Context#

Context is a data structure that is used to store information about the current state of a conversation.

It is used to keep track of the user’s input, the current stage of the conversation, and any other information that is relevant to the current context of a dialog.

The Context data structure provides several key features to make working with data easier. Developers can use the context to store any information that is relevant to the current conversation, such as user data, session data, conversation history, e.t.c. This allows developers to easily access and use this data throughout the conversation flow.

Another important feature of the context is data serialization. The context can be easily serialized to a format that can be stored or transmitted, such as JSON. This allows developers to save the context data and resume the conversation later.

class Context(**data)[source]#

Bases: ContextMainInfo

A structure that is used to store data about the context of a dialog.

id: str#

id is the unique context identifier. By default, randomly generated using uuid4 is used.

labels: LabelContextDict#

labels stores dialog labels. A new label is stored in the dictionary on every turn, the keys are consecutive integers. The first ever (initial) has key 0.

  • key - Label identification numbers.

  • value - Label data: AbsoluteNodeLabel.

requests: MessageContextDict#

requests stores dialog requests. A new request is stored in the dictionary on every turn, the keys are consecutive integers. The first ever (initial) has key 1.

  • key - Request identification numbers.

  • value - Request data: Message.

responses: MessageContextDict#

responses stores dialog responses. A new response is stored in the dictionary on every turn, the keys are consecutive integers. The first ever (initial) has key 1.

  • key - Response identification numbers.

  • value - Response data: Message.

_storage: Optional[DBContextStorage] = ModelPrivateAttr()#

Context storage this context is connected to (if any).

async classmethod connected(storage, start_label=None, id=None)[source]#

Create context connected to the given database storage. If context ID is given, the corresponding context is loaded from the database. If the context does not exist in database or ID is None, a new context with new ID is created. A connected context can be later stored in the database.

Parameters:

  • storage (DBContextStorage) – context storage to connect to.

  • start_label (Optional[AbsoluteNodeLabel]) – new context start label (will be set only if the context is created).

  • id (Optional[str]) – context ID.

Return type:

Context

Returns:

context, connected to the database.

async delete()[source]#

Delete connected context from the context storage and disconnect it. Throw an error if the context is not connected. No local context fields will be affected. If the context is not connected, throw a runtime error.

Return type:

None

property last_label: AbsoluteNodeLabel#

Return label with the highest turn id that is present in labels.

This is not always the label of the transition made during the current turn. For that, use ctx.labels[ctx.current_turn_id].

Returns:

Label with the highest turn id.

Raises:

ContextError – If there are no labels.

property last_response: Message | None#

Return response with the highest turn id that is present in responses.

This is not always the response produced during the current turn. For that, use ctx.responses[ctx.current_turn_id].

Returns:

Response with the highest turn id or None if there are no responses.

property last_request: Message#

Return request with the highest turn id that is present in requests.

This is not always the request that initiated the current turn. For that, use ctx.requests[ctx.current_turn_id].

Returns:

Request with the highest turn id.

Raises:

ContextError – If there are no requests.

property pipeline: Pipeline#

Return FrameworkData.pipeline.

property current_node: Node#

Return FrameworkData.current_node.

class _Turns(ctx)[source]#

Bases: object

An instance of class is returned by turns.

This class only defines a __getitem__ method.

Key for the method may be a single value or a slice.

Keep in mind that the turn id 0 is reserved for start_label and does not have a request or response.

If key is a single value, a tuple of the following is returned:

  1. Request that initiated that turn;

  2. Label of the destination made due to the request;

  3. Response generated by the destination node.

If any of the items are missing, they are replaced by None. If key is negative, corresponding turn is counted from the end (e.g. ctx.turns[-1] is the last turn).

If key is a slice, the slice is applied to the range of all turn ids from 0 to the current turn id, and an iterable of tuples described above is returned.

Examples:

  1. await ctx.turns[0] == None, start_label, None;

  2. await ctx.turns[-2] – request, label, response of the second to last turn;

  3. for request, label, response in await ctx.turns[-5:] – iterate over the last 5 turns.

model_post_init(context: Any, /) → None#

We need to both initialize private attributes and call the user-defined model_post_init method.

Return type:

None

property turns: _Turns#

Return a _Turns object used to slice turns in the context. See the Turn class documentation for more details.

_validate_model(handler, _)[source]#
Return type:

Context

async store()[source]#

Store connected context in the context storage. Depending on the context storage settings (“rewrite_existing” flag in particular), either only write new and deleted values or also modify the changed ones. All the context storage tables are updated asynchronously and simultaneously.

Return type:

None