# MAF-A2A-NVIDIA-NemoAgents Environment Configuration

# ==============================================================================
# LLM Provider Configuration - Choose ONE
# ==============================================================================

# Option A: NVIDIA API (NIM - NVIDIA Inference Microservice)
NVIDIA_API_KEY=your-nvidia-api-key-here

# Option B: Azure OpenAI
# Classic Azure OpenAI: https://your-resource.openai.azure.com/
# Foundry models (OpenAI v1 surface): https://your-resource.services.ai.azure.com/
#   (the /openai/v1 suffix is appended automatically when omitted)
AZURE_OPENAI_ENDPOINT=https://your-resource.openai.azure.com/
AZURE_OPENAI_DEPLOYMENT_NAME=your-deployment-name
AZURE_OPENAI_API_KEY=your-api-key-here
AZURE_OPENAI_API_VERSION=2025-04-01-preview
# Optional. Leave unset for reasoning models (e.g. gpt-5-mini) that only allow the
# default temperature. Set to 0 for deterministic output on classic chat models.
# AZURE_OPENAI_TEMPERATURE=0

# ==============================================================================
# NeMo Agent Configuration
# ==============================================================================
NEMO_HOST=127.0.0.1
NEMO_PORT=8088
NEMO_LOG_LEVEL=INFO
NEMO_WORKFLOW_PROFILE=fast
NEMO_FAST_MODEL_NAME=meta/llama-3.2-3b-instruct

# ==============================================================================
# MAF Agent Configuration
# ==============================================================================
MAF_HOST=127.0.0.1
MAF_PORT=5055
MAF_LOG_LEVEL=Information

# --- Grounded action agent (RAG + optional MCP) ---
# The MAF Action Agent grounds its decisions in a local runbook/policy knowledge base
# (src/MafActionAgent/knowledge/*.md) using LOCAL ONNX embeddings (ElBruno.LocalEmbeddings).
# No cloud embedding deployment is required. The chat model uses Azure OpenAI (above).
# Local sentence-transformers model used for embeddings (384-dim MiniLM by default).
EMBEDDINGS_MODEL=sentence-transformers/all-MiniLM-L6-v2
# Optional Model Context Protocol retrieval. Disabled by default; when true the agent also
# loads tools from MCP_SERVER_ENDPOINT (default: Microsoft Learn). Failures degrade to local RAG.
ENABLE_MCP_RETRIEVAL=false
MCP_SERVER_ENDPOINT=https://learn.microsoft.com/api/mcp

# --- Pitch image agent (MEAI IImageGenerator via GPT-Image-2 / Azure OpenAI) ---
# Additive, isolated from the grounded-action path. Always enabled; it activates automatically
# whenever the endpoint + API key below are set (leave them blank to keep it a no-op). When active it
# pre-renders one "incident hero" image in the background at startup (cached to disk, idempotent)
# served at GET /api/pitch/hero-image. Uses GPT-Image-2 (Azure OpenAI) via ElBruno.Text2Image.Foundry.
# GPT-Image-2 can take several minutes, so generation runs in the background and never blocks startup.
# Note: a bare *.services.ai.azure.com endpoint is auto-normalized to the /openai/v1 path.
FOUNDRY_IMAGE_ENDPOINT=https://your-resource.services.ai.azure.com/openai/v1
FOUNDRY_IMAGE_API_KEY=your-gpt-image-api-key
FOUNDRY_IMAGE_DEPLOYMENT=gpt-image-2
FOUNDRY_IMAGE_MODEL_NAME=GPT-Image-2
# Optional request timeout (seconds) for image generation; defaults to 300.
# FOUNDRY_IMAGE_TIMEOUT_SECONDS=300

# ==============================================================================
# Web UI Configuration
# ==============================================================================
WEB_UI_HOST=127.0.0.1
WEB_UI_PORT=5000
WEB_UI_LOG_LEVEL=Information
AGENT_HTTP_TIMEOUT_SECONDS=120
NEMO_CHAT_TIMEOUT_SECONDS=120
NEMO_WARMUP_ENABLED=true
NEMO_WARMUP_DELAY_SECONDS=0
NEMO_WARMUP_TIMEOUT_SECONDS=120
NEMO_WARMUP_RETRY_DELAY_SECONDS=5
NEMO_WARMUP_MAX_ATTEMPTS=3
NEMO_WARMUP_REQUEST_MAX_WAIT_SECONDS=0
NEMO_WARMUP_MESSAGE=Analyze quarterly revenue values [100, 110, 120, 130] for Q1-Q4 2024 and reply with READY only.
CHAT_ANALYSIS_CONTEXT_TTL_MINUTES=30
CHAT_ANALYSIS_CONTEXT_MAX_LENGTH=1600

# ==============================================================================
# Service URLs (used for internal communication)
# ==============================================================================
NEMO_A2A_ENDPOINT=http://127.0.0.1:8088
NEMO_PUBLIC_BASE_URL=http://127.0.0.1:8088
MAF_AGENT_ENDPOINT=http://127.0.0.1:5055
WEB_UI_ENDPOINT=http://127.0.0.1:5000

# ==============================================================================
# Aspire Configuration
# ==============================================================================
ASPIRE_DASHBOARD_OTLP_HTTP_ENDPOINT_URL=http://localhost:4319
ASPIRE_DASHBOARD_OTLP_ENDPOINT_URL=http://localhost:4317
ASPIRE_RESOURCE_SERVICE_BINDING_OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4317

# ==============================================================================
# Logging Configuration
# ==============================================================================
LOG_LEVEL=Information
ENABLE_OTEL_TRACING=true
OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4317

# ==============================================================================
# Feature Flags
# ==============================================================================
ENABLE_DEMO_MODE=true
ENABLE_HEALTH_CHECKS=true
ENABLE_OTEL_TRACING=true

# ==============================================================================
# Notes
# ==============================================================================
# 1. Copy this file to .env in the repository root
# 2. Fill in the values for your environment
# 3. For local development, you can use the defaults provided
# 4. For Azure OpenAI, ensure you have a deployed model
# 5. For NVIDIA API, get your API key from https://build.nvidia.com