Split 49 modules/suites into independent git repos; untrack from monorepo

Each top-level module/suite folder is now its own private repo on GitHub
(gsinghpal/<name>) and gitea (admin/<name>), with a fresh single initial
commit. The monorepo no longer tracks them (added to .gitignore + git rm
--cached); working-tree files are retained on disk and managed in their
own repos. The monorepo keeps shared root files (CLAUDE.md, docs/, scripts/,
tools/, AGENTS.md, WIP/obsolete dirs) and full history.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

fusion_accounting/fusion_accounting_ai/services/adapters/_base.py

@@ -1,44 +0,0 @@
-"""LLMProvider contract - every adapter must conform.
-
-Phase 1 generalisation: makes local LLM (Ollama, LM Studio, vLLM, llamafile,
-llama.cpp HTTP server) a one-config-line drop-in via the OpenAI-compatible
-HTTP API surface that all of them expose.
-"""
-
-
-class LLMProvider:
-    """Contract every LLM backend must satisfy. Adapters declare capabilities
-    as class attributes; the engine inspects them before calling optional methods."""
-
-    supports_tool_calling: bool = False
-    supports_streaming: bool = False
-    max_context_tokens: int = 4096
-    supports_embeddings: bool = False
-
-    def __init__(self, env):
-        self.env = env
-
-    def complete(self, *, system, messages, max_tokens=2048, temperature=0.0) -> dict:
-        """Plain text completion. Required for ALL providers.
-
-        Returns: {'content': str, 'tokens_used': int, 'model': str}
-        """
-        raise NotImplementedError
-
-    def complete_with_tools(self, *, system, messages, tools, max_tokens=2048) -> dict:
-        """Tool-calling completion. Optional - caller checks supports_tool_calling first.
-
-        Returns: {'content': str, 'tool_calls': [{'name': str, 'arguments': dict}], ...}
-        """
-        raise NotImplementedError(
-            f"{type(self).__name__} does not support tool-calling. "
-            f"Check supports_tool_calling before calling.")
-
-    def embed(self, texts: list[str]) -> list[list[float]]:
-        """Embeddings. Optional - caller checks supports_embeddings first.
-
-        Returns: list of float vectors, one per input text.
-        """
-        raise NotImplementedError(
-            f"{type(self).__name__} does not support embeddings. "
-            f"Check supports_embeddings before calling.")