Function _cleanup_pageindex

openkb/cli.py:1148–1187

Drop a long-doc entry from PageIndex's local SQLite + remove its managed files. Returns ``(did_cleanup, message)``. No-op (returns ``(False, "no PageIndex state")``) when no ``pageindex.db`` exists — short-doc-only KBs never created any. Falls back to matching by ``doc_name`` via `

(
    openkb_dir: Path,
    kb_dir: Path,
    doc_name: str,
    doc_id: str | None,
)

def _cleanup_pageindex(
    openkb_dir: Path,
    kb_dir: Path,
    doc_name: str,
    doc_id: str | None,
) -> tuple[bool, str]:
    """Drop a long-doc entry from PageIndex's local SQLite + remove its
    managed files. Returns ``(did_cleanup, message)``.

    No-op (returns ``(False, "no PageIndex state")``) when no
    ``pageindex.db`` exists — short-doc-only KBs never created any.

    Falls back to matching by ``doc_name`` via ``list_documents()`` when
    the registry entry pre-dates PR #51's ``doc_id`` field. Ambiguous
    multi-match cases are skipped with a warning rather than guessed.
    """
    if not (openkb_dir / "pageindex.db").exists():
        return False, "no PageIndex state"

    from pageindex import PageIndexClient

    _setup_llm_key(kb_dir)
    config = load_config(openkb_dir / "config.yaml")
    model = config.get("model", DEFAULT_CONFIG.get("model", "gpt-5.4"))
    client = PageIndexClient(model=model, storage_path=str(openkb_dir))
    col = client.collection()

    if doc_id is None:
        candidates = [d for d in col.list_documents() if d.get("doc_name") == doc_name]
        if not candidates:
            return False, "no PageIndex doc to delete"
        if len(candidates) > 1:
            return False, (
                f"{len(candidates)} PageIndex docs match doc_name='{doc_name}'; "
                "skipping (re-add to refresh)"
            )
        doc_id = candidates[0]["doc_id"]

    col.delete_document(doc_id)
    return True, f"deleted PageIndex doc ({doc_id[:12]}…)"


def _resolve_doc_identifier(registry, identifier: str) -> list[tuple[str, dict]]: