[Question] How to build vector index incrementally for new/updated graph vertices?

[coolboywcm]

Problem Type (问题类型)

None

Before submit

• I had searched in the issues and found no similar issues.

Environment (环境信息)

Server Version: 1.5
Toolchain Version: hugegraph-rag lastest
Data Size: 2000 vertices, 3000 edges

Your Question (问题描述)

I’m using HugeGraph with a custom RAG pipeline. My graph data is frequently updated by third-party services.

Currently, I rebuild the entire vector index from scratch after every change — which is inefficient.

👉 Does HugeGraph AI natively support incremental index updates (add/update/delete embeddings for changed vertices only)?

If not, is it on the roadmap? Any recommended approach?

Thanks!

Vertex/Edge example (问题点 / 边数据举例)

Schema [VertexLabel, EdgeLabel, IndexLabel] (元数据结构)

[dosubot]

HugeGraph-AI does natively support incremental updates for vector indexes—you can add, update, or delete embeddings for changed vertices without rebuilding the entire index from scratch. The vector index module provides methods for building, updating, and cleaning indexes, and supports batch and parallel processing for efficient handling of frequent updates. This design is well-suited for RAG pipelines where graph data changes often. You can also integrate with external vector databases like Faiss or Milvus if needed. For best results, use the batch and parallel options to process updates in groups and maximize throughput. More details are available in the official documentation.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}    

[imbajin]

@coolboywcm u could use https://deepwiki.com/apache/incubator-hugegraph-ai to ask 🤖 for (code related)details

If u already have vertexes [a, b], then u imported [b, c], the vid embedding method will only embed c (We use set to take the difference set and do not perform a full reconstruction of the entire vid. For details, please refer to the source code)