Create retrieval.py

retrieval.py

@@ -0,0 +1,35 @@
+from sentence_transformers import SentenceTransformer, util
+import chromadb
+from chromadb.utils import embedding_functions
+import numpy as np
+
+class SimpleRetriever:
+    def __init__(self):
+        self.encoder = SentenceTransformer('all-MiniLM-L6-v2')
+        self.client = chromadb.Client()
+        self.collection = self.client.create_collection(
+            name="incidents",
+            embedding_function=embedding_functions.SentenceTransformerEmbeddingFunction()
+        )
+        self._seed_incidents()
+
+    def _seed_incidents(self):
+        incidents = [
+            ("High latency in payment service, caused by database connection pool exhaustion.", "database_pool"),
+            ("Memory leak in API gateway after 24 hours of uptime.", "memory_leak"),
+            ("Authentication service returning 500 errors due to misconfigured OAuth.", "oauth_config"),
+            ("Disk full on logging node, causing log loss.", "disk_full"),
+        ]
+        for text, cause in incidents:
+            self.collection.add(
+                documents=[text],
+                metadatas=[{"cause": cause}],
+                ids=[cause]
+            )
+
+    def get_similarity(self, query: str) -> float:
+        results = self.collection.query(query_texts=[query], n_results=1)
+        if results['distances'] and len(results['distances'][0]) > 0:
+            # Convert L2 distance to similarity (inverse, scaled)
+            return 1.0 / (1.0 + results['distances'][0][0])
+        return 0.0