.env.example

# Java Chat environment variables
#
# Copy to .env:
#   cp .env.example .env

# Environment Configuration
# Set exactly local, dev, or prod so filesystem ingestion state uses the matching profile root.
# Qdrant collection identity is shared by every profile for the active embedding generation.
SPRING_PROFILE=local

# Server Configuration
PORT=8085

# Shared OpenAI-compatible LLM gateway
# Gateway priority tier for live chat turns (X-Tier header); production-z is
# java-chat's recognized live tier. Background/batch work sends "batch".
# LLM_GATEWAY_TIER=production-z
OPENAI_API_KEY=
# Chat and embeddings share the gateway base URL and credential.
OPENAI_BASE_URL=https://api.llm-gateway.iocloudhost.net/v1
OPENAI_MODEL=gpt-5.4
# Local Embedding Server (if you have one running)
APP_LOCAL_EMBEDDING_ENABLED=false
LOCAL_EMBEDDING_SERVER_URL=http://127.0.0.1:8088
APP_LOCAL_EMBEDDING_MODEL=qwen/qwen3-embedding-4b
APP_LOCAL_EMBEDDING_DIMENSIONS=2560

# Qdrant Vector Database (local docker-compose defaults)
QDRANT_HOST=localhost
QDRANT_SSL=false
QDRANT_PORT=8086       # gRPC (app)
QDRANT_REST_PORT=8087  # REST (scripts/monitors)
QDRANT_COLLECTION_BOOKS=java-chat-qwen3-embedding-4b-2560-books
QDRANT_COLLECTION_DOCS=java-chat-qwen3-embedding-4b-2560-docs
QDRANT_COLLECTION_ARTICLES=java-chat-qwen3-embedding-4b-2560-articles
QDRANT_COLLECTION_PDFS=java-chat-qwen3-embedding-4b-2560-pdfs
QDRANT_API_KEY=

# Generation-specific ingestion state (use local/dev/prod to match SPRING_PROFILE).
DOCS_SNAPSHOT_DIR=data/qwen3-embedding-4b-2560/local/snapshots
DOCS_PARSED_DIR=data/qwen3-embedding-4b-2560/local/parsed
DOCS_INDEX_DIR=data/qwen3-embedding-4b-2560/local/index

# RAG Configuration
RAG_CHUNK_MAX_TOKENS=900
RAG_CHUNK_OVERLAP_TOKENS=150
RAG_TOP_K=12
RAG_RETURN_K=6
RAG_CITATIONS_K=3