opvious 0.7.2

Opvious Python SDK

Opvious Python SDK

This package provides a lightweight SDK for solving optimization models with the Opvious API. Its main features are:

• Seamless data import/export via native support for pandas
• Powerful built-in debugging capabilities: automatic infeasibility relaxation, variable pinning, and more
• Non-blocking APIs for performant parallel calls

Quickstart

First, install this package and have an API access token handy (these can be generated here).

pip install opvious[aio] # aio is recommended for improved performance

With these steps out of the way, you are ready to optimize!

import opvious

# Instantiate an API client from an API token
client = opvious.Client.from_token(TOKEN)

# Solve a simple portfolio selection optimization model
outputs = await client.solve(
    sources=[
      r"""
          We find an allocation of assets which minimizes risk while satisfying
          a minimum expected return:

          + A collection of assets: $\S^d_{asset}: A$
          + Covariances: $\S^p_{covariance}: c \in \mathbb{R}^{A \times A}$
          + Expected return: $\S^p_{expectedReturn}: m \in \mathbb{R}^A$
          + Minimum desired return: $\S^p_{desiredReturn}: r \in \mathbb{R}$

          The only output is the allocation per asset
          $\S^v_{allocation}: \alpha \in [0,1]^A$ chosen to minimize risk:
          $\S^o_{risk}: \min \sum_{a, b \in A} c_{a,b} \alpha_a \alpha_b$.

          Subject to the following constraints:

          + $\S^c_{atLeastMinimumReturn}: \sum_{a \in A} m_a \alpha_a \geq r$
          + $\S^c_{totalAllocation}: \sum_{a \in A} \alpha_a = 1$
      """
    ],
    parameters={
        "covariance": {
            ("AAPL", "AAPL"): 0.08,
            # ...
        },
        "expectedReturn": {
            "AAPL": 0.07,
            # ..
        },
        "desiredReturn": 0.05,
    },
)

# Print the optimal allocation, if any
if isinstance(outputs.outcome, opvious.FeasibleOutcome):
  print(outputs.data.variable("allocation"))
else:
  print(f"Problem was {outputs.status}.") # INFEASIBLE, UNBOUNDED

Environments

Clients are compatible with Pyodide environments, for example JupyterLite kernels. Simply install the package as usual in a notebook, omitting the aio optional dependencies:

import piplite
await piplite.install('opvious')

In other environments, prefer using the aiohttp-powered clients as they are more performant (this is the default if the aio dependencies were specified).

Next steps

This SDK is focused on solving optimization models. For convenient access to the rest of Opvious API's functionality, consider using the TypeScript SDK and CLI.