.env.example

# ═══════════════════════════════════════════════════════════════════════
# LaunchStack — Environment Variables
# ═══════════════════════════════════════════════════════════════════════

# ── Core ─────────────────────────────────────────────────────────────

DATABASE_URL="postgresql://user:password@localhost:5432/pdr_ai"
BLOB_READ_WRITE_TOKEN="BLOB_READ_WRITE_TOKEN"

# Clerk Authentication
NEXT_PUBLIC_CLERK_PUBLISHABLE_KEY=NEXT_PUBLIC_CLERK_PUBLISHABLE_KEY
CLERK_SECRET_KEY=CLERK_SECRET_KEY

# ── AI Providers ─────────────────────────────────────────────────────
#
# Option A: Set ONE global provider for everything (simplest)
#   AI_BASE_URL + AI_API_KEY routes embeddings, reranking, NER,
#   transcription, and chat through a single OpenAI-compatible API.
#
# Option B: Use OpenAI as default, override specific capabilities
#   Set OPENAI_API_KEY, then override individual services below.
#
# Option C: Configure each capability independently
#   Set per-capability env vars (RERANK_API_*, NER_API_*, etc.)
#
# Resolution order: per-capability → global fallback → legacy key → default

# Global AI provider (optional — overrides everything when set)
# AI_BASE_URL="https://api.siliconflow.cn/v1"
# AI_API_KEY="sk-..."

# OpenAI (required unless AI_BASE_URL is set)
OPENAI_API_KEY="sk-..."
OPENAI_MODEL="gpt-5-mini"
EMBEDDING_INDEX="legacy-openai-1536"

# 32 random bytes, base64-encoded — encrypts per-company embedding provider
# API keys at rest. Generate once and keep constant across deploys:
#   node -e "console.log(require('crypto').randomBytes(32).toString('base64'))"
EMBEDDING_SECRETS_KEY=

# Anthropic (optional — enables Claude models)
# ANTHROPIC_API_KEY=""
# ANTHROPIC_MODEL=""

# Google AI (optional — enables Gemini models)
# GOOGLE_AI_API_KEY=""
# GOOGLE_MODEL=""

# Hugging Face embeddings (optional)
# HUGGINGFACE_API_KEY=
# HUGGINGFACE_EMBEDDING_MODEL=
# HUGGINGFACE_EMBEDDING_DIMENSION=
# HUGGINGFACE_EMBEDDING_VERSION="v1"

# Ollama (optional — local models)
# OLLAMA_BASE_URL="http://localhost:11434"
# OLLAMA_MODEL="llama3.1:8b"

# ── Per-Capability Overrides (optional) ──────────────────────────────
# Only set these if you want a specific capability to use a different
# provider than your global/OpenAI default.

# Embeddings — defaults to OpenAI text-embedding-3-large
# EMBEDDING_API_BASE_URL="https://api.siliconflow.cn/v1"
# EMBEDDING_API_KEY="sk-..."
# EMBEDDING_MODEL="BAAI/bge-m3"

# Reranking — defaults to Jina AI
# RERANK_API_BASE_URL="https://api.siliconflow.cn/v1"
# RERANK_API_KEY="sk-..."
# RERANK_MODEL="BAAI/bge-reranker-v2-m3"
JINA_API_KEY=""

# NER (entity extraction) — defaults to gpt-4o-mini via OpenAI
# Set to SiliconFlow + Qwen3.5-4B for FREE NER:
# NER_API_BASE_URL="https://api.siliconflow.cn/v1"
# NER_API_KEY="sk-..."
# NER_MODEL="Qwen/Qwen3.5-4B"

# Sidecar ML Compute (Optional)
# SIDECAR_URL="http://localhost:8080"
# SIDECAR_EMBEDDING_MODEL=
# SIDECAR_EMBEDDING_DIMENSION=
# SIDECAR_EMBEDDING_VERSION="v1"

