.env.example

# ── OpenAI ────────────────────────────────────────────────────────────────────
OPENAI_API_KEY=

# ── Models ────────────────────────────────────────────────────────────────────
MODEL_NAME=gpt-3.5-turbo
EMBEDDING_MODEL=text-embedding-ada-002

# LLM sampling temperatures (0 = deterministic, higher = more creative)
TEMPERATURE=0.7          # main answer LLM
QUERY_TEMPERATURE=0.0    # query rewriting LLM
SUMMARY_TEMPERATURE=0.2  # summarisation LLM

# Token limits per LLM call
MAX_TOKENS=500
SUMMARY_MAX_TOKENS=300

# ── Text processing ───────────────────────────────────────────────────────────
CHUNK_SIZE=1000
CHUNK_OVERLAP=200
SNIPPET_MAX_CHARS=1200   # max chars from a chunk shown in the prompt

# ── Retrieval ─────────────────────────────────────────────────────────────────
RETRIEVAL_K=6            # FAISS candidates per query
BM25_K=6                 # BM25 candidates per query
FINAL_K=6                # top-K docs sent to the LLM
MULTIQUERY_COUNT=3       # extra query variants for multi-query retrieval
MAX_CONTEXT_CHARS=8000   # total character budget for the context block

ENABLE_QUERY_REWRITE=true
ENABLE_MULTIQUERY=true
ENABLE_COMPRESSION=true

# ── Persistence ───────────────────────────────────────────────────────────────
INDEX_DIR=data/indexes

# ── Summarisation ─────────────────────────────────────────────────────────────
SUMMARY_MAX_CHUNKS=6