ai insight rework #2

backend/.env.example

@@ -70,7 +70,8 @@ OLLAMA_BASE_URL=http://127.0.0.1:11434
 LLM_MODEL_NAME=qwen2.5:3b
 LLM_TIMEOUT_SECONDS=1200
 # Plan §6 CPU inference tuning (Profile A 8GB: qwen2.5:1.5b; Profile B 16GB: qwen2.5:3b)
-LLM_NUM_CTX=2048
+LLM_NUM_CTX=8192
+LLM_NUM_PREDICT=768
 LLM_NUM_THREAD=4
 LLM_SEED=42
 LLM_TEMPERATURE=0.2