[BUG] 简洁阐述问题 / Concise description of the issue

[shengmengru]

这些报错怎么解决啊
附加信息 / Additional Information
2026-01-08 08:48:12.466 | ERROR | chatchat.server.utils:check_embed_model:355 - failed to access embed model 'quentinz/bge-large-zh-v1.5': Error raised by inference API HTTP code: 404, {"error":"model "quentinz/bge-large-zh-v1.5" not found, try pulling it first"}
ERROR: Exception in ASGI application
Traceback (most recent call last):
File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/sse_starlette/sse.py", line 269, in call
await wrap(partial(self.listen_for_disconnect, receive))
File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/sse_starlette/sse.py", line 258, in wrap
await func()
File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/sse_starlette/sse.py", line 215, in listen_for_disconnect
message = await receive()
File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/uvicorn/protocols/http/httptools_impl.py", line 563, in receive
await self.message_event.wait()
File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/asyncio/locks.py", line 309, in wait
await fut
asyncio.exceptions.CancelledError

During handling of the above exception, another exception occurred:

• Exception Group Traceback (most recent call last):
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/uvicorn/protocols/http/httptools_impl.py", line 409, in run_asgi
  | result = await app( # type: ignore[func-returns-value]
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/uvicorn/middleware/proxy_headers.py", line 60, in call
  | return await self.app(scope, receive, send)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/fastapi/applications.py", line 1054, in call
  | await super().call(scope, receive, send)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/applications.py", line 123, in call
  | await self.middleware_stack(scope, receive, send)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/middleware/errors.py", line 186, in call
  | raise exc
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/middleware/errors.py", line 164, in call
  | await self.app(scope, receive, _send)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/middleware/exceptions.py", line 62, in call
  | await wrap_app_handling_exceptions(self.app, conn)(scope, receive, send)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/_exception_handler.py", line 64, in wrapped_app
  | raise exc
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/_exception_handler.py", line 53, in wrapped_app
  | await app(scope, receive, sender)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/routing.py", line 758, in call
  | await self.middleware_stack(scope, receive, send)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/routing.py", line 778, in app
  | await route.handle(scope, receive, send)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/routing.py", line 299, in handle
  | await self.app(scope, receive, send)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/routing.py", line 79, in app
  | await wrap_app_handling_exceptions(app, request)(scope, receive, send)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/_exception_handler.py", line 64, in wrapped_app
  | raise exc
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/_exception_handler.py", line 53, in wrapped_app
  | await app(scope, receive, sender)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/routing.py", line 77, in app
  | await response(scope, receive, send)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/sse_starlette/sse.py", line 269, in call
  | await wrap(partial(self.listen_for_disconnect, receive))
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/anyio/_backends/_asyncio.py", line 685, in aexit
  | raise BaseExceptionGroup(
  | exceptiongroup.ExceptionGroup: unhandled errors in a TaskGroup (1 sub-exception)
  +-+---------------- 1 ----------------
  | Traceback (most recent call last):
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/sse_starlette/sse.py", line 258, in wrap
  | await func()
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/sse_starlette/sse.py", line 246, in stream_response
  | chunk = ensure_bytes(data, self.sep)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/sse_starlette/sse.py", line 137, in ensure_bytes
  | return ServerSentEvent(**data).encode()
  | TypeError: init() got an unexpected keyword argument 'code'
  +------------------------------------
  2026-01-08 08:48:12.504 | ERROR | chatchat.webui_pages.utils:ret_sync:208 - RemoteProtocolError: API通信遇到错误：peer closed connection withoung complete message body (incomplete chunked read)

[dosubot]

Hi @shengmengru! I'm Dosu and I’m helping the Langchain-Chatchat team.

