TLDR

Add NL query parameters to search API and complete NL search models CRUD operations.

Change Summary

Added Features:

1. New Natural Language Query Support in src/typesense/types/document.py:

   • NLLanguageParameters: Configuration for natural language queries with caching, model selection, and debug options
   • LLMResponse: Schema for raw LLM responses with content, extraction method, and model info
   • ParsedNLQuery: Schema for processed NL queries with timing, generated/augmented parameters, and LLM response
   • Extended SearchParameters to include NLLanguageParameters
   • Added parsed_nl_query field to SearchResponse for NL query metadata

2. New NL Search Models Management System:

   • NLSearchModel class in src/typesense/nl_search_model.py: Individual model operations (retrieve, update, delete)
   • NLSearchModels class in src/typesense/nl_search_models.py: Collection operations (create, retrieve, getitem)
   • Type definitions in src/typesense/types/nl_search_model.py: Support for multiple LLM providers (OpenAI, Google, GCP Vertex AI, Cloudflare)

Code Changes:

1. In src/typesense/client.py:

   • Integrated NLSearchModels into main Client class with proper imports
   • Added self.nl_search_models = NLSearchModels(self.api_call) to client initialization

2. In tests/fixtures/nl_search_model_fixtures.py:

   • Added test fixtures for both individual and collection NL search model operations
   • Includes cleanup utilities and actual API test fixtures

3. In tests/nl_search_model_test.py and tests/nl_search_models_test.py:

   • Complete test coverage for individual model operations (retrieve, update, delete)
   • Complete test coverage for collection operations (create, retrieve, getitem)
   • Integration tests with actual Typesense server