The index.py module#

Summary#

ApiEntry

One row of api_objects.json.

ApiSearchHit

One search hit from the index.

ApiIndex

Lightweight searchable index over PyFluent’s api_objects.json.

get_default_api_index

Return the default API index.

reset_default_api_index

Test hook that drops the cached index so that the next call rebuilds.

Description#

Lexical fallback over the PyFluent API catalog (the default in practice).

This module is the SOLE actively-running retriever for every default install.

It reads ansys/fluent/core/generated/api_tree/api_objects.json (shipped with ansys-fluent-core) and offers three operations used by orchestration:

  • ApiIndex.search(): BM25 ranking over the union of path tokens and PyFluent class docstring tokens. (See ansys.fluent.mcp.common.api_help). Cheap, dependency free, and much stronger than the previous token-overlap scorer for free-text queries. For example, “temperature of incoming gas” → velocity_inlet.thermal.t now works because the leaf class’s docstring contains the word temperature.

  • ApiIndex.lookup(): Exact dotted-path resolution.

  • ApiIndex.children_of(): Immediate API children of a path For example, used to walk a boundary condition kind to enumerate its property leaves.

The scoring formula follows. Also see _score_entry()):

bm25(query_tokens, entry_tokens)
  + 0.25 * (#query_token substring matches in path)
  - 0.05 * (path depth in dots)

The substring bonus preserves the historic behavior where exact segment matches (energy → setup.models.energy) outrank deeper family matches. The depth penalty discourages picking generic ancestors over leaf parameters.

Module detail#

index.get_default_api_index() → ApiIndex#

Return the default API index.

Returns:
ApiIndex

API index produced by the operation.

index.reset_default_api_index() → None#

Test hook that drops the cached index so that the next call rebuilds.

Returns:
None

The function completes through its side effects.