julee.services

Services for julee domain.

This module provides service classes and factory functions for interacting with external services in the Capture, Extract, Assemble, Publish workflow.

The services layer handles external integrations while the repository layer handles local metadata persistence. Services are organized by service type into submodules, each with their own protocols and implementations.

Submodules

Classes

KnowledgeService

Protocol for interacting with external knowledge services.

Package Contents

class julee.services.KnowledgeService[source]

Bases: Protocol

Protocol for interacting with external knowledge services.

This protocol defines the interface for external operations that were moved out of the repository layer. Implementations handle the specifics of different knowledge service APIs (Anthropic, OpenAI, etc.).

async execute_query(config, query_text, output_schema=None, service_file_ids=None, query_metadata=None, assistant_prompt=None)[source]

Execute a query against the external knowledge service.

This method executes a text query against the knowledge service, optionally scoping the query to specific documents that have been previously registered with the service.

Parameters:

  • config (julee.domain.models.knowledge_service_config.KnowledgeServiceConfig) – KnowledgeServiceConfig for the service to use

  • query_text (str) – The query to execute (natural language or structured)

  • output_schema (dict[str, Any] | None) – Optional JSON schema for structured response. When provided, the service will attempt to return results conforming to this schema using structured outputs or schema-guided prompting.

  • service_file_ids (list[str] | None) – Optional list of service file IDs to provide as context for the query. These are the IDs returned by the knowledge service from register_file operations, and are included in the query to give the service access to specific documents.

  • query_metadata (dict[str, Any] | None) – Optional service-specific metadata and configuration options such as model selection, temperature, max_tokens, etc. The structure depends on the specific knowledge service being used.

  • assistant_prompt (str | None) – Optional assistant message content to constrain or prime the model’s response. This is added as the final assistant message before the model generates its response, allowing control over response format and structure.

Returns:

QueryResult containing query results and execution metadata

Return type:

QueryResult

Implementation Notes

  • Must be idempotent: same query returns consistent results

  • Service file IDs are provided as context to enhance query responses

  • Should handle service unavailability gracefully

  • Query results should be structured as domain objects

  • Should track execution time and metadata

  • Must handle various query formats (natural language, structured, etc.)

  • Should validate that service_file_ids exist in the service before including them in the query context

Workflow Context

In Temporal workflows, this method is implemented as an activity to ensure query results are durably stored and can be replayed consistently.

async register_file(config, document)[source]

Register a document file with the external knowledge service.

This method registers a document with the external knowledge service, allowing that service to analyze and index the document content for future queries.

Parameters:
Returns:

FileRegistrationResult containing registration details and the service’s internal file identifier

Return type:

FileRegistrationResult

Implementation Notes

  • Must be idempotent: re-registering same document returns same result

  • Should handle service unavailability gracefully

  • Must return the service’s internal file ID for future queries

  • Document content is accessed directly from the Document object

  • Should handle various document formats and sizes

Workflow Context

In Temporal workflows, this method is implemented as an activity to ensure registration results are durably stored and consistent across workflow replays.