File Upload / KnowledgeBase Deployment Feedback | 知识库部署问题反馈

[arvinxx]

This issue is only for feedback on deployment problems. For feature requests or bug-related issues, please open a new issue.

此 issue 只用于反馈部署问题，功能需求 / bug 类问题请新开 issue。

---

The first phase of the knowledge base has been released, and here is a brief deployment guide:

1. Create a pg instance, which needs to include the pgvector plugin (Note: The following command is for demonstration purposes only, as this pg instance does not include the persistence part. Please build a production-grade pg instance that meets your requirements.)

If your pg instance is deployed via Docker, you can use the pgvector/pgvector:pg16 image instead. (Platforms like Vercel / Neon / Supabase do not require any additional operations, as they come with the pgvector plugin by default.)

docker run --name my-postgres --network pg -e POSTGRES_PASSWORD=mysecretpassword -p 5432:5432 -d pgvector/pgvector:pg16

2. Create a lobe-chat.env file to store environment variables:

# Website Domain Name
APP_URL=https://your-prod-domain.com

# DB Required
KEY_VAULTS_SECRET=jgwsK28dspyVQoIf8/M3IIHl1h6LYYceSYNXeLpy6uk=
DATABASE_URL=postgres://postgres:mysecretpassword@my-postgres:5432/postgres


# NEXT_AUTH related, you can use auth0, and if there are other integration requests, feel free to submit a PR
NEXT_AUTH_SECRET=3904039cd41ea1bdf6c93db0db96e250
NEXT_AUTH_SSO_PROVIDERS=auth0
NEXTAUTH_URL=https://your-prod-domain.com/api/auth
AUTH0_CLIENT_ID=xxxxxx
AUTH0_CLIENT_SECRET=cSX_xxxxx
AUTH0_ISSUER=https://lobe-chat-demo.us.auth0.com

# S3 Related
S3_ACCESS_KEY_ID=xxxxxxxxxx
S3_SECRET_ACCESS_KEY=xxxxxxxxxx
S3_ENDPOINT=https://xxxxxxxxxx.r2.cloudflarestorage.com
S3_BUCKET=lobechat
S3_PUBLIC_DOMAIN=https://s3-for-lobechat.your-domain.com

3. Start the lobe-chat-database docker image

docker run -it -d -p 3210:3210 --network pg --env-file lobe-chat.env --name lobe-chat-database lobehub/lobe-chat-database

I'm sorry, but I need the source text to provide a translation. Please provide the text you would like translated

Other considerations:

• Vectorization requires the use of OpenAI's text-embedding-3-small model, please confirm in advance whether your API supports this model;
• If you encounter a chunk failure due to an unsupported chunk type, you can provide feedback at Better File Chunk | 更加强大的文件分块 #3550
• Minio needs to add S3_ENABLE_PATH_STYLE=1

---

知识库一期已发布，以下是一个简版部署指南：

1. 创建 pg 实例，需要包含 pgvector 插件（注意：以下命令仅用于演示，因为这个 pg 实例并没有包含持久化部分，请自行构建一个符合你诉求的 生产级 pg 实例）

如果你的 pg 实例是通过 docker 部署，可以使用 pgvector/pgvector:pg16 镜像替代。（Vercel / Neon / Supabase 这些不需要额外操作，默认就带有 pgvector 插件）

docker run --name my-postgres --network pg -e POSTGRES_PASSWORD=mysecretpassword -p 5432:5432 -d pgvector/pgvector:pg16

2. 创建一个 lobe-chat.env 文件用于存放环境变量：

#网站域名
APP_URL=https://your-prod-domain.com

# DB 必须
KEY_VAULTS_SECRET=jgwsK28dspyVQoIf8/M3IIHl1h6LYYceSYNXeLpy6uk=
DATABASE_URL=postgres://postgres:mysecretpassword@my-postgres:5432/postgres


# NEXT_AUTH 相关，可以使用 auth0 等，如有其他接入诉求欢迎提 PR
NEXT_AUTH_SECRET=3904039cd41ea1bdf6c93db0db96e250
NEXT_AUTH_SSO_PROVIDERS=auth0
NEXTAUTH_URL=https://your-prod-domain.com/api/auth
AUTH0_CLIENT_ID=xxxxxx
AUTH0_CLIENT_SECRET=cSX_xxxxx
AUTH0_ISSUER=https://lobe-chat-demo.us.auth0.com

# S3 相关
S3_ACCESS_KEY_ID=xxxxxxxxxx
S3_SECRET_ACCESS_KEY=xxxxxxxxxx
S3_ENDPOINT=https://xxxxxxxxxx.r2.cloudflarestorage.com
S3_BUCKET=lobechat
S3_PUBLIC_DOMAIN=https://s3-for-lobechat.your-domain.com

3. 启动 lobe-chat-database docker 镜像

docker run -it -d -p 3210:3210 --network pg --env-file lobe-chat.env --name lobe-chat-database lobehub/lobe-chat-database

---

其余注意事项：

• 向量化需要使用 OpenAI 的 text-embedding-3-small 模型，请提前确认你的 API 是否支持该模型；
• 如果遇到分块失败，是不支持的 chunk 类型，可以在 Better File Chunk | 更加强大的文件分块 #3550 这里反馈
• minio 需要添加 S3_ENABLE_PATH_STYLE=1

[arvinxx]

我想问一下，连接数据库后报错。是不是因为迁移数据库报的错，我的表没有自动生成，是我用“src/database/server/migrations/0000_init.sql”文件手动生成的。

对。选择手动生成的话，以后每次数据库变更你都要自己执行一遍迁移的 sql

我又重新执行了一遍，然后我使用oauth2.0登录为什么没有把我的用户信息存进去表里，需要我自己写方法吗？还是profile这里定义好对应的表字段就行？

@cy948 来看下？

[lobehubbot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

I would like to ask about an error after connecting to the database. Is it because of the error reported in the migration database? My table was not automatically generated. I generated it manually using the "src/database/server/migrations/0000_init.sql" file.

