Self Checks

• I have read the Contributing Guide and Language Policy.
• This is only for bug report, if you would like to ask a question, please head to Discussions.
• I have searched for existing issues search for existing issues, including closed ones.
• I confirm that I am using English to submit this report, otherwise it will be closed.
• 【中文用户 & Non English User】请使用英语提交，否则会被关闭 ：）
• Please do not modify this template :) and fill in all the required fields.

Dify version

1.14.0

Cloud or Self Hosted

Self Hosted (Docker)

Steps to reproduce

Describe the bug When two different documents contain exactly the same text in one of their segments (e.g., both have a child segment with the content "Hello"), the Weaviate vector database uses only the text content (page_content) to generate the UUID. Because the content is identical, the generated UUID is identical. As a result, when the second document is parsed and embedded, it completely overwrites the vector node of the first document. This causes the first document's segment to become unretrievable.

To Reproduce

1. Upload Document A1 and configure it with parent-child segmentation. Suppose one of the child segments has the exact content: "Hello".
2. Upload Document B2 and also configure it with parent-child segmentation. Ensure it also generates a child segment with the exact content: "Hello".
3. Wait for the indexing process to finish.
4. Perform a retrieval test using the keyword "Hello".
5. Only Document B2 is retrieved (with a high score). Document A1 is missing.
6. If you disable Document B2 in the knowledge base and test again, Document A1 is still not retrieved.

✔️ Expected Behavior

Identical text segments from different documents should be stored as separate nodes in the vector database and both should be retrievable based on their respective document contexts.

❌ Actual Behavior

Additional context / Root cause analysis* In api/providers/vdb/vdb-weaviate/src/dify_vdb_weaviate/weaviate_vector.py, the _get_uuids method generates deterministic UUIDs using _uuid.uuid5 purely based on doc.page_content.

    def _get_uuids(self, documents: list[Document]) -> list[str]:
        URL_NAMESPACE = _uuid.UUID("6ba7b811-9dad-11d1-80b4-00c04fd430c8")
        uuids = []
        for doc in documents:
            uuid_val = _uuid.uuid5(URL_NAMESPACE, doc.page_content)
            uuids.append(str(uuid_val))
        return uuids

When Document B2's "Hello" is embedded, it calculates the exact same UUID as A1's "Hello" and overwrites A1's node metadata (including document_id).

Suggested fixes Modify the UUID generation logic to include the unique doc_id in the hash calculation. This ensures that even if the text content is identical, the UUID will be unique per segment across different documents.

    def _get_uuids(self, documents: list[Document]) -> list[str]:
        URL_NAMESPACE = _uuid.UUID("6ba7b811-9dad-11d1-80b4-00c04fd430c8")
        uuids = []
        for doc in documents:
            doc_id = (doc.metadata or {}).get("doc_id", "")
            # Combine doc_id and page_content to avoid overwriting identical text blocks
            text_to_hash = f"{doc_id}_{doc.page_content}" if doc_id else doc.page_content
            uuid_val = _uuid.uuid5(URL_NAMESPACE, text_to_hash)
            uuids.append(str(uuid_val))
        return uuids