Make sure the given string is json format and contains the 3 keys: [“encoder_name”, “encoder_type”, “routes”]
Generates a RouterConfig object that can be used for initializing routers.
Initialize a RouterConfig object.
Arguments:
routes
(List[Route]
): A list of routes.encoder_type
(str
): The type of encoder to use.encoder_name
(Optional[str]
): The name of the encoder to use.Initialize a RouterConfig from a file. Expects a JSON or YAML file with file
extension .json, .yaml, or .yml.
Arguments:
path
(str
): The path to the file to load the RouterConfig from.Initialize a RouterConfig from a list of tuples of routes and
utterances.
Arguments:
route_tuples
(List[Tuple[str, str]]
): A list of tuples, each containing a route name and an
associated utterance.encoder_type
(str, optional
): The type of encoder to use, defaults to “openai”.encoder_name
(Optional[str], optional
): The name of the encoder to use, defaults to None.Initialize a RouterConfig from a BaseIndex object.
Arguments:
index
(BaseIndex
): The index to initialize the RouterConfig from.encoder_type
(str, optional
): The type of encoder to use, defaults to “openai”.encoder_name
(Optional[str], optional
): The name of the encoder to use, defaults to None.Convert the RouterConfig to a dictionary.
Returns:
Dict[str, Any]
: A dictionary representation of the RouterConfig.
Save the routes to a file in JSON or YAML format.
Arguments:
path
(str
): The path to save the RouterConfig to.Convert the routes to a list of Utterance objects.
Returns:
List[Utterance]
: A list of Utterance objects.
Add a route to the RouterConfig.
Arguments:
route
(Route
): The route to add.Get a route from the RouterConfig by name.
Arguments:
name
(str
): The name of the route to get.Returns:
Optional[Route]
: The route if found, otherwise None.
Remove a route from the RouterConfig by name.
Arguments:
name
(str
): The name of the route to remove.Get the hash of the RouterConfig. Used for syncing.
Returns:
ConfigParameter
: The hash of the RouterConfig.
Reshape the query vector to be a 2D numpy array.
Arguments:
xq
(List[float] | np.ndarray
): The query vector.Returns:
np.ndarray
: The reshaped query vector.
Base class for all routers.
Initialize a BaseRouter object. Expected to be used as a base class only,
not directly instantiated.
Arguments:
encoder
(Optional[DenseEncoder]
): The encoder to use.sparse_encoder
(Optional[SparseEncoder]
): The sparse encoder to use.llm
(Optional[BaseLLM]
): The LLM to use.routes
(Optional[List[Route]]
): The routes to use.index
(Optional[BaseIndex]
): The index to use.Optional[DenseEncoder]
0 (Optional[DenseEncoder]
1): The number of routes to return.Optional[DenseEncoder]
2 (Optional[DenseEncoder]
3): The aggregation method to use.Optional[DenseEncoder]
4 (Optional[DenseEncoder]
5): The auto sync mode to use.Check for a matching route in the routes list.
Arguments:
top_class
(str
): The top class to check for.Returns:
Optional[Route]
: The matching route if found, otherwise None.
Call the router to get a route choice.
Arguments:
text
(Optional[str]
): The text to route.vector
(Optional[List[float] | np.ndarray]
): The vector to route.simulate_static
(bool
): Whether to simulate a static route.route_filter
(Optional[List[str]]
): The route filter to use.limit
(int | None
): The number of routes to return, defaults to 1. If set to None, no
limit is applied and all routes are returned.Returns:
Optional[str]
0: The route choice.
Asynchronously call the router to get a route choice.
Arguments:
text
(Optional[str]
): The text to route.vector
(Optional[List[float] | np.ndarray]
): The vector to route.simulate_static
(bool
): Whether to simulate a static route (ie avoid dynamic route
LLM calls during fit or evaluate).route_filter
(Optional[List[str]]
): The route filter to use.Returns:
RouteChoice
: The route choice.
Runs a sync of the local routes with the remote index.
Arguments:
sync_mode
(str
): The mode to sync the routes with the remote index.force
(bool, optional
): Whether to force the sync even if the local and remote
hashes already match. Defaults to False.wait
(int
): The number of seconds to wait for the index to be unlocked
before proceeding with the sync. If set to 0, will raise an error if
index is already locked/unlocked.Returns:
List[str]
: A list of diffs describing the addressed differences between
the local and remote route layers.
Runs a sync of the local routes with the remote index.
Arguments:
sync_mode
(str
): The mode to sync the routes with the remote index.force
(bool, optional
): Whether to force the sync even if the local and remote
hashes already match. Defaults to False.wait
(int
): The number of seconds to wait for the index to be unlocked
before proceeding with the sync. If set to 0, will raise an error if
index is already locked/unlocked.Returns:
List[str]
: A list of diffs describing the addressed differences between
the local and remote route layers.
Load a RouterConfig from a JSON file.
