Llm/integration tests fixes

[nuwangeek]

This pull request refactors how integration tests handle Vault secrets and connection IDs for LLM and embedding models. The changes ensure that test secrets, database rows, and API requests all use a single, deterministic vault_uuid value. This removes previous mismatches and makes the test environment more robust and less dependent on database-assigned IDs. Additionally, the Vault token and file system mounts in the test Docker Compose setup are updated for consistency.

Key changes:

Secret management and test consistency

• All LLM and embedding Vault secrets for integration tests now use a fixed TEST_VAULT_UUID instead of environment-based or database-generated IDs, ensuring that secrets, DB rows, and API requests are aligned. (tests/integration_tests/conftest.py, DSL/Ruuter.private/rag-search/POST/inference/test.yml, [1] [2] [3] [4] [5] [6] [7] [8] [9]
• The integration test setup forcibly updates the vault_uuid field in the relevant test database row to match TEST_VAULT_UUID by introducing a new _pin_vault_uuid helper. (tests/integration_tests/conftest.py, [1] [2]

Docker Compose test environment improvements

• Docker Compose mounts for the Vault agent and LLM orchestrator are updated so that the Vault token is consistently accessible at /agent/llm-token/token, matching the default expectation of the code. (docker-compose-test.yml, [1] [2]

Model and API changes

• The connectionId field in the test request model is changed from Optional[int] to Optional[str] to reflect the use of UUIDs instead of integer IDs. (src/models/request_models.py, src/models/request_models.pyL272-R274)

Miscellaneous

• The PostgreSQL client in tests now sets the search path to include the rag_search schema, ensuring unqualified table queries work after a schema migration. (tests/integration_tests/conftest.py, tests/integration_tests/conftest.pyR1185-R1188)
• Minor cleanup in test imports. (tests/integration_tests/test_indexing.py, tests/integration_tests/test_indexing.pyL11-L20)