# ============================================================
# 🎯 RedNote Insight — Environment Configuration
# ============================================================
# Setup:
#   1. Copy this file:  cp .env.example .env
#   2. Fill in your API key below
#   3. Get a free key at: https://siliconflow.cn (supports DeepSeek)
#
# All fields are compatible with any OpenAI-format API provider:
#   - SiliconFlow (硅基流动):  https://api.siliconflow.cn/v1
#   - Alibaba Bailian (阿里云百炼): https://dashscope.aliyuncs.com/compatible-mode/v1
#   - DeepSeek Official:     https://api.deepseek.com/v1
#   - OpenAI:                https://api.openai.com/v1
# ============================================================

# ===== LLM Configuration =====
OPENAI_BASE_URL=https://api.siliconflow.cn/v1
OPENAI_API_KEY=your_api_key_here

# Optional: use a dedicated provider for report generation. Confirm the
# current model ids through DeepSeek's official `/models` endpoint.
LLM_BASE_URL=https://api.deepseek.com
LLM_API_KEY=your_deepseek_api_key_here
LLM_MODEL=deepseek-v4-flash

# ===== Embedding Model =====
# For semantic search (Chinese-optimized)
EMBEDDING_MODEL=BAAI/bge-m3

# ===== Reranker Model =====
# Cross-Encoder for relevance scoring
RERANKER_MODEL=BAAI/bge-reranker-v2-m3

# ===== Hybrid Retrieval Tuning =====
# RRF fusion constant (Cormack et al. 2009 paper recommends 60)
RRF_K=60
# Number of candidates from BM25 keyword search
BM25_K=25
# Number of candidates from vector search
VECTOR_K=25
# Final documents returned after RRF fusion
FINAL_K=10
# Re-ranker relevance threshold (docs below this score are discarded)
RERANKER_THRESHOLD=0.1

# ===== PostgreSQL（Docker Compose 内自动配置）=====
# 本地开发（非 Docker）：用自己的 PG 实例
# DATABASE_URL=postgresql+asyncpg://postgres:postgres@localhost:5432/rednote_insight
#
# Docker Compose 部署：自动通过 docker-compose.yml 注入
# DATABASE_URL=postgresql+asyncpg://postgres:postgres@db:5432/rednote_insight

# ===== Redis（Docker Compose 内自动配置）=====
# 本地开发（非 Docker）：
# REDIS_URL=redis://localhost:6379/0
#
# Docker Compose 部署：自动通过 docker-compose.yml 注入
# REDIS_URL=redis://redis:6379/0

# ===== LangFuse 可观测性（可选）=====
# 注册 https://cloud.langfuse.com 获取
# LANGFUSE_PUBLIC_KEY=pk-...
# LANGFUSE_SECRET_KEY=sk-...
# LANGFUSE_HOST=https://cloud.langfuse.com

# ===== 限流配置（可选）=====
# RATE_LIMIT_ENABLED=true
# RATE_LIMIT_INSIGHT=10/minute
# RATE_LIMIT_QA=20/minute
# RATE_LIMIT_CRAWL=5/minute

# ===== 小红书 Cookie（Streamlit Cloud 部署用）=====
# 在本地运行: uv run python scripts/export_cookies.py
# 复制输出的 JSON 粘贴到 Streamlit Cloud Secrets → XHS_COOKIES
# 注意: 本地 .env 中不需要此项，仅用于 Streamlit Cloud