relationalai.experimental.solvers.Provider

Signature

class Provider

The Provider class is used to manage solver engines using the Python API.

Example

Create an instance of the Provider class and use it to create a new solver engine:

from relationalai.experimental import solvers


# Create a Provider instance.
provider = solvers.Provider()

# Create a solver engine named "my_solver_engine".
provider.create_solver("my_solver_engine")

Methods

The Provider class has methods for viewing and managing solver engines and their jobs.

.__init__()

Signature

Provider.__init__(self) -> None

Constructs a new Provider instance.

For example:

from relationalai.experimental import solvers


# Create a Provider instance.
provider = solvers.Provider()

.create_solver()

Signature

Provider.create_solver(
    name: str,
    size: str,
    settings: dict[str, Any] | None = None,
    auto_suspend_mins: int | None = None,
) -> None

Instruct the RAI Native App to provision a new solver engine with the specified name and size, and block until the engine is ready. Note that the name is case-sensitive and must be unique across all solver engines in the RAI Native App. The size parameter specifies the instance family of the Snowflake SPCS compute pool on which the engine is hosted. Currently, the following sizes are supported:

• "HIGHMEM_X64_S"
• "HIGHMEM_X64_M"

Resource

Refer to the Snowflake docs for memory and CPU specifications of these instance families.

Use the optional settings dictionary to enable or disable specific solver backends on the engine. If None, the following defaults are used:

{
    "gurobi": False,  # Disabled by default
    "highs": True,
    "ipopt": True,
    "minizinc": True,
}

Open-source backends are all enabled by default. Gurobi, which is commercial and requires a license, is disabled by default. You can enable it by passing a dictionary with the "gurobi" key set to True:

from relationalai.experimental import solvers


# Create a Provider instance.
provider = solvers.Provider()

# Create a new solver engine with Gurobi enabled.
provider.create_solver("my_solver_engine", settings={"gurobi": True})

The optional auto_suspend_mins specifies the number of minutes after which the engine should automatically suspend itself if it is idle. If None, the engine will auto-suspend after 60 minutes of inactivity. Set to 0 to disable auto-suspend and keep the engine running indefinitely:

# Create a new solver engine that never suspends.
provider.create_solver("my_solver_engine", auto_suspend_mins=0)

.create_solver_async()

Signature

Provider.create_solver_async(
    name: str,
    size: str,
    settings: dict[str, Any] | None = None,
    auto_suspend_mins: int | None = None,
) -> None

Asynchronously instruct the RAI Native App to provision a new solver engine with the specified name and size. This method returns immediately and does not block until the engine is ready. The parameters are the same as for .create_solver().

.list_solvers()

Signature

Provider.list_solvers(
    self,
    state: str | None = None
) -> list[dict[str, Any]]

List all solver engines available in the RAI Native App. If state is specified, only engines in that state are returned. Possible values for state are:

• "PENDING": The engine is being created and is not yet ready.
• "READY": The engine is ready to use and can run jobs.
• "SUSPENDED": The engine is suspended and not running.

The state parameter is case-insensitive. For example, to list all solver engines that are currently ready:

from pprint import pprint

from relationalai.experimental import solvers


# Create a Provider instance.
provider = solvers.Provider()

# List all solver engines.
pprint(provider.list_solvers(state="ready"))

output

Output:
[{'auto_suspend_mins': 60,
  'created_by': 'jl.picard@starfleet.org',
  'created_on': datetime.datetime(2025, 6, 14, 15, 7, 5, 164000, tzinfo=<DstTzInfo 'America/Los_Angeles' PST-1 day, 16:00:00 STD>),
  'id': '2150b30323a3',
  'name': 'make_it_solve',
  'size': 'HIGHMEM_X64_S',
  'solvers': ['gurobi', 'minizinc'],
  'state': 'READY',
  'updated_on': datetime.datetime(2025, 6, 27, 4, 47, 40, 988000, tzinfo=<DstTzInfo 'America/Los_Angeles' PST-1 day, 16:00:00 STD>)},
 {'auto_suspend_mins': 60,
  'created_by': 'catherine.janeway@starfleet.org',
  'created_on': datetime.datetime(2025, 6, 27, 5, 38, 32, 871000, tzinfo=<DstTzInfo 'America/Los_Angeles' PST-1 day, 16:00:00 STD>),
  'id': 'ee88b24407c1',
  'name': 'my_solver_engine',
  'size': 'HIGHMEM_X64_S',
  'solvers': ['highs', 'minizinc', 'ipopt'],
  'state': 'READY',

.get_solver()

Signature

Provider.get_solver(name: str) -> dict[str, Any]

Get the details of the solver engine with the specified name. Returns a dictionary with the engine’s details, including its state, size, and settings.

