I was trying to build a RAG with the combination of both Llamaindex and Milvus, and I was referring to the code from this Llamaindex document: https://docs.llamaindex.ai/en/stable/examples/node_postprocessor/MetadataReplacementDemo/

Based on the code above, I modified it to use mainly Milvus with my own data inserted. However, when I attempted to make the index, some of the nodes were giving errors, saying how the length of dynamic field exceeds max length.

I thought this was due to the fact that the nodes were too big, thus I tried to decrease their sizes but still got the same error.
I was wondering if there is something I can do on the Milvus end, for example increasing the max limit.

The error suggests ‘Please reduce your prompt; or completion length.’, but I am not sure how to do that.

Here is my code for reference:

from llama_index.core import VectorStoreIndex
from llama_index.vector_store.milvus import MilvusVectorStore
from llama_index.core import StorageContext
from pymilvus import MilvusClient

client = MilvusClient(
    uri = "http://localhost:19530"
)

if client.has_collection(collection_name = "SWR_collection_for_all_nodes")
    client.drop_collection(collection_name = "SWR_collection_for_all_nodes")

count = 0
fails = 0
successes = 0

for x in nodes:
    count += 1
    try:
        del vector_store, storage_context, sentence_index
    except:
        pass
    vector_store = MiluvsVectorStore(uri = "http://localhost:19530", collection_name = "SWR_collection_for_all_nodes", dim = 1536, overwrite = False)
    storage_context = StorageContext.from_defaults(vector_store = vector_store)

    try:
        sentence_index = VectorStoreIndex([x], storage_context = storage_context)
    exception Exception as e:
        print(f"failed to add node {count} to storage")
        print(e)
        fails += 1
        continue
    print(f"node {count} added to storage")
    successes += 1

print(f"Added {success} nodes to storage, failed to add {fails} nodes")