✨ feat: 允许用户自行定义 Embedding 模型 #5177
Conversation
|
@cookieY is attempting to deploy a commit to the LobeHub Team on Vercel. A member of the Team first needs to authorize it. |
dosubot
bot
added
size:L
|
👍 @cookieY Thank you for raising your pull request and contributing to our Community |
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #5177 +/- ##
==========================================
- Coverage 92.20% 92.02% -0.18%
==========================================
Files 568 571 +3
Lines 43395 43596 +201
Branches 2706 2707 +1
==========================================
+ Hits 40012 40120 +108
- Misses 3383 3476 +93
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. |
|
支不支持本地部署的embedding模型？例如 api 地址是 http://127.0.0.1/v1 的类 openai 接口？ |
Does it support locally deployed embedding models? For example, the api address is an openai-like interface at http://127.0.0.1/v1? |
基于现有的框架设计无法实现这样的自定义需求，本地部署可以使用 ollama |
Such custom requirements cannot be achieved based on the existing framework design. For local deployment, ollama can be used. |
| @@ -4,10 +4,14 @@ import { z } from 'zod'; | |||
| export const getKnowledgeConfig = () => { | |||
| return createEnv({ | |||
| runtimeEnv: { | |||
| DEFAULT_FILES_CONFIG: !!process.env.DEFAULT_FILES_CONFIG | |||
| ? process.env.DEFAULT_FILES_CONFIG | |||
| : 'embedding_model=openai/embedding-text-3-small,reranker_model=cohere/rerank-english-v3.0,query_mode=full_text', | |||
There was a problem hiding this comment.
默认的 embedding model 、 query model 以及reranker model 不应该用 DEFAULT_SYSTEM_AGENT_ITEM 吧？
There was a problem hiding this comment.
这里不能用任何服务端方法。因为 schema 既会在 server 端跑，也会在 client端执行
如果要改默认向量化模型，这里就不需要加 $defaultFn(() => DEFAULT_EMBEDDING_MODEL), 了
| @@ -4,6 +4,7 @@ export interface EmbeddingsPayload { | |||
| * supported in `text-embedding-3` and later models. | |||
| */ | |||
| dimensions?: number; | |||
| index?: number; | |||
| @@ -275,22 +275,46 @@ export const LobeOpenAICompatibleFactory = <T extends Record<string, any> = any> | |||
| .filter(Boolean) as ChatModelCard[]; | |||
| } | |||
|
|
|||
| /** | |||
| * Due to the limitation of a maximum of 8k tokens per request in | |||
There was a problem hiding this comment.
8k 不是 request 的要求，而是 openai embedding 模型的上下文大小，因此这里不应该改变。要改也是上层调用 embedding 的地方做 8k 的分块
There was a problem hiding this comment.
这里的实现逻辑并不合理。system agent 和 embedding 是两个不同的业务功能模块，不应该混在一起。单独写一个 parseFileConfig
There was a problem hiding this comment.
这个模块感觉没有太大必要，而且未来提供 ui 配置还没法合并，不应该这么写。
💻 变更类型 | Change Type
🔀 变更说明 | Description of Change
由于之前代码commit差距过大 重新提交，提交说明参考 #4370
📝 补充信息 | Additional Information