# Transcription — defaults to Groq whisper-large-v3-turbo
# TRANSCRIPTION_API_BASE_URL="https://api.groq.com/openai/v1"
# TRANSCRIPTION_API_KEY="gsk-..."
# TRANSCRIPTION_MODEL="whisper-large-v3-turbo"
GROQ_API_KEY=""

# ── Token System ─────────────────────────────────────────────────────

# Free tokens granted to each new company on signup (default: 10,000,000)
# TOKEN_SIGNUP_BONUS="10000000"

# ── Document Processing — OCR ────────────────────────────────────────

# Paid providers (optional — used in cloud mode)
# AZURE_DOC_INTELLIGENCE_ENDPOINT="https://..."
# AZURE_DOC_INTELLIGENCE_KEY="..."
# LANDING_AI_API_KEY="..."
# DATALAB_API_KEY="..."

# Self-hosted OCR (optional — for docker-compose deployments)
# OCR_WORKER_URL="http://localhost:8001"
# OCR_DEFAULT_PROVIDER="MARKER"   # "MARKER" | "DOCLING"
# APP_PUBLIC_URL="http://app:3000"

# ── DOCX Redlining (Adeu) ───────────────────────────────────────────

# Docker:   http://sidecar:8000
# Local:    http://localhost:8000
ADEU_SERVICE_URL="http://localhost:8000"

# ── Web Search (optional) ───────────────────────────────────────────

# SEARCH_PROVIDER="tavily"  # "tavily" | "serper" | "fallback" | "parallel"
# TAVILY_API_KEY="tvly-..."
# SERPER_API_KEY="..."

# ── Marketing Pipeline (optional) ───────────────────────────────────

# REDDIT_CLIENT_ID=""
# REDDIT_CLIENT_SECRET=""
# REDDIT_USER_AGENT=""
# TWITTER_BEARER_TOKEN=""
# LINKEDIN_ACCESS_TOKEN=""
# BLUESKY_HANDLE=""
# BLUESKY_APP_PASSWORD=""

# ── LangSmith Tracing (optional) ────────────────────────────────────

# LANGCHAIN_TRACING_V2="false"
# LANGCHAIN_API_KEY="..."
# LANGCHAIN_PROJECT="pdr_ai_v2"

# ── File Upload (optional) ──────────────────────────────────────────

NEXT_PUBLIC_UPLOADTHING_ENABLED="true"
# UPLOADTHING_TOKEN="..."

# ── Background Jobs ─────────────────────────────────────────────────

JOB_RUNNER="inngest"
INNGEST_EVENT_KEY="local"
# INNGEST_SIGNING_KEY="signkey-dev-xxxxx"

# ── Graph RAG (optional) ────────────────────────────────────────────

# ENABLE_GRAPH_RETRIEVER="false"
# NEO4J_URI=""
# NEO4J_USERNAME=""
# NEO4J_PASSWORD=""

# ── File Storage ────────────────────────────────────────────────────
# "s3" → any S3-compatible endpoint (AWS, MinIO, SeaweedFS)
# "database" → base64 in Postgres (no-S3 fallback)
# Unset → auto-detect based on S3 vars

# NEXT_PUBLIC_STORAGE_PROVIDER="s3"
# NEXT_PUBLIC_S3_ENDPOINT="http://localhost:8333"
# S3_REGION="us-east-1"
# S3_ACCESS_KEY="pdr_local_key"
# S3_SECRET_KEY="pdr_local_secret"
# S3_BUCKET_NAME="pdr-documents"

# ── CORS ─────────────────────────────────────────────────────────────
# Comma-separated list of allowed origins for API CORS headers
# CORS_ALLOWED_ORIGINS="https://yourdomain.com,https://app.yourdomain.com"

# ── Logging ──────────────────────────────────────────────────────────
# Log level: fatal, error, warn, info, debug, trace (default: info in prod, debug in dev)
# LOG_LEVEL="info"

# ── Legacy (deprecated) ─────────────────────────────────────────────
# SIDECAR_URL — ML models removed from sidecar. Use cloud providers above.