luplo.core.extract.pipeline
===========================

.. py:module:: luplo.core.extract.pipeline

.. autoapi-nested-parse::

   Extraction pipeline — transcript text to structured items and glossary candidates.

   v0.5 implementation is a stub that returns empty results.  The function
   signatures are the stable contract that the MCP tool ``luplo_save_decisions``
   and the background worker will call.  Post-v0.5, these will use the Claude
   API to parse transcripts into decisions, knowledge, and terminology.



Functions
---------

.. autoapisummary::

   luplo.core.extract.pipeline.extract_decisions
   luplo.core.extract.pipeline.extract_glossary_candidates


Module Contents
---------------

.. py:function:: extract_decisions(text: str, *, project_id: str, actor_id: str, work_unit_id: str | None = None) -> list[luplo.core.models.ItemCreate]
   :async:


   Extract structured decisions from a conversation transcript.

   Scans *text* for decision-like statements and returns ``ItemCreate``
   objects ready for insertion.  The caller (MCP tool or worker) handles
   the actual ``create_item`` calls.

   :param text: Raw transcript or conversation text.
   :param project_id: Project to scope the extracted items to.
   :param actor_id: Actor who authored the transcript.
   :param work_unit_id: Optional work unit to attach items to.

   :returns: always returns ``[]``.**
   :rtype: List of ``ItemCreate`` objects.  **v0.5


.. py:function:: extract_glossary_candidates(text: str, *, project_id: str) -> list[tuple[str, str]]
   :async:


   Extract glossary term candidates from text.

   Returns ``(surface_form, context_snippet)`` pairs.  The worker feeds
   these into the glossary strict-first pipeline for normalisation,
   matching, and human curation.

   :param text: Source text to extract terms from.
   :param project_id: Project scope (for future filtering).

   :returns: always returns ``[]``.**
   :rtype: List of ``(surface, context)`` tuples.  **v0.5