL__PAGE_RENDERING_ENABLED=true
AURALYNQ_VISUAL__RENDER_DPI=144

# ── Optional paid providers ──────────────────────────────────────────────────
COHERE_API_KEY=
OPENAI_API_KEY=
ANTHROPIC_API_KEY=
HUGGINGFACE_TOKEN=