env.donotcommit.example

# NeMo Microservices Configuration
# Copy this file to env.donotcommit and fill in your values
# env.donotcommit is git-ignored and will NOT be committed

# REQUIRED: Namespace for cluster services
# Replace with your actual OpenShift namespace/project name
# Find your namespace: oc projects
NMS_NAMESPACE=your-namespace

# OPTIONAL: NeMo Data Store token
# Default is "token" - update if your deployment uses a different token
NDS_TOKEN=token

# OPTIONAL: Dataset name for RAG tutorial documents
DATASET_NAME=rag-tutorial-documents

# OPTIONAL: RAG Configuration
# Number of documents to retrieve
RAG_TOP_K=5
# Similarity threshold for retrieval
RAG_SIMILARITY_THRESHOLD=0.3

# ----- Optional: RHOAI LlamaStack (for rag-tutorial-rhoai.ipynb) -----
# If using the RHOAI-deployed LlamaStack (copilot-llama-stack), set these.
# The notebook rag-tutorial-rhoai.ipynb sets these by default; override here if needed.
# LLAMASTACK_URL=http://copilot-llama-stack-service.anemo-rhoai.svc.cluster.local:8321
# LLAMASTACK_CHAT_MODEL=vllm-inference/redhataillama-31-8b-instruct
# Leave LLAMASTACK_API_KEY empty (no client auth for copilot-llama-stack)
# LLAMASTACK_API_KEY=

# REQUIRED: NIM Service Account Token (for embeddings and nemo-instances LlamaStack)
# Kubernetes service account token (JWT) for authenticating with KServe InferenceService
# This token is REQUIRED for LlamaStack to work with authenticated KServe services
# Get your token: oc create token <service-account-name> -n <your-namespace> --duration=8760h
# Example: oc create token my-model-sa -n my-namespace --duration=8760h
# The service account name is typically: <inferenceservice-name>-sa
# Find your service account: oc get sa -n <your-namespace> | grep model
NIM_SERVICE_ACCOUNT_TOKEN=

# OPTIONAL: API Keys (only needed if using external APIs as fallback)
# OPENAI_API_KEY=
# NVIDIA_API_KEY=
# HF_TOKEN=