.env.example

# NeuroLink AI Toolkit - Environment Configuration
# Copy this file to .env and fill in your credentials

# =============================================================================
# ENTERPRISE PROXY CONFIGURATION (Optional)
# =============================================================================
# Corporate proxy support - automatically detected when set
# HTTPS_PROXY=http://proxy.company.com:8080
# HTTP_PROXY=http://proxy.company.com:8080
# NO_PROXY=localhost,127.0.0.1,.company.com

# For authenticated proxies:
# HTTPS_PROXY=http://username:password@proxy.company.com:8080
# HTTP_PROXY=http://username:password@proxy.company.com:8080

# =============================================================================
# OPENAI CONFIGURATION
# =============================================================================
# Required for OpenAI provider
OPENAI_API_KEY=sk-your-openai-api-key-here

# Optional: Override default model
OPENAI_MODEL=gpt-4o

# Optional: Maximum number of tools to send to OpenAI (default: 150)
# Lower values improve performance but limit tool availability
OPENAI_MAX_TOOLS=150

# =============================================================================
# AI ENHANCEMENT FEATURES (Optional)
# =============================================================================
# Model used for response quality evaluation (Lighthouse-compatible)
NEUROLINK_EVALUATION_MODEL=gemini-2.5-flash

# =============================================================================
# NEUROLINK EVALUATION SYSTEM (Enhanced Multi-Provider Support)
# =============================================================================

# Primary Evaluation Configuration
NEUROLINK_EVALUATION_PROVIDER=google-ai        # Primary provider choice
NEUROLINK_EVALUATION_MODE=fast                 # fast|balanced|quality

# Fallback Strategy
NEUROLINK_EVALUATION_FALLBACK_ENABLED=true     # Enable automatic fallbacks
NEUROLINK_EVALUATION_FALLBACK_PROVIDERS=openai,anthropic,vertex,bedrock  # Comma-separated fallback order

# Performance Tuning
NEUROLINK_EVALUATION_TIMEOUT=10000             # Evaluation timeout in ms
NEUROLINK_EVALUATION_MAX_TOKENS=500            # Max tokens for evaluation response
NEUROLINK_EVALUATION_TEMPERATURE=0.1           # Low temperature for consistent scoring

# Cost Optimization
NEUROLINK_EVALUATION_PREFER_CHEAP=true         # Prefer cost-effective models
NEUROLINK_EVALUATION_MAX_COST_PER_EVAL=0.01    # Max cost per evaluation in USD

# Advanced Configuration
NEUROLINK_EVALUATION_RETRY_ATTEMPTS=2          # Retry failed evaluations
NEUROLINK_EVALUATION_CACHE_ENABLED=false       # Cache evaluation results (future)

# =============================================================================
# AMAZON BEDROCK CONFIGURATION
# =============================================================================
# Required AWS credentials
AWS_ACCESS_KEY_ID=your-aws-access-key-id
AWS_SECRET_ACCESS_KEY=your-aws-secret-access-key

# Optional: Required for temporary credentials
AWS_SESSION_TOKEN=your-aws-session-token-if-using-temp-credentials

# Optional: Override defaults
AWS_REGION=us-east-2
BEDROCK_MODEL_ID=anthropic.claude-3-7-sonnet-20250219-v1:0

# =============================================================================
# GOOGLE VERTEX AI CONFIGURATION (Gemini 3)
# Supports all Gemini 3 models - see model list below
# Choose ONE of the three authentication methods below
# =============================================================================

# METHOD 1: Service Account File (Recommended for Production)
# Uncomment and provide path to your service account JSON file
GOOGLE_APPLICATION_CREDENTIALS=/path/to/your/service-account.json

# METHOD 2: Service Account JSON String (Good for Containers/Cloud)
# Uncomment and provide the full JSON content as a string
# GOOGLE_SERVICE_ACCOUNT_KEY={"type":"service_account","project_id":"your-project","private_key_id":"...","private_key":"-----BEGIN PRIVATE KEY-----\n...\n-----END PRIVATE KEY-----\n","client_email":"service-account@your-project.iam.gserviceaccount.com","client_id":"...","auth_uri":"https://accounts.google.com/o/oauth2/auth","token_uri":"https://oauth2.googleapis.com/token"}

# METHOD 3: Individual Environment Variables (Good for CI/CD)
# Uncomment and provide individual service account details
# GOOGLE_AUTH_CLIENT_EMAIL=service-account@your-project.iam.gserviceaccount.com
# GOOGLE_AUTH_PRIVATE_KEY="-----BEGIN PRIVATE KEY-----\nYOUR_PRIVATE_KEY_HERE\n-----END PRIVATE KEY-----"