Right. If you choose to generate it manually, you will have to execute the migration SQL yourself every time the database changes in the future.

I executed it again, and then I used oauth2.0 to log in. Why was my user information not saved in the table? Do I need to write a method myself? Or should the corresponding table fields be defined here in the profile?

@cy948 Let’s take a look?

[mujiannan]

这个项目的bug修复速度奇快，Server版的DALLE.3图片存储功能也正常工作了

[asd4259682]

Cannot read properties of null (reading 'toString') 知识库里面传的是一个pdf，chunk已经完成了，提问的时候报这个错误。 最新的v.12.8版本，编译部署启动

刚才测试了下，发现知识库里面有向量化失败的pdf就有这个问题，把这个失败的pdf删除就没有问题了。

我也遇到类似问题，但是删除完所有向量化失败和分块失败的pdf后，仍然会报错……

[lobehubbot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

Cannot read properties of null (reading 'toString') The knowledge base is uploading a PDF, and the chunk has been completed. This error is reported when asking questions. The latest v.12.8 version, compilation and deployment started! [image](https://private-user-images.githubusercontent.com/81293467/360732696-6f460fca-d1ca-4faf-a87a-0c6c751ee80c.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpX VCJ9.. ltSLji9XVRpqnQ7BL74W0BcJkH5Tig1-RqRQh6hUGn0)

I just tested it and found that there is a pdf with failed vectorization in the knowledge base that has this problem. Delete the failed pdf and there will be no problem.

I also encountered a similar problem, but after deleting all PDFs that failed to vectorize and fail to block, an error still occurred...

[arvinxx]

@asd4259682 专门开个 issue，这个不是部署问题。是使用问题

[lobehubbot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

@asd4259682 opened an issue specifically, this is not a deployment issue. It's a matter of usage

[wcu1117]

我想问一下，连接数据库后报错。是不是因为迁移数据库报的错，我的表没有自动生成，是我用“src/database/server/migrations/0000_init.sql”文件手动生成的。

对。选择手动生成的话，以后每次数据库变更你都要自己执行一遍迁移的 sql

我又重新执行了一遍，然后我使用oauth2.0登录为什么没有把我的用户信息存进去表里，需要我自己写方法吗？还是profile这里定义好对应的表字段就行？

@cy948 来看下？

这个问题能帮忙解答一下吗？我现在授权回调回来直接报，但是网页上面已经是登录成功了：
{"level":30,"time":1724662603462,"pid":15288,"hostname":"鍗庡","msg":"Error in tRPC handler (lambda) on path: user.getUserState, type: query"} UserNotFoundError [TRPCError]: user not found at UserModel.getUserState (webpack-internal:///(rsc)/./src/database/server/models/user.ts:117:23)

[zdt3476]

你好，我这边使用阿里 OSS 作为存储，上传能正常上传，但是分块的时候报错了，但是这边没有更详细的错误

[lobehubbot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

Hello, I use Alibaba OSS as storage. The upload can be uploaded normally, but an error is reported when dividing into chunks, but there are no more detailed errors here.

[wcu1117]

我想问一下，连接数据库后报错。是不是因为迁移数据库报的错，我的表没有自动生成，是我用“src/database/server/migrations/0000_init.sql”文件手动生成的。

对。选择手动生成的话，以后每次数据库变更你都要自己执行一遍迁移的 sql

我又重新执行了一遍，然后我使用oauth2.0登录为什么没有把我的用户信息存进去表里，需要我自己写方法吗？还是profile这里定义好对应的表字段就行？

@cy948 来看下？

这个问题能帮忙解答一下吗？我现在授权回调回来直接报，但是网页上面已经是登录成功了： {"level":30,"time":1724662603462,"pid":15288,"hostname":"鍗庡","msg":"Error in tRPC handler (lambda) on path: user.getUserState, type: query"} UserNotFoundError [TRPCError]: user not found at UserModel.getUserState (webpack-internal:///(rsc)/./src/database/server/models/user.ts:117:23)

我现在解决这个问题的方法是把这个文件的代码改了：src/server/routers/lambda/user.ts

if (enableClerk && error instanceof UserNotFoundError) {

把这个判断clerk的部分删了，加入了nextauth的处理逻辑，我不懂是否会影响后续的流程，现在使用一切正常。

[lobehubbot]

✅ @arvinxx

This issue is closed, If you have any questions, you can comment and reply.
此问题已经关闭。如果您有任何问题，可以留言并回复。

[lobehubbot]

🎉 This issue has been resolved in version 1.14.3 🎉

The release is available on:

• npm package (@latest dist-tag)
• GitHub release

Your semantic-release bot 📦🚀

[zwjzxh520]

向量化需要使用 OpenAI 的 text-embedding-3-small 模型，请提前确认你的 API 是否支持该模型；

向量化有国产的api可以替代吗？

[lobehubbot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

Vectorization requires the use of OpenAI's text-embedding-3-small model. Please confirm in advance whether your API supports this model;

Is there any domestic API that can replace vectorization?

[wcu1117]

在哪里修改，可以计算统计某个用户的token总使用量

[lobehubbot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

Where to modify, you can calculate and count the total token usage of a certain user