For example:

from pprint import pprint

from relationalai.experimental import solvers


# Create a Provider instance.
provider = solvers.Provider()

# Get details for a specific solver engine.
pprint(provider.get_solver("my_solver_engine"))

output

{'auto_suspend_mins': 60,
 'created_by': 'catherine.janeway@starfleet.org',
 'created_on': datetime.datetime(2025, 6, 27, 5, 38, 32, 871000, tzinfo=<DstTzInfo 'America/Los_Angeles' PST-1 day, 16:00:00 STD>),
 'id': 'ee88b24407c1',
 'name': 'my_solver_engine',
 'size': 'HIGHMEM_X64_S',
 'solvers': ['highs', 'minizinc', 'ipopt'],
 'state': 'READY',
 'updated_on': datetime.datetime(2025, 6, 27, 5, 38, 32, 871000, tzinfo=<DstTzInfo 'America/Los_Angeles' PST-1 day, 16:00:00 STD>)}

.resume_solver_async()

Signature

Provider.resume_solver_async(name: str) -> None

Resume the solver engine with the specified name. This method does not block and returns immediately. If the engine is suspended, then it is resumed. Otherwise, this method has no effect.

.delete_solver()

Signature

Provider.delete_solver(name: str) -> None

Delete the solver engine with the specified name. Running jobs are cancelled and terminated. For example, to delete a solver engine named "my_solver_engine":

from relationalai.experimental import solvers


# Create a Provider instance.
provider = solvers.Provider()

# Delete the solver engine named "my_solver_engine".
provider.delete_solver("my_solver_engine")

.list_jobs()

Signature

Provider.list_jobs(
    self,
    state: str | None = None,
    limit: int | None = None
) -> list[dict[str, Any]]

Returns a list of all jobs submitted to the solver engines in the RAI Native App. Each job is represented as a dictionary containing its details, such as job ID, state, and creation time. Jobs are sorted by creation time in descending order.

For example:

from pprint import pprint

from relationalai.experimental import solvers


# Create a Provider instance.
provider = solvers.Provider()

# View all solver engine jobs across the RAI Native App.
pprint(provider.list_jobs())

output

[{'created_by': 'jl.picard@starfleet.org',
  'created_on': datetime.datetime(2025, 6, 27, 5, 46, 46, 887000, tzinfo=<DstTzInfo 'America/Los_Angeles' PST-1 day, 16:00:00 STD>),
  'duration': 14836,
  'engine': 'make_it_solve',
  'finished_at': datetime.datetime(2025, 6, 27, 5, 47, 1, 723000, tzinfo=<DstTzInfo 'America/Los_Angeles' PST-1 day, 16:00:00 STD>),
  'id': '01bd4f7a-0105-c8b9-0003-0acf0aaf4ca6',
  'solver': 'minizinc',
  'state': 'COMPLETED'},
 {'created_by': 'catherine.janeway@starfleet.org',
  'created_on': datetime.datetime(2025, 6, 27, 4, 19, 47, 432000, tzinfo=<DstTzInfo 'America/Los_Angeles' PST-1 day, 16:00:00 STD>),
  'duration': 401,
  'engine': 'my_solver_engine',
  'finished_at': datetime.datetime(2025, 6, 27, 4, 19, 47, 833000, tzinfo=<DstTzInfo 'America/Los_Angeles' PST-1 day, 16:00:00 STD>),
  'id': '01bd4f23-0105-c682-0003-0acf0aae4866',
  'solver': 'highs',
  'state': 'COMPLETED'}]

If the optional state parameter is specified, only jobs in that state are returned. Possible values for state are:

• "CREATED"
• "QUEUED"
• "RUNNING"
• "COMPLETED"
• "CANCELLED"
• "FAILED"

The optional limit parameter specifies the maximum number of jobs to return. If None, all jobs are returned.

.get_job()

Signature

Provider.get_job(job_id: str) -> dict[str, Any]

Returns a dictionary containing the details of the job with the specified job_id. The dictionary includes information such as job state, creation time, and the solver engine it was submitted to.

For example:

from pprint import pprint

from relationalai.experimental import solvers


# Create a Provider instance.
provider = solvers.Provider()

# View details for a specific job.
pprint(provider.get_job("01bd4f23-0105-c682-0003-0acf0aae4866"))

output

{'created_by': 'catherine.janeway@starfleet.org',
 'created_on': datetime.datetime(2025, 6, 27, 4, 19, 47, 432000, tzinfo=<DstTzInfo 'America/Los_Angeles' PST-1 day, 16:00:00 STD>),
 'duration': 401,
 'engine': 'my_solver_engine',
 'finished_at': datetime.datetime(2025, 6, 27, 4, 19, 47, 833000, tzinfo=<DstTzInfo 'America/Los_Angeles' PST-1 day, 16:00:00 STD>),
 'id': '01bd4f23-0105-c682-0003-0acf0aae4866',
 'solver': 'highs',
 'state': 'COMPLETED'}

.cancel_job()

Signature

Provider.cancel_job(job_id: str) -> None

Asynchronously cancels the job with the specified job_id. If the job is running, it is terminated. If the job is already completed or cancelled, this method has no effect.