Arguments:
file_path
(str
): The path to the JSON file.Returns:
RouterConfig
: The RouterConfig object.
Load a RouterConfig from a YAML file.
Arguments:
file_path
(str
): The path to the YAML file.Returns:
RouterConfig
: The RouterConfig object.
Create a Router from a RouterConfig object.
Arguments:
config
(RouterConfig
): The RouterConfig object.index
(Optional[BaseIndex]
): The index to use.Add a route to the local SemanticRouter and index.
Arguments:
route
(Route
): The route to add.Add a route to the local SemanticRouter and index asynchronously.
Arguments:
route
(Route
): The route to add.Updates the route specified in name. Allows the update of
threshold and/or utterances. If no values are provided via the threshold or utterances parameters, those fields are not updated. If neither field is provided raises a ValueError.
The name must exist within the local SemanticRouter, if not a KeyError will be raised.
Arguments:
name
(str
): The name of the route to update.threshold
(Optional[float]
): The threshold to update.utterances
(Optional[List[str]]
): The utterances to update.Deletes a route given a specific route name.
Arguments:
route_name
: the name of the route to be deletedDeletes a route given a specific route name asynchronously.
Arguments:
route_name
: the name of the route to be deletedCheck if the local and remote route layer instances are
synchronized.
Returns:
bool
: True if the local and remote route layers are synchronized,
False otherwise.
Check if the local and remote route layer instances are
synchronized asynchronously.
Returns:
bool
: True if the local and remote route layers are synchronized,
False otherwise.
Get the difference between the local and remote utterances. Returns
a list of strings showing what is different in the remote when compared to the local. For example:
[” route1: utterance1”, ” route1: utterance2”, ”- route2: utterance3”, ”- route2: utterance4”]
Tells us that the remote is missing “route2: utterance3” and “route2: utterance4”, which do exist locally. If we see:
[” route1: utterance1”, ” route1: utterance2”, ”+ route2: utterance3”, ”+ route2: utterance4”]
This diff tells us that the remote has “route2: utterance3” and “route2: utterance4”, which do not exist locally.
Arguments:
include_metadata
(bool
): Whether to include metadata in the diff.Returns:
List[str]
: A list of strings showing the difference between the local and remote
utterances.
Get the difference between the local and remote utterances asynchronously.
Returns a list of strings showing what is different in the remote when compared to the local. For example:
[” route1: utterance1”, ” route1: utterance2”, ”- route2: utterance3”, ”- route2: utterance4”]
Tells us that the remote is missing “route2: utterance3” and “route2: utterance4”, which do exist locally. If we see:
[” route1: utterance1”, ” route1: utterance2”, ”+ route2: utterance3”, ”+ route2: utterance4”]
This diff tells us that the remote has “route2: utterance3” and “route2: utterance4”, which do not exist locally.
Arguments:
include_metadata
(bool
): Whether to include metadata in the diff.Returns:
List[str]
: A list of strings showing the difference between the local and remote
utterances.
Get a route by name.
Arguments:
name
(str
): The name of the route to get.Returns:
Optional[Route]
: The route.
Group the scores by class.
Arguments:
query_results
(List[Dict]
): The query results to group. Expected format is a list of
dictionaries with “route” and “score” keys.Returns:
Dict[str, List[float]]
: A dictionary of route names and their associated scores.
Set the score threshold for a specific route or all routes. A threshold
of 0.0
will mean that the route will be returned no matter how low it scores whereas a threshold of 1.0 will mean that a route must contain an exact utterance match to be returned.
Arguments:
threshold
(float
): The threshold to set.route_name
(str | None
): The name of the route to set the threshold for. If None, the
threshold will be set for all routes.Convert the router to a RouterConfig object.
Returns:
RouterConfig
: The RouterConfig object.
Convert the router to a JSON file.
Arguments:
file_path
(str
): The path to the JSON file.Convert the router to a YAML file.
Arguments:
file_path
(str
): The path to the YAML file.Get the score thresholds for each route.
Returns:
Dict[str, float]
: A dictionary of route names and their associated thresholds.
Fit the router to the data. Works best with a large number of examples for each
route and with many None
utterances.
Arguments:
X
(List[str]
): The input data.y
(List[str]
): The output data.batch_size
(int
): The batch size to use for fitting.max_iter
(int
): The maximum number of iterations to use for fitting.local_execution
(X
0): Whether to execute the fitting locally.Evaluate the accuracy of the route selection.
Arguments:
X
(List[str]
): The input data.y
(List[str]
): The output data.batch_size
(int
): The batch size to use for evaluation.Returns:
float
: The accuracy of the route selection.
Performs a random search iteration given a route layer and a search range.
Arguments:
route_layer
(BaseRouter
): The route layer to search.search_range
(Union[int, float]
): The search range to use.Returns:
Dict[str, float]
: A dictionary of route names and their associated thresholds.