awaredb 1.0.1

Python client for AwareDB — a data-modeling and calculation platform with unit-aware formulas, scenarios, optimization, and reversible history.

awaredb

Python client for AwareDB — a data-modeling and calculation platform with unit-aware formulas, reversible history, and a rich analyses toolbox (scenarios, optimisation, sensitivity, projections, traces).

pip install awaredb

from awaredb import AwareDBClient

with AwareDBClient(db="my_db", token="…") as client:
    client.calculate(formula="car.power * 2")

---

Features

• Connect by token or by username/password.
• Read commands: get, query, calculate.
• Write commands: update, remove, flush, load (bulk-load JSON from disk).
• Analyses: scenarios, impact, projection, goal_seek, optimize, sensitivity, trace.
• History: history against the reversible audit database.
• Typed exceptions: AwareDBError, TransportError, AuthError, InvalidCommandError, UnexpectedError.
• Persistent httpx-backed transport with built-in retries and connection pooling.
• Ships a py.typed marker — works with mypy / pyright out of the box.

---

Requirements

• Python 3.12+
• httpx

---

Quick start

from awaredb import AwareDBClient

# Token auth
client = AwareDBClient(db="my_db", token="…")

# Username / password — fetches a token at construction time
client = AwareDBClient(db="my_db", user="alice", password="…")

# Custom host + timeout, no upfront connection check
client = AwareDBClient(
    db="my_db",
    token="…",
    host="https://aware.example.com",
    timeout=30,
    check_connection=False,
)

# Close the underlying HTTP client when finished
client.close()

# Or use the client as a context manager
with AwareDBClient(db="my_db", token="…") as client:
    client.flush()

---

Read commands

get(path, states=None)

Return the value at a path.

client.get(path="car.power")
# "250 hp"

query(nodes=None, conditions=None, properties=None, states=None, show_abstract=False)

Return nodes matching the filters. nodes defaults to ["*"] (all).

client.query(
    nodes=["employee"],
    conditions=["node.salary.gross > 60000"],
    properties=["salary"],
)

calculate(formula, states=None)

Evaluate one or more formulas against the live database.

client.calculate(formula="car.power * 2")
# "500 hp"

---

Write commands

update(data, partial=False)

Create or update nodes / relations / relation types. data accepts a single item or a list.

client.update([
    {"uid": "fan", "power": "=sum(this.children.power)"},
])

remove(ids)

Delete by id.

client.remove(ids=["7037a8a5-…", "7037a8a5-…"])

flush()

Drop every node and relation in the database.

load(filepath, recursive=False, flush=False)

Push the contents of a JSON file (or folder of JSON files) through update.

client.load("./seed", recursive=True, flush=True)

---

Analyses

All analyses are read-only — they recalculate the graph against a temporary context and return per-variant results. None mutate the database.

scenarios(edits, states=None)

Apply path/value overrides, recalculate, return per-scenario results.

client.scenarios(edits=[{"path": "motor.power", "value": "30 kW"}])

impact(edit, states=None, **extra)

Sweep an input across N steps and observe outputs.

projection(start_from, unit="day", step=1, points=10, edits=None, states=None)

Walk forward in time. Per-step edits apply between points.

goal_seek(target, decision, states=None, **extra)

Find the input value that drives target to a desired value.

optimize(objectives, decisions, constraints=None, states=None, **extra)

Vary decision variables to minimise / maximise objectives subject to constraints.

sensitivity(target, inputs, states=None, **extra)

Perturb each input by ±delta and report the response per variant.

trace(target, depth=5, states=None)

Walk the upstream dependency tree of a target path.

client.trace(target="car.range", depth=10)

---

History

history(change_id=None, ids=None, start=None, end=None, from_date=None, to_date=None)

List changes recorded in the reversible history database.

from datetime import datetime

client.history(from_date=datetime(2026, 1, 1))

---

Errors

Everything funnels through AwareDBError. Subclasses indicate the failure mode:

Exception	When it fires
AuthError	401 / 403, or username/password rejected.
InvalidCommandError	400 — bad payload or unknown command.
TransportError	Network failure or other non-2xx HTTP status.
UnexpectedError	2xx body that isn't JSON.
AwareDBError	Base class. Catch this to handle everything above.

---

Development

uv drives dependency management. The lockfile (uv.lock) is committed and pins exact versions.

# One-time: install uv
brew install uv                            # macOS / Linuxbrew
# or:
curl -LsSf https://astral.sh/uv/install.sh | sh

# Provision the project env (creates .venv from uv.lock — runtime + dev group)
uv sync

# Runtime-only (excludes dev group: pytest, ruff, mypy, …)
uv sync --no-dev

# Tests, lint, format, type check
uv run pytest
uv run ruff check awaredb/ tests/
uv run ruff format --check awaredb/ tests/
uv run mypy awaredb/

# Add / remove deps (auto-updates pyproject.toml + uv.lock)
uv add some-package
uv add --group dev some-package
uv remove some-package

Note: python -m build is the pip-era invocation and requires the build package installed in the env. With uv, use uv build instead — it uses uv's own PEP 517 frontend, no extra dependency needed.

---

Releasing

Releases are cut by pushing a v* git tag. The .github/workflows/release.yml workflow builds the sdist + wheel with uv build and uploads them to PyPI via Trusted Publishing (OIDC, no token in CI).

One-time PyPI setup

1. Create the project on PyPI: https://pypi.org/manage/account/publishing/.
2. Add a pending trusted publisher with:
   • PyPI Project Name: awaredb
   • Owner: djinalab
   • Repository name: awaredb-python
   • Workflow name: release.yml
   • Environment name: pypi
3. In GitHub repo settings → Environments, create an environment named pypi (optionally with required reviewers for an approval gate).

Cutting a release

# 1. Bump version in awaredb/__init__.py (e.g. 0.1.0 → 0.1.1)
# 2. Move the [Unreleased] block in CHANGELOG.md under a new [X.Y.Z] - YYYY-MM-DD heading
# 3. Commit, push to main
git commit -am "Release 0.1.1"
git push

# 4. Tag and push — release.yml fires automatically
git tag v0.1.1
git push origin v0.1.1

Manual publish (fallback)

If you need to publish from a workstation:

uv build                                   # writes sdist + wheel to dist/
uv publish                                 # uses UV_PUBLISH_TOKEN env var
# or:
uv publish --token "$PYPI_API_TOKEN"

The dist/ folder is gitignored — never commit build artifacts.

See CONTRIBUTING.md and CHANGELOG.md.