# Required for ALL Google Vertex AI authentication methods
GOOGLE_VERTEX_PROJECT=your-gcp-project-id

# Optional: Override defaults
GOOGLE_VERTEX_LOCATION=us-central1
VERTEX_MODEL_ID=claude-sonnet-4@20250514

# Gemini 3 Models (Google AI Studio / Vertex AI)
# gemini-3-flash          - Fast, efficient model
# gemini-3-flash-preview  - Preview version of flash
# gemini-3-flash-latest   - Latest flash version
# gemini-3-pro            - Most capable Gemini 3 model
# gemini-3-pro-preview    - Preview version of pro
# gemini-3-pro-latest     - Latest pro version
#
# Examples:
# VERTEX_MODEL_ID=gemini-3-flash            # Fast, cost-effective Gemini 3
# VERTEX_MODEL_ID=gemini-3-pro              # Most capable Gemini 3

# =============================================================================
# PROVIDER SELECTION AND BEHAVIOR
# =============================================================================
# Optional: Set preferred provider order
DEFAULT_PROVIDER=auto
FALLBACK_PROVIDER=openai

# Optional: Enable/disable features
ENABLE_STREAMING=true
ENABLE_FALLBACK=true

# =============================================================================
# APPLICATION SETTINGS
# =============================================================================
# Optional: Application environment and configuration
PUBLIC_APP_ENVIRONMENT=dev
PUBLIC_APP_NAME=neurolink
PUBLIC_APP_URL=http://localhost:5173

# Optional: Debug and logging
NEUROLINK_DEBUG=false
LOG_LEVEL=info

# =============================================================================
# ADVANCED CONFIGURATION (Optional)
# =============================================================================
# Port for demo server
PORT=3000

# Custom model overrides
# OPENAI_MODEL=gpt-4o-mini
# BEDROCK_MODEL_ID=anthropic.claude-3-5-sonnet-20241022-v2:0
# BEDROCK_MODEL_ID=anthropic.claude-3-haiku-20240307-v1:0
# VERTEX_MODEL_ID=claude-opus-4@20250514
# VERTEX_MODEL_ID=gemini-2.5-flash

# Rate limiting and performance
# MAX_RETRIES=3
# TIMEOUT_MS=30000
# RATE_LIMIT_RPM=60

# =============================================================================
# AUTHENTICATION EXAMPLES
# =============================================================================

# Example 1: OpenAI Only
# OPENAI_API_KEY=sk-proj-abc123...
# DEFAULT_PROVIDER=openai

# Example 2: AWS Bedrock with IAM User
# AWS_ACCESS_KEY_ID=your-aws-access-key-id
# AWS_SECRET_ACCESS_KEY=your-aws-secret-access-key
# AWS_REGION=us-east-1
# DEFAULT_PROVIDER=bedrock

# Example 3: Google Vertex AI with Service Account File
# GOOGLE_APPLICATION_CREDENTIALS=/home/user/gcp-service-account.json
# GOOGLE_VERTEX_PROJECT=my-ai-project-123456
# DEFAULT_PROVIDER=vertex

# Example 4: LiteLLM for 100+ Models via Proxy
# LITELLM_BASE_URL=http://localhost:4000
# LITELLM_API_KEY=sk-anything
# LITELLM_MODEL=openai/gpt-4o-mini
# DEFAULT_PROVIDER=litellm

# Example 5: SageMaker Custom Model Deployment
# AWS_ACCESS_KEY_ID=your-aws-access-key-id
# AWS_SECRET_ACCESS_KEY=your-aws-secret-access-key
# AWS_REGION=us-east-1
# SAGEMAKER_DEFAULT_ENDPOINT=my-custom-model-endpoint
# DEFAULT_PROVIDER=sagemaker

# Example 6: All Providers with Fallback Chain
# OPENAI_API_KEY=sk-proj-abc123...
# AWS_ACCESS_KEY_ID=your-aws-access-key-id
# AWS_SECRET_ACCESS_KEY=your-aws-secret-access-key
# GOOGLE_APPLICATION_CREDENTIALS=/path/to/service-account.json
# GOOGLE_VERTEX_PROJECT=my-project
# LITELLM_BASE_URL=http://localhost:4000
# LITELLM_API_KEY=sk-anything
# SAGEMAKER_DEFAULT_ENDPOINT=my-custom-model
# DEFAULT_PROVIDER=bedrock
# FALLBACK_PROVIDER=openai
# ENABLE_FALLBACK=true

# =============================================================================
# TROUBLESHOOTING TIPS
# =============================================================================
# 1. For AWS "not authorized" errors: Check Bedrock/SageMaker access in your region
# 2. For Google auth errors: Verify service account permissions
# 3. For OpenAI rate limits: Check your API tier and usage
# 4. For LiteLLM proxy errors: Ensure proxy server is running on specified port
# 5. For SageMaker endpoint errors: Verify endpoint exists and is "InService"
# 6. Enable NEUROLINK_DEBUG=true for detailed logging
# 7. Check the README.md for comprehensive setup instructions

# =============================================================================
# GOOGLE AI STUDIO CONFIGURATION (UPDATED 2025-01-03)
# =============================================================================
# Required for Google AI Studio provider (free Google AI)
GOOGLE_AI_API_KEY=your-google-ai-studio-api-key
# or alternative name
GOOGLE_GENERATIVE_AI_API_KEY=your-google-ai-studio-api-key