你的 embedding 模型报错是因为 'quentinz/bge-large-zh-v1.5' 没有在本地推理服务（如 xinference 或 ollama）注册和启动。你需要先注册并启动该模型，确保模型名和配置一致。具体步骤如下：

1. 注册 embedding 模型：编辑并运行模型注册脚本（如 model_registrations_emb.sh），内容类似这样（路径和模型名根据实际情况修改）：

   curl 'http://127.0.0.1:9997/v1/model_registrations/embedding' \
     -H 'Content-Type: application/json' \
     --data-raw '{"model":"{\"model_name\":\"bge-large-zh-v1.5\",\"dimensions\":768,\"max_tokens\":512,\"model_uri\":\"/root/model/bge-large-zh\",\"language\":[\"en\",\"zh\"]}","persist":true}'

   参考

2. 启动 embedding 模型：运行启动脚本（如 start_models_emb.sh），内容类似：

   curl 'http://127.0.0.1:9997/v1/models' \
     -H 'Content-Type: application/json' \
     --data-raw '{"model_uid":"bge-large-zh-v1.5","model_name":"bge-large-zh-v1.5","model_type":"embedding","replica":1,"n_gpu":"auto"}'

   参考

3. 模型名要和配置一致：比如在 settings.py 或 model_providers.yaml 里配置的模型名要和注册/启动的名字一致，否则会找不到模型。参考

4. 切换 embedding 模型后需重建知识库：知识库会记录创建时用的 embedding 模型，如果你换了 embedding，需要用 python chatchat/cli.py kb --recreate-vs 重建知识库，否则会报错。参考

5. SSE TypeError 和 RemoteProtocolError：这些报错很可能是模型加载失败导致 API 响应不完整或格式错误，解决 embedding 模型问题后一般会一起消失。参考

6. Faiss 兼容性问题：如果你用的是 Faiss 向量库，quentinz/bge-large-zh-v1.5 可能有兼容性问题，建议尝试其他 embedding 或向量库组合。参考

如果你用的是 ollama 平台，确保模型已通过 ollama 拉取并启动，命令类似 ollama pull quentinz/bge-large-zh-v1.5。

修复 embedding 模型后，其他报错一般会一起解决。如果还有 SSE 相关报错，可以再反馈具体信息。

To reply, just mention @dosu.

---

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

   

[shengmengru]

您好，我修改了使用的embedding，并且使用openai平台，还是一直在报错：
2026-01-09 11:12:00.435 | INFO | chatchat.startup:start_main_server:257 - 正在启动服务：
2026-01-09 11:12:00.439 | INFO | chatchat.startup:start_main_server:258 - 如需查看 llm_api 日志，请前往 /newVol/users/shengmr/llm/Langchain-Chter/data/logs
/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/langchain/_api/module_import.py:87: LangChainDeprecationWarning: Importing GuardrailsOutputParser from langchain.output_parsers is deprecated. Please replace the import with the following:
from langchain_community.output_parsers.rail_parser import GuardrailsOutputParser
warnings.warn(
/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/pydantic/_internal/fields.py:160: UserWarning: Field "model_name" has conflict with protected namespace "model".

You may be able to resolve this warning by setting model_config['protected_namespaces'] = ().
warnings.warn(
2026-01-09 11:12:06.558 | INFO | chatchat.startup:run_api_server:55 - Api MODEL_PLATFORMS: [PlatformConfig(platform_name='oneapi', platform_type='oneapi', api_base_url='http://127.0.0.1:8000/v1', api_key='sk-4aa129d7b0b84f8e9ec0fbcfe3e041f6', api_proxy='', api_concurrencies=5, auto_detect_model=False, llm_models=['chatglm_pro', 'chatglm_turbo', 'chatglm_std', 'chatglm_lite', 'qwen-turbo', 'qwen-plus', 'qwen-max', 'qwen-max-longcontext', 'ERNIE-Bot', 'ERNIE-Bot-turbo', 'ERNIE-Bot-4', 'SparkDesk'], embed_models=['text-embedding-v1', 'Embedding-V1'], text2image_models=[], image2text_models=[], rerank_models=[], speech2text_models=[], text2speech_models=[])]
INFO: Started server process [127027]
INFO: Waiting for application startup.
INFO: Application startup complete.
INFO: Uvicorn running on http://10.168.6.34:7861 (Press CTRL+C to quit)
2026-01-09 11:12:11.030 | INFO | chatchat.startup:run_webui:80 - Webui MODEL_PLATFORMS: [PlatformConfig(platform_name='oneapi', platform_type='oneapi', api_base_url='http://127.0.0.1:8000/v1', api_key='sk-4aa129d7b0b84f8e9ec0fbcfe3e041f6', api_proxy='', api_concurrencies=5, auto_detect_model=False, llm_models=['chatglm_pro', 'chatglm_turbo', 'chatglm_std', 'chatglm_lite', 'qwen-turbo', 'qwen-plus', 'qwen-max', 'qwen-max-longcontext', 'ERNIE-Bot', 'ERNIE-Bot-turbo', 'ERNIE-Bot-4', 'SparkDesk'], embed_models=['text-embedding-v1', 'Embedding-V1'], text2image_models=[], image2text_models=[], rerank_models=[], speech2text_models=[], text2speech_models=[])]

You can now view your Streamlit app in your browser.

