legal-rag-starter-kit/.env.hosted.example at feature/retrieval-quality-pass · eliv1982/legal-rag-starter-kit · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
# ============================================================
# legal-rag-starter-kit — hosted OpenAI-compatible profile
# Example only. Copy values into .env when using hosted mode.
# ============================================================

# --- Provider ---
LLM_API_KEY=your-api-key-here
LLM_BASE_URL=https://api.proxyapi.ru/openai/v1

# --- Models ---
RAG_CHAT_MODEL=gpt-4o-mini
RAG_EMBEDDING_MODEL=text-embedding-3-small

# --- Vector DB / corpus version ---
# Keep hosted vector DB separate from local embeddings.
RAG_CHROMA_PATH=runtime/chroma_db_hosted_text_embedding_3_small
RAG_CORPUS_VERSION=hosted-text-embedding-3-small-v1

# --- Retrieval / generation ---
# Retrieval quality
RAG_RAW_TOP_K=10
RAG_MAX_DISTANCE=0.44
RAG_FINAL_TOP_K=5
# Legacy fallback:
# RAG_TOP_K=5
RAG_MAX_TOKENS=1500
RAG_TEMPERATURE=0.2

# --- Chunking / embeddings ---
RAG_CHUNK_SIZE=800
RAG_CHUNK_OVERLAP=200
RAG_MIN_CHUNK_LEN=80
RAG_EMBED_BATCH_SIZE=16

# --- Client behavior ---
OPENAI_TIMEOUT=180
OPENAI_MAX_RETRIES=5
OPENAI_EMBED_RETRIES=5