Update .env

.env

@@ -1,20 +1,4 @@
 # --- Core LLM backend (switch here only) ---
-BACKEND_LLM=llamacpp
-LLAMACPP_MODEL_PATH=models/qwen2.5-1.5b-instruct-q4_k_m.gguf           #tinyllama-1.1b-chat-v1.0.Q4_K_M.gguf
-HF_MODEL_REPO=Qwen/Qwen2.5-1.5B-Instruct-GGUF
-
-# llama.cpp runtime knobs
-N_CTX=4096
-N_THREADS=4
-N_GPU_LAYERS=0
-
-# ASR device: 'cpu' is safest across Mac/HF; use 'mps' only if you’ve verified support
-ASR_DEVICE=cpu
-
-# TTS preference (choose ONE)
-TTS_ENGINE=piper
-PIPER_MODEL=models/piper/en_US-amy-medium.onnx
-
 
 # Files written by TTS go here (and get cleaned by our code)
 VOICE_AUDIO_DIR=runtime/audio
@@ -23,15 +7,34 @@ VOICE_AUDIO_DIR=runtime/audio
 # OPENAI_API_KEY=
 # GROQ_API_KEY=
 
-# Environment flags
-IS_HF_SPACE=false
-DEBUG=true
-CAFE_UNRELATED_LIMIT=3
-
 API_BACKEND=sim   # sim | mock | http
 
 # Switch between old rule router and LLM-driven flow (no code changes needed)
 ROUTER_MODE=llm
 
-PIPER_BIN=/home/user/bin/piper
-SAY_VOICE=Samantha
+SAY_VOICE=Samantha
+
+
+# Which backend to use
+BACKEND_LLM=llamacpp
+
+# Path where the GGUF model will be saved after download
+LLAMACPP_MODEL_PATH=models/qwen2.5-1.5b-instruct-q4_k_m.gguf
+
+# HF repo to download the model from
+HF_MODEL_REPO=Qwen/Qwen2.5-1.5B-Instruct-GGUF
+HF_MODEL_FILE=qwen2.5-1.5b-instruct-q4_k_m.gguf
+
+# llama.cpp runtime knobs
+N_CTX=4096
+N_THREADS=4
+N_GPU_LAYERS=0
+
+# Audio + misc
+ASR_DEVICE=cpu
+TTS_ENGINE=piper
+PIPER_MODEL=models/piper/en_US-amy-medium.onnx
+PIPER_BIN=piper
+
+IS_HF_SPACE=true
+DEBUG=false