# Recommended models (UPDATED):
GOOGLE_AI_MODEL=gemini-2.5-pro        # ✅ Recommended for most tasks
# GOOGLE_AI_MODEL=gemini-pro-vision   # ✅ For multimodal (text + images)

# ⚠️ AVOID these deprecated models:
# GOOGLE_AI_MODEL=gemini-2.5-pro-preview-05-06  # ❌ Causes empty responses
# GOOGLE_AI_MODEL=gemini-pro                     # ❌ May be deprecated

# Enhancement Features (OPTIONAL)
# These control default behavior - can be overridden per request
NEUROLINK_ENABLE_ANALYTICS=false      # Enable analytics by default
NEUROLINK_ENABLE_EVALUATION=false     # Enable evaluation by default

# =============================================================================
# HUGGING FACE CONFIGURATION
# =============================================================================
# Required for Hugging Face provider
HUGGINGFACE_API_KEY=hf_your-hugging-face-api-key-here
# or alternative name
HF_TOKEN=hf_your-hugging-face-api-key-here

# Optional: Override default model
HUGGINGFACE_MODEL=microsoft/DialoGPT-large

# =============================================================================
# OLLAMA LOCAL AI CONFIGURATION
# =============================================================================
# Ollama requires local installation and running service
# Download from: https://ollama.ai

# Optional: Override default base URL (if running on different host/port)
OLLAMA_BASE_URL=http://localhost:11434

# Optional: Override default model
OLLAMA_MODEL=llama3.2:latest

# Optional: Override request timeout (default 60 seconds)
OLLAMA_TIMEOUT=60000

# =============================================================================
# MISTRAL AI CONFIGURATION
# =============================================================================
# Required for Mistral AI provider (European GDPR-compliant AI)
MISTRAL_API_KEY=your-mistral-api-key-here

# Optional: Override default model
MISTRAL_MODEL=mistral-small
# Available models:
# - mistral-tiny (fastest, most cost-effective)
# - mistral-small (balanced performance, default)
# - mistral-medium (enhanced capabilities)
# - mistral-large (most capable model)

# Optional: Override default endpoint
MISTRAL_ENDPOINT=https://api.mistral.ai

# =============================================================================
# OPENAI COMPATIBLE CONFIGURATION
# =============================================================================
# Required for OpenAI-compatible providers (vLLM, etc.)
OPENAI_COMPATIBLE_BASE_URL=https://api.example.com/v1
OPENAI_COMPATIBLE_API_KEY=your-api-key-here

# Optional: Override default model (will auto-discover if not specified)
OPENAI_COMPATIBLE_MODEL=gpt-4o-mini
# Example configurations:
# vLLM: http://localhost:8000/v1
# LiteLLM proxy: http://localhost:4000/v1

# =============================================================================
# OPENROUTER CONFIGURATION (300+ AI Models from 60+ Providers)
# =============================================================================
# OpenRouter provides unified access to 300+ AI models from 60+ providers
# through a single API with automatic fallbacks and intelligent routing.
# Get your API key at: https://openrouter.ai/keys

# Required: Your OpenRouter API key
OPENROUTER_API_KEY=sk-or-v1-your-api-key-here

# Optional: Default model (format: provider/model-name)
OPENROUTER_MODEL=anthropic/claude-3-5-sonnet
# Popular model options:
# - anthropic/claude-3-5-sonnet (recommended for complex tasks)
# - anthropic/claude-3-5-haiku (fast and cost-effective)
# - openai/gpt-4o (OpenAI's latest)
# - openai/gpt-4o-mini (fast and cheap)
# - google/gemini-2.0-flash (Google's multimodal)
# - meta-llama/llama-3.1-70b-instruct (open source)
# - mistralai/mistral-large (European GDPR-compliant)
# Browse all models: https://openrouter.ai/models

# Optional: Attribution headers for OpenRouter dashboard
OPENROUTER_REFERER=https://yourapp.com
OPENROUTER_APP_NAME=YourAppName

# Optional: Request timeout in milliseconds (default: 60000)
OPENROUTER_TIMEOUT=60000

# =============================================================================
# LITELLM CONFIGURATION (100+ AI Models via Proxy)
# =============================================================================
# LiteLLM provides unified access to 100+ AI models through a proxy server
# Start local proxy: pip install litellm && litellm --port 4000

# Required: LiteLLM proxy server URL
LITELLM_BASE_URL=http://localhost:4000

# Required: API key for proxy (any value works for local proxy)
LITELLM_API_KEY=sk-anything

# Optional: Default model to use
LITELLM_MODEL=openai/gpt-4o-mini

# Optional: Request timeout (default 60 seconds)
LITELLM_TIMEOUT=60000

# Available model formats (examples):
# OpenAI: openai/gpt-4o, openai/gpt-4o-mini
# Anthropic: anthropic/claude-3-5-sonnet, anthropic/claude-3-haiku
# Google: google/gemini-2.0-flash, vertex_ai/gemini-pro
# Mistral: mistral/mistral-large, mistral/mixtral-8x7b

