legal-rag-starter-kit/.env.local.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
39
40
41
42
43
# ============================================================
# legal-rag-starter-kit — local Ollama profile
# Default profile: fast local mode
# ============================================================

# --- Provider ---
LLM_API_KEY=local-placeholder
LLM_BASE_URL=http://localhost:11434/v1

# --- Models ---
# Fast local mode:
RAG_CHAT_MODEL=gemma3:4b

# Quality local mode:
# RAG_CHAT_MODEL=deepseek-r1:8b

RAG_EMBEDDING_MODEL=bge-m3

# --- Vector DB / corpus version ---
# Keep a separate Chroma path and corpus version for each embedding model.
RAG_CHROMA_PATH=runtime/chroma_db_local_bge_m3
RAG_CORPUS_VERSION=local-bge-m3-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=1

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