update langchain rag (#2357)

1. to align with this PR:

langchain-ai/langchain@1c31234#diff-072c45917082c5132a4631ee3fa519e3732985c436c1ca60fce0c4e03d21d2ba

2. support text completion for llama_index

notebooks/llm-agent-react/llm-agent-rag-llamaindex.ipynb

@@ -309,8 +309,8 @@
     "    return f\"<|system|><|end|><|user|>{completion}<|end|><|assistant|>\\n\"\n",
     "\n",
     "\n",
-    "def llama_completion_to_prompt(completion):\n",
-    "    return f\"<|begin_of_text|><|start_header_id|>system<|end_header_id|><|eot_id|><|start_header_id|>user<|end_header_id|>{completion}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\\n\"\n",
+    "def llama3_completion_to_prompt(completion):\n",
+    "    return f\"<|begin_of_text|><|start_header_id|>system<|end_header_id|>\\n\\n<|eot_id|><|start_header_id|>user<|end_header_id|>\\n\\n{completion}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\\n\\n\"\n",
     "\n",
     "\n",
     "llm = OpenVINOLLM(\n",
@@ -319,7 +319,7 @@
     "    max_new_tokens=1000,\n",
     "    model_kwargs={\"ov_config\": ov_config},\n",
     "    generate_kwargs={\"do_sample\": False, \"temperature\": None, \"top_p\": None},\n",
-    "    completion_to_prompt=phi_completion_to_prompt if llm_model_path == \"Phi-3-mini-4k-instruct-int4-ov\" else llama_completion_to_prompt,\n",
+    "    completion_to_prompt=phi_completion_to_prompt if llm_model_path == \"Phi-3-mini-4k-instruct-int4-ov\" else llama3_completion_to_prompt,\n",
     "    device_map=llm_device.value,\n",
     ")"
    ]