# =============================================================================
# AMAZON SAGEMAKER CONFIGURATION (Custom Model Deployment)
# =============================================================================
# Amazon SageMaker allows you to deploy and use your own custom trained models

# Required: AWS credentials (same as Bedrock)
# AWS_ACCESS_KEY_ID=your-aws-access-key-id
# AWS_SECRET_ACCESS_KEY=your-aws-secret-access-key
# AWS_REGION=us-east-1

# Required: Your SageMaker endpoint name
SAGEMAKER_DEFAULT_ENDPOINT=your-sagemaker-endpoint-name

# Optional: Timeout and retry settings
SAGEMAKER_TIMEOUT=30000             # Request timeout in milliseconds (default 30s)
SAGEMAKER_MAX_RETRIES=3             # Number of retry attempts (default 3)

# Optional: Model-specific settings
SAGEMAKER_MODEL=custom-model-name   # Model identifier for logging
SAGEMAKER_CONTENT_TYPE=application/json      # Request content type
SAGEMAKER_ACCEPT=application/json            # Response content type

# Optional: Session token for temporary credentials (IAM roles)
# AWS_SESSION_TOKEN=your-session-token

# =============================================================================
# IMAGE CACHE CONFIGURATION (Optional)
# =============================================================================
# Image caching reduces bandwidth usage by caching downloaded images from URLs
# Uses LRU (Least Recently Used) eviction strategy with TTL expiration
# Addresses IMG-026: No Caching issue - prevents redundant downloads

# Enable/disable image caching (default: false)
# When enabled, images downloaded from URLs are cached to avoid re-downloading
NEUROLINK_IMAGE_CACHE_ENABLED=false

# Maximum number of cached images (default: 100, min: 1, max: 1000)
# When cache is full, least recently used entries are evicted
NEUROLINK_IMAGE_CACHE_SIZE=100

# Time-to-live for cached images in milliseconds (default: 1800000 = 30 minutes)
# Min: 1000 (1 second), Max: 86400000 (24 hours)
# Entries older than TTL are automatically expired
NEUROLINK_IMAGE_CACHE_TTL_MS=1800000

# Maximum size per image in bytes (default: 10485760 = 10MB)
# Min: 1024 (1KB), Max: 104857600 (100MB)
# Images exceeding this size are not cached
NEUROLINK_IMAGE_MAX_SIZE=10485760

# Cache Features:
# ✅ LRU eviction when cache is full (least recently accessed removed first)
# ✅ Automatic TTL-based expiration (old entries cleaned up)
# ✅ Content hash deduplication (same image from different URLs cached once)
# ✅ URL normalization (removes tracking parameters like utm_source, fbclid)
# ✅ Cache hit/miss statistics tracking
# ✅ Rate limiting bypass for cached images (improves performance)

# Performance Impact:
# - Cache hits: ~1ms (no network request)
# - Cache misses: ~2-10s (network download + rate limiting)
# - Typical hit rate: 40-60% for repeated image URLs

# Example configurations:
# NEUROLINK_IMAGE_CACHE_ENABLED=false           # Disable caching entirely (testing)
# NEUROLINK_IMAGE_CACHE_SIZE=200                # Cache up to 200 images (high volume)
# NEUROLINK_IMAGE_CACHE_TTL_MS=3600000          # 1 hour TTL (longer retention)
# NEUROLINK_IMAGE_MAX_SIZE=5242880              # 5MB max per image (smaller limit)

# =============================================================================
# CONVERSATION TITLE GENERATION (Optional)
# =============================================================================
# Customize the prompt used to generate conversation titles from user messages.
# Use ${userMessage} placeholder to include the user's message in the prompt.
# If not set, uses default prompt optimized for 20-character titles.
#
NEUROLINK_TITLE_PROMPT='Create a short 3-word title for: ${userMessage}'
#

# =============================================================================
# CONVERSATION MEMORY CONFIGURATION (Optional)
# =============================================================================
# Enable conversation memory feature (ENABLED BY DEFAULT)
NEUROLINK_MEMORY_ENABLED=true                 # Enable conversation memory (default: true)

# Memory storage type (memory or redis)
STORAGE_TYPE=memory                            # Options: memory, redis

# Memory limits
NEUROLINK_MEMORY_MAX_SESSIONS=50              # Maximum number of sessions to keep in memory
# Redis Storage Configuration (used when STORAGE_TYPE=redis)
REDIS_HOST=localhost                           # Redis server hostname
REDIS_PORT=6379                                # Redis server port
REDIS_PASSWORD=                                # Redis password (if required)
REDIS_DB=0                                     # Redis database number
REDIS_KEY_PREFIX=neurolink:conversation:       # Key prefix for Redis
REDIS_TTL=86400                                # TTL in seconds (24 hours)

# Redis Connection Options
REDIS_CONNECT_TIMEOUT=30000                    # Connection timeout in ms
REDIS_MAX_RETRIES=3                            # Maximum connection retry attempts
REDIS_RETRY_DELAY=100                          # Retry delay in ms

