pyproject.toml

[project]
name = "tribrid-rag"
version = "0.1.0"
description = "Tri-brid RAG engine combining vector, sparse, and graph search"
readme = "README.md"
requires-python = ">=3.11,<3.13"
license = { file = "LICENSE" }
authors = [
    { name = "TriBridRAG Team" }
]

dependencies = [
    # FastAPI
    "fastapi>=0.110.0",
    "uvicorn[standard]>=0.27.0",
    "python-multipart>=0.0.9",
    # MCP (Model Context Protocol) - official Python SDK
    "mcp>=1.26.0",
    # Database
    "asyncpg>=0.29.0",
    "pgvector>=0.2.4",
    "neo4j>=5.15.0",
    # AI/ML
    "openai>=1.12.0",
    "cohere>=4.47.0",
    "sentence-transformers>=2.3.0",
    "transformers>=4.37.0",
    "torch>=2.1.0",
    "einops>=0.8.0",
    "tiktoken>=0.6.0",
    # Data Processing
    "pydantic>=2.5.0",
    "pydantic-settings>=2.1.0",
    # Observability
    "prometheus-client>=0.19.0",
    "opentelemetry-api>=1.22.0",
    "opentelemetry-sdk>=1.22.0",
    "opentelemetry-exporter-otlp>=1.22.0",
    "opentelemetry-instrumentation-fastapi>=0.43b0",
    "langfuse>=4.7.0",
    # Utilities
    "httpx>=0.26.0",
    "python-dotenv>=1.0.0",
    # Repair-parse malformed JSON replies from the vision alias at index time
    "json-repair>=0.44.1",
    # Document extraction (Parquet ingestion; rich documents go through Docling)
    "pandas>=2.2.0",
    "pyarrow>=15.0.0",
    # Retrieval/indexing lane: Haystack + Docling + Qdrant
    "haystack-ai>=2.26.0",
    "qdrant-client>=1.17.0",
    "qdrant-haystack>=10.3.0",
    "docling>=2.81.0",
    "pypdfium2>=5.6.0",
    "pillow>=12.1",
    "neo4j>=5.28.4,<6.0.0",
    "neo4j-graphrag[openai]==1.19.0",
    "fastembed-haystack>=2.6.0",
    "ragas>=0.3.1",
    "langchain-community<1",
    "langchain-core<1",
    "langchain<1",
    "pyroscope-io>=1.2.1",
]

[project.optional-dependencies]
dev = [
    # Test fixture that synthesizes XLSX inputs for the Docling extractor tests
    "openpyxl>=3.1.5",
    "pytest>=8.0.0",
    "pytest-asyncio>=0.23.0",
    "pytest-cov>=4.1.0",
    "httpx>=0.26.0",
    "ruff>=0.2.0",
    "mypy>=1.8.0",
    "datamodel-code-generator>=0.25.0",
]

flyte = [
    # Registration-time only: packages infra/flyte/workflows into the Flyte control plane
    "flytekit>=1.16,<2",
]

mlx = [
    "mlx>=0.12.0",
    "mlx-lm>=0.17.0",
    "mlx-data>=0.0.6",
    "sentencepiece>=0.2.0",
    "datasets>=2.20.0",
    "mlx-embeddings>=0.0.5",
]

[build-system]
requires = ["hatchling"]
build-backend = "hatchling.build"

[tool.hatch.build.targets.wheel]
packages = ["server"]

[tool.uv]
dev-dependencies = [
    "pytest>=8.0.0",
    "pytest-asyncio>=0.23.0",
    "pytest-cov>=4.1.0",
    "ruff>=0.2.0",
    "mypy>=1.8.0",
    "datamodel-code-generator>=0.25.0",
]

[tool.ruff]
line-length = 100
target-version = "py311"

[tool.ruff.lint]
select = ["E", "F", "I", "N", "W", "UP", "B"]
ignore = ["E501"]

[tool.mypy]
python_version = "3.11"
strict = true
warn_return_any = true
warn_unused_ignores = true

[[tool.mypy.overrides]]
module = [
    "asyncpg",
    "asyncpg.*",
    "pgvector",
    "pgvector.*",
    "openpyxl",
    "openpyxl.*",
    "pyarrow",
    "pyarrow.*",
    "mlx",
    "mlx.*",
    "mlx_embeddings",
    "mlx_embeddings.*",
    "mlx_lm",
    "mlx_lm.*",
    "neo4j_graphrag",
    "neo4j_graphrag.*",
    "sentence_transformers",
    "sentence_transformers.*",
    "pypdfium2",
    "pyroscope",
    "ragas",
    "ragas.*",
]
ignore_missing_imports = true

[tool.pytest.ini_options]
asyncio_mode = "auto"
asyncio_default_fixture_loop_scope = "session"
asyncio_default_test_loop_scope = "session"
testpaths = ["tests"]
addopts = "-v --cov=server --cov-report=term-missing"
markers = [
    "integration: exercises a cross-component integration contract",
    "requires_postgres: requires a live authenticated PostgreSQL connection",
    "requires_neo4j: requires a live authenticated Neo4j connection",
    "requires_pg_search: requires the optional ParadeDB pg_search extension",
    "requires_qdrant: requires a live Qdrant vector-store service",
    "requires_flyte: requires the live Compose-owned Flyte control plane",
]