Skip to content
For the complete documentation index, see llms.txt. This page is also available as Markdown.

Data, Privacy, and Maintenance

The knowledge base stores hosted copies of imported materials, parsed text, Chunks, and retrieval indexes. Whether data leaves the local machine depends on the services chosen for parsing, OCR, embedding, re-ranking, and chat.

Keeping the original files local does not mean the entire knowledge base workflow is offline. As long as one processing step uses a cloud service, files, chunks, or queries needed to complete the task may be sent out.

What is saved after import

Content
Purpose
Update method

Copies of files in a file or directory

For the knowledge base to continue processing and displaying

Re-add or replace after the original file is updated

Webpage and note snapshots

Preserve the content as it was when imported

Re-import after the source is updated

Parsed text and Chunks

Preview and retrieval

Re-index after changing the processor or chunking

BM25 keyword index

Exact word search

Rebuilt when re-indexing

Vector index

Semantic retrieval

Generated or rebuilt after configuring or changing the embedding model

包含多条已处理资料的员工差旅制度知识库
Items in the material list are the objects hosted and indexed by the knowledge base, not a live synchronized view of the original directory.

What boundaries may a single query cross

展示解析、关键词检索、向量检索、重排和回答之间数据流的知识库架构图
The parsing, vector, re-ranking, and answer nodes in the diagram may all choose local or cloud services; check each one individually to determine the data boundary.
Selected capability
Possible received content

Cloud document processor

File content used for parsing

Cloud embedding model

Material chunks and retrieval queries

Cloud re-ranking model

Queries and candidate chunks

Cloud chat model

Questions, conversation context, and retrieved chunks

Local equivalent capability

Perform the corresponding processing on the local machine

Perform a full maintenance check

1

1. Record the current configuration

Record the knowledge base name, file processor, OCR, embedding model, re-ranking model, and key chunking settings. Use them to verify the environment after migration.

2

2. Clean up duplicates and old versions

Keep only the current version for the same policy; if historical audits are needed, clearly mark the year or version in the title to avoid search collisions.

3

3. Check abnormal entries

Handle materials marked [Error] or stuck [Processing] for a long time, and spot-check the text and Chunks. If an app interruption left indexing incomplete, run [Re-index].

4

4. Create the right backup

Open [Settings] → [Data]. Use a full backup that includes the knowledge base data files when migrating devices or preparing to delete materials.

5

5. Restore and verify in the target environment

Do not just confirm that restoration completed. Check the knowledge base entries, text, and Chunks, and run fixed retrieval questions.

6

6. Keep a rollback baseline

Validate new processors or models on a small set of materials first, then re-index in batches. Do not delete the most recent full backup before confirming the new results are stable.

Full backup vs. slim backup

Backup method
Included content
Applicable scenario
Limitations

Full backup

Data files such as chats, settings, and knowledge bases

Device migration, pre-deletion protection, full restore

Large file size, takes longer

Slim backup

Mainly chat logs and settings, skipping data files such as images and knowledge bases

Quickly preserve common settings and chats

Cannot restore a complete knowledge base on its own

How updates and deletions are handled

Source content updated

  1. Re-add the same-named material.

  2. Choose [Replace] when you need to overwrite the old version; choose [Keep all] only when you truly need both to coexist.

  3. Wait for the material to become ready.

  4. Spot-check the text and Chunks.

  5. Run fixed retrieval regression questions.

Processing settings updated

When only the processor, OCR, chunking, or model settings are changed, run [Re-index] on existing entries. Changing the configuration itself does not automatically reprocess old materials.

Delete materials or knowledge base

Deletion removes the content and indexes hosted by the knowledge base, but it does not delete the files in the original path or the original notes. Before proceeding, make sure the original source can still be found, a full backup is available, and check whether any agents are still bound to this knowledge base.

Configuration notes

Item
Recommended starting point
Acceptance method
Risk

Material version

Keep only the current version for the same use

Fixed questions still hit the correct version

Mixing old and new rules

Cloud service

Confirm item by item according to sensitivity level

Check the processor and model settings

Documents or chunks are sent to external services

Backup

Create a full backup before major changes

Check entries and retrieval after restoration

Slim backup lacks knowledge base files

Re-index

Process representative materials in batches

Compare the same question set before and after

One-time rebuild loses a usable baseline

User case

Xiaolin wants to migrate the team policy repository to a new computer. First he records the processor and model settings, creates a full backup, then checks the text and Chunks of three materials after restoring on the new computer and repeats the original five retrieval questions. Only after everything passes does he clean up the old environment.

The completion criteria are: the number of materials and titles match, key questions still hit the same source, and the team confirms that all cloud services meet data requirements.

Fully offline checklist

  • Document parsing and OCR use system, local, or self-hosted capabilities.

  • The embedding model runs locally.

  • Do not use cloud re-ranking, or use local re-ranking capabilities.

  • Conversations and agents use a local chat model.

  • No MCP, web search, or channels that send content to external systems are enabled.

FAQ

Will the knowledge base update automatically after modifying the original file?

No. Files, web pages, and notes are turned into materials based on the content at import time. Re-add and replace, or re-index as needed.

Does using a local embedding model mean it is completely offline?

Not necessarily. If any step in parsing, OCR, re-ranking, or chat uses a cloud service, the content needed to complete the task may be sent out.

Can a slim backup restore a knowledge base?

It cannot restore the complete knowledge base files. Use a full backup before migration or deletion, and actually verify the materials and retrieval after restoration.

Continue reading

Last updated

Was this helpful?