# Storage Notes:
# - Memory storage: Fast access, no persistence between restarts
# - Redis storage: Persistent storage, survives restarts, supports clustering

# =============================================================================
# HIPPOCAMPUS MEMORY ENGINE (Embedded SDK)
# =============================================================================
# HC_CONDENSATION_PROMPT=             # Custom condensation prompt (placeholders: {{OLD_MEMORY}}, {{NEW_CONTENT}}, {{MAX_WORDS}})
# HC_LOG_LEVEL=warn                   # Memory SDK log level (debug, info, warn, error)

# =============================================================================
# ANTHROPIC CLAUDE CONFIGURATION
# =============================================================================
# Required for Anthropic Claude provider
ANTHROPIC_API_KEY=sk-ant-your-anthropic-api-key-here

# Optional: Override default model
ANTHROPIC_MODEL=claude-3-5-sonnet-20241022
# Available models:
# - claude-3-5-sonnet-20241022 (Latest and most capable)
# - claude-3-5-haiku-20241022 (Fast and cost-effective)
# - claude-3-opus-20240229 (Most powerful, slower)
# - claude-3-sonnet-20240229 (Balanced performance)
# - claude-3-haiku-20240307 (Fast and economical)

# =============================================================================
# AZURE OPENAI CONFIGURATION
# =============================================================================
# Required for Azure OpenAI provider
AZURE_OPENAI_API_KEY=your-azure-openai-32-char-hex-key
AZURE_OPENAI_ENDPOINT=https://your-resource.openai.azure.com

# Optional: Override default deployment/model
AZURE_OPENAI_MODEL=gpt-4o
# Available deployments (depends on your Azure setup):
# - gpt-4o (Latest GPT-4 Omni model)
# - gpt-4o-mini (Faster, cost-effective GPT-4 Omni)
# - gpt-4-turbo (GPT-4 Turbo)
# - gpt-35-turbo (GPT-3.5 Turbo)
# - text-embedding-ada-002 (For embeddings)

# Note: Azure OpenAI uses deployment names, not model names
# Create deployments in Azure OpenAI Studio first

# =============================================================================
# PERFORMANCE TUNING (Optional)
# =============================================================================
# Duration in milliseconds to cache the list of available tools
NEUROLINK_TOOL_CACHE_DURATION=20000

# =============================================================================
# NEUROLINK - MCP TOOLING (Optional)
# =============================================================================
# Disable MCP tools entirely if needed
NEUROLINK_DISABLE_MCP_TOOLS=false

# Maximum number of tools per provider (for performance)
NEUROLINK_MAX_TOOLS_PER_PROVIDER=5

# =============================================================================
# NEUROLINK ADVANCED CONFIGURATION (Optional)
# =============================================================================
# Config File Location
NEUROLINK_CONFIG_FILE=./neurolink.config.json

# Summarization (Conversation Memory) - TOKEN-BASED MEMORY ENABLED BY DEFAULT
# Token-based memory is enabled by default and uses 80% of each model's context window
NEUROLINK_SUMMARIZATION_ENABLED=true             # Enable summarization (default: true)
NEUROLINK_TOKEN_THRESHOLD=50000                  # Optional: Override token threshold (default: 80% of model context)
NEUROLINK_SUMMARIZATION_PROVIDER=vertex          # Provider for summarization (default: vertex)
NEUROLINK_SUMMARIZATION_MODEL=gemini-2.5-flash   # Model for summarization (default: gemini-2.5-flash)

# Deprecated: Turn-based memory settings (use TOKEN_THRESHOLD instead)
# NEUROLINK_SUMMARIZATION_THRESHOLD_TURNS=20     # Deprecated: Use token threshold
# NEUROLINK_SUMMARIZATION_TARGET_TURNS=10        # Deprecated: Use token threshold

# Default Generation Parameters
NEUROLINK_DEFAULT_MAX_TOKENS=4096
NEUROLINK_DEFAULT_TEMPERATURE=0.7

# Tool Configuration
NEUROLINK_TOOL_DESCRIPTION_MAX_LENGTH=1024
NEUROLINK_TOOL_NAME_MAX_LENGTH=64
NEUROLINK_DISABLE_DIRECT_TOOLS=false
NEUROLINK_DISABLE_BUILTIN_TOOLS=false
NEUROLINK_DISABLE_CUSTOM_TOOLS=false

# Evaluation Configuration
NEUROLINK_RAGAS_EVALUATION_MODEL=gemini-2.5-flash
NEUROLINK_RAGAS_EVALUATION_PROVIDER=google-ai
NEUROLINK_EVALUATION_THRESHOLD=7.0

# Context Processing
NEUROLINK_CONTEXT_EXCLUDE_FIELDS=password,secret,token
NEUROLINK_CONTEXT_INCLUDE_FIELDS=
NEUROLINK_CONTEXT_OVERRIDE_FIELDS=

# Factory & Caching
NEUROLINK_FACTORY_CACHE_SIZE=100
NEUROLINK_LOG_LEVEL=info

# Telemetry
NEUROLINK_TELEMETRY_ENABLED=false

# =============================================================================
# PROVIDER PERFORMANCE RATINGS (Optional - Advanced)
# =============================================================================
# Override default cost and performance ratings for providers
# Used for automatic provider selection and cost optimization

