Install

A bare install is a lightweight remote SDK — smrititantra.connect(), models, exceptions, and the sync facade. No database drivers, no web framework, no scale backends:

pip install smrititantra

Everything heavier lives behind an extra, enforced by scripts/test-layering.sh so nothing ever lands in the wrong layer:

Extra	Adds	Typical use
local	embedded engine, Postgres store, numpy-backed snapshot compile	running SmritiTantra in-process
server	FastAPI HTTP service, MCP server	running it as a service
s3	S3/MinIO artifact store	snapshot/dataset storage at scale
qdrant	Qdrant vector-index backend	dense recall at scale
neo4j	Neo4j graph-serving tier	graph traversal at scale
mongo	MongoDB source connector	mining a Mongo-backed workload
otel	OpenTelemetry tracing spans	production observability
all	every extra above	development, demos
dev	ruff, mypy, pytest, pre-commit	contributing

Combine what you need:

pip install "smrititantra[local,server]"

Requirements

• Python 3.12+
• Postgres (for the local extra's persistent store — the embedded engine also runs against an in-memory store with no database at all, e.g. for smrititantra demo or tests)

From source

git clone https://github.com/NoesisResearch/smrititantra.git
cd smrititantra
uv venv && uv pip install -e ".[all,dev]"