URL: http://10.168.6.34:8501

INFO: 10.168.6.34:41208 - "GET /tools HTTP/1.1" 200 OK
2026-01-09 11:12:18,069 httpx 127295 INFO HTTP Request: GET http://10.168.6.34:7861/tools "HTTP/1.1 200 OK"
/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/langchain/_api/module_import.py:87: LangChainDeprecationWarning: Importing GuardrailsOutputParser from langchain.output_parsers is deprecated. Please replace the import with the following:
from langchain_community.output_parsers.rail_parser import GuardrailsOutputParser
warnings.warn(
INFO: 10.168.6.34:41210 - "POST /knowledge_base/search_docs HTTP/1.1" 200 OK
2026-01-09 11:12:26,393 httpx 127295 INFO HTTP Request: POST http://10.168.6.34:7861/knowledge_base/search_docs "HTTP/1.1 200 OK"
INFO: 10.168.6.34:41212 - "POST /knowledge_base/recreate_vector_store HTTP/1.1" 200 OK
2026-01-09 11:12:27,157 httpx 127295 INFO HTTP Request: POST http://10.168.6.34:7861/knowledge_base/recreate_vector_store "HTTP/1.1 200 OK"
2026-01-09 11:12:27,990 openai._base_client 127027 INFO Retrying request to /embeddings in 0.404542 seconds
2026-01-09 11:12:28,398 openai._base_client 127027 INFO Retrying request to /embeddings in 0.883340 seconds
2026-01-09 11:12:29,284 openai._base_client 127027 INFO Retrying request to /embeddings in 1.871602 seconds
2026-01-09 11:12:31,159 chatchat.server.localai_embeddings 127027 WARNING Retrying chatchat.server.localai_embeddings.embed_with_retry.._embed_with_retry in 4.0 seconds as it raised APIConnectionError: Connection error..
2026-01-09 11:12:35,164 openai._base_client 127027 INFO Retrying request to /embeddings in 0.491765 seconds
2026-01-09 11:12:35,659 openai._base_client 127027 INFO Retrying request to /embeddings in 0.912875 seconds
2026-01-09 11:12:36,574 openai._base_client 127027 INFO Retrying request to /embeddings in 1.886101 seconds
2026-01-09 11:12:38,465 chatchat.server.localai_embeddings 127027 WARNING Retrying chatchat.server.localai_embeddings.embed_with_retry.._embed_with_retry in 4.0 seconds as it raised APIConnectionError: Connection error..
2026-01-09 11:12:42,473 openai._base_client 127027 INFO Retrying request to /embeddings in 0.439051 seconds
2026-01-09 11:12:42,916 openai._base_client 127027 INFO Retrying request to /embeddings in 0.754007 seconds
2026-01-09 11:12:43,675 openai._base_client 127027 INFO Retrying request to /embeddings in 1.919182 seconds
2026-01-09 11:12:45.599 | ERROR | chatchat.server.utils:check_embed_model:355 - failed to access embed model 'text-embedding-v1': Connection error.
ERROR: Exception in ASGI application
Traceback (most recent call last):
File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/sse_starlette/sse.py", line 289, in call
await wrap(partial(self.listen_for_disconnect, receive))
File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/sse_starlette/sse.py", line 278, in wrap
await func()
File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/sse_starlette/sse.py", line 228, in listen_for_disconnect
message = await receive()
File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/uvicorn/protocols/http/httptools_impl.py", line 563, in receive
await self.message_event.wait()
File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/asyncio/locks.py", line 309, in wait
await fut
asyncio.exceptions.CancelledError

During handling of the above exception, another exception occurred:

