[project]
name = "ssearch"
version = "0.1.0"
description = "Semantic search over a personal journal archive and a clippings library (RAG: LlamaIndex + HuggingFace embeddings + Ollama/OpenAI)."
readme = "README.md"
requires-python = ">=3.12"
license = { text = "MIT" }
authors = [{ name = "E. M. Furst" }]

dependencies = [
    # --- RAG framework (version-sensitive: pinned exactly) ---
    "llama-index-core==0.14.14",
    "llama-index-embeddings-huggingface==0.6.1",
    "llama-index-readers-file==0.5.6",
    "llama-index-llms-ollama==0.9.1",
    "llama-index-llms-openai==0.6.18",
    "llama-index-retrievers-bm25==0.6.5",
    "llama-index-vector-stores-chroma==0.5.5",
    # --- Vector store (clippings) ---
    "chromadb==1.5.1",
    # --- Embeddings / cross-encoder re-ranking (pulls torch) ---
    "sentence-transformers>=5.1.0",
    # transformers + huggingface-hub pinned: a transformers update once broke
    # offline model loading (AutoTokenizer phoning home). See devlog 2026-02-20.
    "transformers==4.55.0",
    "huggingface-hub==0.34.4",
    # --- Keyword search / document loaders ---
    "nltk>=3.9.1",
    "beautifulsoup4>=4.13.4",
    "striprtf>=0.0.26",
    "pypdf>=6.7.1",
]

[dependency-groups]
# Analysis + exploration notebooks (not needed to run the search pipeline).
# Install with: uv sync --group notebook
notebook = [
    "jupyter>=1.1.1",
    "ipykernel>=6.30.0",
    "ipywidgets>=8.1.0",
    "matplotlib>=3.10.0",
    "seaborn>=0.13.0",
    "pandas>=2.2.0",
    "scikit-learn>=1.7.0",
    "scipy>=1.16.0",
]