# Google AI Studio
GOOGLE_AI_DEFAULT_INPUT_COST=0.00001
GOOGLE_AI_DEFAULT_OUTPUT_COST=0.00003
GOOGLE_AI_SPEED_RATING=9
GOOGLE_AI_QUALITY_RATING=8
GOOGLE_AI_COST_RATING=10

# Google Vertex AI
GOOGLE_VERTEX_DEFAULT_INPUT_COST=0.000125
GOOGLE_VERTEX_DEFAULT_OUTPUT_COST=0.000375
GOOGLE_VERTEX_SPEED_RATING=9
GOOGLE_VERTEX_QUALITY_RATING=9
GOOGLE_VERTEX_COST_RATING=8

# OpenAI
OPENAI_DEFAULT_INPUT_COST=0.000001
OPENAI_DEFAULT_OUTPUT_COST=0.000002
OPENAI_SPEED_RATING=8
OPENAI_QUALITY_RATING=9
OPENAI_COST_RATING=7

# Anthropic
ANTHROPIC_DEFAULT_INPUT_COST=0.000003
ANTHROPIC_DEFAULT_OUTPUT_COST=0.000015
ANTHROPIC_SPEED_RATING=7
ANTHROPIC_QUALITY_RATING=10
ANTHROPIC_COST_RATING=6

# Vertex (Anthropic on Vertex)
VERTEX_DEFAULT_INPUT_COST=0.000003
VERTEX_DEFAULT_OUTPUT_COST=0.000015
VERTEX_SPEED_RATING=7
VERTEX_QUALITY_RATING=10
VERTEX_COST_RATING=7

# AWS Bedrock
BEDROCK_DEFAULT_INPUT_COST=0.000003
BEDROCK_DEFAULT_OUTPUT_COST=0.000015
BEDROCK_SPEED_RATING=7
BEDROCK_QUALITY_RATING=9
BEDROCK_COST_RATING=7

# Azure OpenAI
AZURE_DEFAULT_INPUT_COST=0.000001
AZURE_DEFAULT_OUTPUT_COST=0.000002
AZURE_SPEED_RATING=8
AZURE_QUALITY_RATING=9
AZURE_COST_RATING=7

# Ollama (Local)
OLLAMA_DEFAULT_INPUT_COST=0
OLLAMA_DEFAULT_OUTPUT_COST=0
OLLAMA_SPEED_RATING=6
OLLAMA_QUALITY_RATING=6
OLLAMA_COST_RATING=10

# Hugging Face
HUGGINGFACE_DEFAULT_INPUT_COST=0.00001
HUGGINGFACE_DEFAULT_OUTPUT_COST=0.00003
HUGGINGFACE_SPEED_RATING=5
HUGGINGFACE_QUALITY_RATING=6
HUGGINGFACE_COST_RATING=9

# Mistral AI
MISTRAL_DEFAULT_INPUT_COST=0.000001
MISTRAL_DEFAULT_OUTPUT_COST=0.000003
MISTRAL_SPEED_RATING=8
MISTRAL_QUALITY_RATING=8
MISTRAL_COST_RATING=9

# =============================================================================
# DYNAMIC MODEL CONFIGURATION (Optional)
# =============================================================================
MODEL_CONFIG_URL=https://your-server.com/models.json
MODEL_CONFIG_GITHUB_REPO=owner/repo
MODEL_CONFIG_GITHUB_BRANCH=main
MODEL_SERVER_PORT=3001

# =============================================================================
# ADDITIONAL PROVIDER CONFIGURATIONS (Optional)
# =============================================================================
# Azure OpenAI Deployments
AZURE_OPENAI_DEPLOYMENT=gpt-4o-deployment-name
AZURE_OPENAI_DEPLOYMENT_ID=your-deployment-id
AZURE_API_VERSION=2024-02-15-preview
AZURE_MODEL=gpt-4o

# Google Cloud Extended
GOOGLE_VOICE_AI_MODEL=gemini-2.0-flash-thinking-exp-01-21
GOOGLE_CLOUD_LOCATION=us-central1
VERTEX_LOCATION=us-central1
GOOGLE_CLOUD_PROJECT_ID=your-project-id
GOOGLE_PROJECT_ID=your-project-id
GOOGLE_CLOUD_PROJECT=your-project-id
GCLOUD_PROJECT=your-project-id

# AWS Extended
AWS_PROFILE=default
AWS_DEFAULT_REGION=us-east-1
BEDROCK_ENDPOINT_URL=https://bedrock-runtime.us-east-1.amazonaws.com
BEDROCK_MODEL=anthropic.claude-3-5-sonnet-20241022-v2:0

# Vertex Model Override
VERTEX_MODEL=claude-3-5-sonnet@20241022
VERTEX_PROJECT_ID=your-gcp-project-id

# =============================================================================
# GOOGLE VERTEX AI - ALTERNATIVE AUTH METHODS (Optional)
# =============================================================================
# For environments where service account file is not available

