feat(graphiti): Implement Graphiti client and related steps for episode ingestion and querying

- Added `graphiti_client.py` to manage the Graphiti client as a singleton.
- Created `ingest_episode_event_step.py` to handle episode ingestion from HTTP webhook events.
- Implemented `ingest_episode_step.py` for validating and enqueuing episode payloads via a POST request.
- Developed `query_graph_step.py` for performing semantic searches in the Graphiti Knowledge-Graph.
- Introduced an `__init__.py` file for the graphiti steps module.

services/graphiti_client.py

@@ -0,0 +1,100 @@
+"""Graphiti Knowledge-Graph – zentraler Client-Singleton.
+
+Wird von ingest_episode_event_step und query_graph_step genutzt.
+Lazy-initialisiert beim ersten Aufruf.
+"""
+import asyncio
+import os
+from typing import Any, Optional
+
+from graphiti_core import Graphiti
+from graphiti_core.llm_client import LLMConfig
+from graphiti_core.llm_client.openai_generic_client import OpenAIGenericClient
+from graphiti_core.embedder import OpenAIEmbedder, OpenAIEmbedderConfig
+
+
+class GraphitiError(Exception):
+    """Fehler beim Zugriff auf den Graphiti-Client."""
+
+
+_graphiti_client: Graphiti | None = None
+_graphiti_init_lock = asyncio.Lock()
+
+
+async def get_graphiti(ctx: Optional[Any] = None) -> Graphiti:
+    """Gibt den gecachten Graphiti-Client zurück (Singleton).
+
+    Args:
+        ctx: Optionaler Motia-Context für Logging während der Initialisierung.
+
+    Benötigte Umgebungsvariablen:
+        NEO4J_URI            – bolt://host:7687
+        NEO4J_USER           – neo4j
+        NEO4J_PASSWORD       – Pflicht
+        XAI_API_KEY          – xAI-Key für LLM (grok)
+        XAI_BASE_URL         – optional, Standard: https://api.x.ai/v1
+        XAI_MODEL            – optional, Standard: grok-4-1-fast-reasoning
+        OPENAI_API_KEY       – OpenAI-Key für Embeddings
+        GRAPHITI_EMBED_BASE_URL  – optional, Standard: https://api.openai.com/v1
+        GRAPHITI_EMBED_MODEL     – optional, Standard: text-embedding-3-small
+    """
+    global _graphiti_client
+    if _graphiti_client is None:
+        async with _graphiti_init_lock:
+            if _graphiti_client is None:
+                _log(ctx, "Initialisiere Graphiti-Client...")
+                try:
+                    _graphiti_client = await _build_graphiti()
+                    _log(ctx, "Graphiti-Client bereit.")
+                except KeyError as e:
+                    raise GraphitiError(f"Konfigurationsfehler – Umgebungsvariable fehlt: {e}") from e
+    return _graphiti_client
+
+
+def _log(ctx: Optional[Any], message: str, level: str = "info") -> None:
+    """Loggt via ctx.logger falls vorhanden, sonst print."""
+    if ctx is not None and hasattr(ctx, "logger"):
+        getattr(ctx.logger, level)(f"[GraphitiClient] {message}")
+    else:
+        print(f"[GraphitiClient] {message}")
+
+
+async def _build_graphiti() -> Graphiti:
+    neo4j_uri = os.environ["NEO4J_URI"]
+    neo4j_user = os.environ.get("NEO4J_USER", "neo4j")
+    neo4j_password = os.environ["NEO4J_PASSWORD"]
+
+    xai_base_url = os.environ.get("XAI_BASE_URL", "https://api.x.ai/v1")
+    xai_api_key = os.environ["XAI_API_KEY"]
+    xai_model = os.environ.get("XAI_MODEL", "grok-4-1-fast-reasoning")
+
+    embed_api_key = os.environ.get("OPENAI_API_KEY") or os.environ.get("GRAPHITI_EMBED_API_KEY", xai_api_key)
+    embed_base_url = os.environ.get("GRAPHITI_EMBED_BASE_URL", "https://api.openai.com/v1")
+    embed_model = os.environ.get("GRAPHITI_EMBED_MODEL", "text-embedding-3-small")
+
+    llm_client = OpenAIGenericClient(
+        config=LLMConfig(
+            api_key=xai_api_key,
+            base_url=xai_base_url,
+            model=xai_model,
+        )
+    )
+
+    embedder = OpenAIEmbedder(
+        config=OpenAIEmbedderConfig(
+            api_key=embed_api_key,
+            base_url=embed_base_url,
+            embedding_model=embed_model,
+            embedding_dim=1536,
+        )
+    )
+
+    client = Graphiti(
+        uri=neo4j_uri,
+        user=neo4j_user,
+        password=neo4j_password,
+        llm_client=llm_client,
+        embedder=embedder,
+    )
+    await client.build_indices_and_constraints()
+    return client