Mahdi Bazrafshan
55907a8dec
Why: - Support Text PDFs in the same DocumentTree/markdown contract as DOCX. Changes: - PyMuPDF parser, text-layer gate, shared heading heuristics, upload dispatch for .pdf. Impact: - Scanned/image PDFs are rejected at upload; requires pymupdf installed. Co-authored-by: Cursor <cursoragent@cursor.com>
32 lines
645 B
Plaintext
32 lines
645 B
Plaintext
# OpenAI API authentication (required)
|
|
OPENAI_API_KEY=sk-...
|
|
|
|
# Qdrant configuration
|
|
QDRANT_URL=http://localhost:6333
|
|
QDRANT_API_KEY=
|
|
|
|
# Fixed models (not configurable per strategy)
|
|
EMBEDDING_MODEL=text-embedding-3-small
|
|
LLM_MODEL=gpt-4o-mini
|
|
|
|
# Retrieval defaults
|
|
TOP_K=5
|
|
|
|
# LLM generation parameters
|
|
TEMPERATURE=0.0
|
|
MAX_TOKENS=1024
|
|
|
|
# Chunking defaults
|
|
CHUNK_SIZE=512
|
|
CHUNK_OVERLAP=50
|
|
|
|
# Semantic chunking parameters
|
|
SEMANTIC_THRESHOLD=0.3
|
|
SEMANTIC_MIN_CHUNK_SIZE=3
|
|
|
|
# SQLite database path
|
|
DATABASE_URL=sqlite:///./data/chunking_benchmark.db
|
|
# Text PDF gate (reject scanned/image PDFs)
|
|
PDF_MIN_TOTAL_CHARS=100
|
|
PDF_MIN_MEDIAN_CHARS_PER_PAGE=40
|