# Individual Auth Components
GOOGLE_AUTH_TYPE=service_account
GOOGLE_AUTH_BREEZE_PROJECT_ID=your-project-id
GOOGLE_AUTH_CLIENT_ID=123456789012345678901
GOOGLE_AUTH_CLIENT_EMAIL=service-account@your-project.iam.gserviceaccount.com
GOOGLE_AUTH_PRIVATE_KEY="-----BEGIN PRIVATE KEY-----\n...\n-----END PRIVATE KEY-----\n"
GOOGLE_AUTH_AUTH_URI=https://accounts.google.com/o/oauth2/auth
GOOGLE_AUTH_TOKEN_URI=https://oauth2.googleapis.com/token
GOOGLE_AUTH_AUTH_PROVIDER_CERT_URL=https://www.googleapis.com/oauth2/v1/certs
GOOGLE_AUTH_CLIENT_CERT_URL=https://www.googleapis.com/robot/v1/metadata/x509/service-account%40project.iam.gserviceaccount.com
GOOGLE_AUTH_UNIVERSE_DOMAIN=googleapis.com

# Alternative API Key Names
GEMINI_API_KEY=your-api-key-here        # Alternative to GOOGLE_AI_API_KEY

# =============================================================================
# SAGEMAKER EXTENDED CONFIGURATION (Optional)
# =============================================================================
# Endpoint Selection
SAGEMAKER_REGION=us-east-1
SAGEMAKER_ENDPOINT=your-endpoint-name
SAGEMAKER_ENDPOINT_NAME=your-endpoint-name
SAGEMAKER_MODEL_NAME=your-model-name
SAGEMAKER_MODEL_TYPE=huggingface         # Model type hint

# Multi-Environment Endpoints
SAGEMAKER_PRODUCTION_ENDPOINT=prod-endpoint
SAGEMAKER_TEST_ENDPOINT=test-endpoint

# Request Configuration
SAGEMAKER_CUSTOM_ATTRIBUTES={}
SAGEMAKER_INPUT_FORMAT=json
SAGEMAKER_OUTPUT_FORMAT=json
SAGEMAKER_MAX_TOKENS=2048
SAGEMAKER_TEMPERATURE=0.7
SAGEMAKER_TOP_P=0.9
SAGEMAKER_STOP_SEQUENCES=["\\n\\nHuman:"]

# Streaming Configuration
SAGEMAKER_BASE_STREAMING_DELAY_MS=50
SAGEMAKER_MAX_STREAMING_DELAY_MS=200

# =============================================================================
# MULTI-ENVIRONMENT AWS CREDENTIALS (Optional)
# =============================================================================
# For managing multiple AWS environments

# Development
DEV_AWS_ACCESS_KEY_ID=your-dev-aws-access-key-id
DEV_AWS_SECRET_ACCESS_KEY=your-dev-aws-secret-access-key

# Staging
STAGING_AWS_ACCESS_KEY_ID=your-staging-aws-access-key-id
STAGING_AWS_SECRET_ACCESS_KEY=your-staging-aws-secret-access-key

# Production
PROD_AWS_ACCESS_KEY_ID=your-prod-aws-access-key-id
PROD_AWS_SECRET_ACCESS_KEY=your-prod-aws-secret-access-key
PROD_AWS_REGION=us-east-1
PROD_SAGEMAKER_ENDPOINT=prod-endpoint-name

# =============================================================================
# EXTERNAL SERVICE INTEGRATIONS (Optional)
# =============================================================================
# Bitbucket
BITBUCKET_USERNAME=your-username
BITBUCKET_TOKEN=your-app-password
BITBUCKET_BASE_URL=https://api.bitbucket.org/2.0

# Slack
SLACK_BOT_TOKEN=xoxb-your-bot-token
SLACK_SIGNING_SECRET=your-signing-secret

# Database
DATABASE_URL=postgresql://user:password@localhost:5432/neurolink

# =============================================================================
# EXTENDED PROXY CONFIGURATION (Optional)
# =============================================================================
ALL_PROXY=socks5://proxy.company.com:1080
SOCKS_PROXY=socks5://proxy.company.com:1080

# =============================================================================
# OPENTELEMETRY CONFIGURATION (Optional)
# =============================================================================
OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4318
OTEL_SERVICE_NAME=neurolink
OTEL_SERVICE_VERSION=7.48.1
OTEL_EXPORTER_OTLP_TRACES_ENDPOINT=http://localhost:4318/v1/traces

# =============================================================================
# LITELLM EXTENDED (Optional)
# =============================================================================
LITELLM_FALLBACK_MODELS=openai/gpt-4o-mini,anthropic/claude-3-haiku
PROVIDER=litellm                       # Override default provider

# =============================================================================
# ADVANCED PROVIDER SETTINGS (Optional)
# =============================================================================
PROVIDER_FAILURE_THRESHOLD=3           # Failed requests before circuit breaker
OLLAMA_API_BASE=http://localhost:11434 # Alternative to OLLAMA_BASE_URL

