Add a Complete Ingestion Pipeline Example (encode → insert → search) with Correct Result Iteration

bootcamp/ingestion_pipeline/main.py

@@ -0,0 +1,48 @@
+from encoder import TextEncoder
+from milvus_client import MilvusClient
+import numpy as np
+
+def main():
+    encoder = TextEncoder()
+
+    client = MilvusClient(
+        host="127.0.0.1",
+        port="19530",
+        collection_name="text_rag",
+        dim=384
+    )
+
+    texts = [
+        "Milvus makes vector search scalable.",
+        "LangChain helps build LLM applications.",
+        "Vector databases are essential for AI systems."
+    ]
+
+    print("\nEncoding texts and inserting into Milvus...\n")
+    vecs = encoder.encode(texts)
+
+    # ⭐ 强制所有向量转为 float32，Milvus 只接受 float32
+    vecs = np.array(vecs, dtype=np.float32).tolist()
+
+    client.insert(vecs)
+    print("\nInserted documents successfully!\n")
+
+    query = "What can scale vector search?"
+    print(f"\nSearching for: {query}\n")
+
+    query_vec = encoder.encode(query)[0]
+    query_vec = np.array(query_vec, dtype=np.float32).tolist()   # ⭐ 关键：强制为 float32
+
+    # ⭐ 你的 milvus_client 会自动封装成 [query_vec]，这里千万不能再包一层
+    res = client.search(query_vec, topk=3)
+
+    print("\nSearch results:")
+    for hits in res:          # res = [[hit1, hit2, hit3]]
+        for hit in hits:
+            print(f"ID={hit.id}, distance={hit.distance}")
+
+    print("\nDone!")
+
+
+if __name__ == "__main__":
+    main()