# =============================================================================
# TESTING & DEVELOPMENT (Optional)
# =============================================================================
NEUROLINK_PACKAGE=@juspay/neurolink    # Package identifier
PROMPT_TYPE=default                    # Prompt template type

# =============================================================================
# OBSERVABILITY & ANALYTICS
# =============================================================================
# Langfuse Configuration
# Sign up at: https://cloud.langfuse.com or self-host
LANGFUSE_PUBLIC_KEY=pk-lf-your-public-key-here
LANGFUSE_SECRET_KEY=sk-lf-your-secret-key-here
LANGFUSE_BASE_URL=https://cloud.langfuse.com
LANGFUSE_ENABLED=false

# OpenTelemetry Configuration (Optional)
# Telemetry is automatically enabled when OTEL_EXPORTER_OTLP_ENDPOINT is set
# OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4318
# OTEL_SERVICE_NAME=neurolink-ai

# =============================================================================
# REAL-TIME VOICE SERVER (voice-server command)
# =============================================================================

# Cartesia TTS — https://cartesia.ai
CARTESIA_API_KEY=
# Optional: API version pin used in Cartesia WS query params
CARTESIA_API_VERSION=2025-04-16
# Optional: override if Cartesia provides a different WS endpoint for your account/region
CARTESIA_WS_BASE_URL=wss://api.cartesia.ai/tts/websocket

# Soniox STT — https://soniox.com
SONIOX_API_KEY=
# Optional: override Soniox streaming WS endpoint
SONIOX_WS_URL=wss://stt-rt.soniox.com/transcribe-websocket

# Picovoice Cobra VAD — https://picovoice.ai
PICOVOICE_ACCESS_KEY=

# Optional: override LLM provider and model used by the voice pipeline
VOICE_LLM_MODEL=gpt-4o-automatic
VOICE_LLM_PROVIDER=azure

# =============================================================================
# DEEPSEEK CONFIGURATION
# =============================================================================
# Get an API key at https://platform.deepseek.com/api_keys
DEEPSEEK_API_KEY=
# Optional: override default model (deepseek-chat | deepseek-reasoner)
DEEPSEEK_MODEL=deepseek-chat
# Optional: override default base URL
# DEEPSEEK_BASE_URL=https://api.deepseek.com

# =============================================================================
# NVIDIA NIM CONFIGURATION
# =============================================================================
# Get an API key at https://build.nvidia.com/settings/api-keys
NVIDIA_NIM_API_KEY=
# Optional: override default model (browse https://build.nvidia.com/models)
NVIDIA_NIM_MODEL=meta/llama-3.3-70b-instruct
# Optional: override default base URL (use for self-hosted NIM)
# NVIDIA_NIM_BASE_URL=https://integrate.api.nvidia.com/v1
# Optional NIM extras (rarely needed, leave commented)
# NVIDIA_NIM_TOP_K=
# NVIDIA_NIM_MIN_P=
# NVIDIA_NIM_REPETITION_PENALTY=
# NVIDIA_NIM_MIN_TOKENS=
# NVIDIA_NIM_CHAT_TEMPLATE=

# =============================================================================
# LM STUDIO CONFIGURATION (local provider)
# =============================================================================
# Install LM Studio: https://lmstudio.ai/
# Load a model in the app and click "Start Server"
LM_STUDIO_BASE_URL=http://localhost:1234/v1
# Optional: explicit model id (blank = auto-discover from /v1/models)
LM_STUDIO_MODEL=
# Optional: API key (not required for stock LM Studio; use only when running
# behind an auth-proxying reverse-proxy)
# LM_STUDIO_API_KEY=

# =============================================================================
# LLAMA.CPP CONFIGURATION (local provider)
# =============================================================================
# Run: ./llama-server -m model.gguf --port 8080 (add --jinja for tool support)
LLAMACPP_BASE_URL=http://localhost:8080/v1
# Optional: explicit model id (blank = use whatever model llama-server has loaded)
LLAMACPP_MODEL=
# Optional: API key (not required for stock llama-server; use only when running
# behind an auth-proxying reverse-proxy)
# LLAMACPP_API_KEY=

# =============================================================================
# VOICE PROVIDER CREDENTIALS (TTS / STT / Realtime)
# =============================================================================

# ElevenLabs TTS (https://elevenlabs.io)
ELEVENLABS_API_KEY=

# Deepgram STT (https://deepgram.com)
DEEPGRAM_API_KEY=

# Azure Cognitive Services Speech (TTS + STT)
# Get from: Azure Portal → Cognitive Services → Speech → Keys and Endpoint
AZURE_SPEECH_KEY=
AZURE_SPEECH_REGION=eastus

# Google STT + Gemini Live realtime voice
# Both providers accept GOOGLE_AI_API_KEY (defined above) or GEMINI_API_KEY as
# aliases — no need to set a separate GOOGLE_API_KEY. For service-account auth
# with Google STT, set GOOGLE_APPLICATION_CREDENTIALS instead.