• Exception Group Traceback (most recent call last):
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/uvicorn/protocols/http/httptools_impl.py", line 409, in run_asgi
  | result = await app( # type: ignore[func-returns-value]
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/uvicorn/middleware/proxy_headers.py", line 60, in call
  | return await self.app(scope, receive, send)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/fastapi/applications.py", line 1054, in call
  | await super().call(scope, receive, send)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/applications.py", line 123, in call
  | await self.middleware_stack(scope, receive, send)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/middleware/errors.py", line 186, in call
  | raise exc
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/middleware/errors.py", line 164, in call
  | await self.app(scope, receive, _send)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/middleware/exceptions.py", line 62, in call
  | await wrap_app_handling_exceptions(self.app, conn)(scope, receive, send)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/_exception_handler.py", line 64, in wrapped_app
  | raise exc
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/_exception_handler.py", line 53, in wrapped_app
  | await app(scope, receive, sender)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/routing.py", line 758, in call
  | await self.middleware_stack(scope, receive, send)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/routing.py", line 778, in app
  | await route.handle(scope, receive, send)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/routing.py", line 299, in handle
  | await self.app(scope, receive, send)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/routing.py", line 79, in app
  | await wrap_app_handling_exceptions(app, request)(scope, receive, send)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/_exception_handler.py", line 64, in wrapped_app
  | raise exc
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/_exception_handler.py", line 53, in wrapped_app
  | await app(scope, receive, sender)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/starlette/routing.py", line 77, in app
  | await response(scope, receive, send)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/sse_starlette/sse.py", line 289, in call
  | await wrap(partial(self.listen_for_disconnect, receive))
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/anyio/_backends/_asyncio.py", line 685, in aexit
  | raise BaseExceptionGroup(
  | exceptiongroup.ExceptionGroup: unhandled errors in a TaskGroup (1 sub-exception)
  +-+---------------- 1 ----------------
  | Traceback (most recent call last):
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/sse_starlette/sse.py", line 278, in wrap
  | await func()
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/sse_starlette/sse.py", line 259, in stream_response
  | chunk = ensure_bytes(data, self.sep)
  | File "/gfs/users/shengmr/miniconda3/envs/llm/lib/python3.8/site-packages/sse_starlette/sse.py", line 142, in ensure_bytes
  | return ServerSentEvent(**data).encode()
  | TypeError: init() got an unexpected keyword argument 'code'
  +------------------------------------
  2026-01-09 11:12:45.629 | ERROR | chatchat.webui_pages.utils:ret_sync:208 - RemoteProtocolError: API通信遇到错误：peer closed connection withoung complete message body (incomplete chunked read)
  INFO: 10.168.6.34:41238 - "POST /knowledge_base/search_docs HTTP/1.1" 200 OK
  2026-01-09 11:12:45,962 httpx 127295 INFO HTTP Request: POST http://10.168.6.34:7861/knowledge_base/search_docs "HTTP/1.1 200 OK"
  INFO: 10.168.6.34:41240 - "GET /tools HTTP/1.1" 200 OK
  2026-01-09 11:13:11,350 httpx 127295 INFO HTTP Request: GET http://10.168.6.34:7861/tools "HTTP/1.1 200 OK"
  INFO: 10.168.6.34:41242 - "GET /tools HTTP/1.1" 200 OK
  2026-01-09 11:13:11,653 httpx 127295 INFO HTTP Request: GET http://10.168.6.34:7861/tools "HTTP/1.1 200 OK"
  INFO: 10.168.6.34:41244 - "GET /tools HTTP/1.1" 200 OK
  2026-01-09 11:13:14,211 httpx 127295 INFO HTTP Request: GET http://10.168.6.34:7861/tools "HTTP/1.1 200 OK"
  INFO: 10.168.6.34:41246 - "GET /tools HTTP/1.1" 200 OK
  2026-01-09 11:13:21,924 httpx 127295 INFO HTTP Request: GET http://10.168.6.34:7861/tools "HTTP/1.1 200 OK"
  INFO: 10.168.6.34:41248 - "GET /tools HTTP/1.1" 200 OK
  2026-01-09 11:13:31,706 httpx 127295 INFO HTTP Request: GET http://10.168.6.34:7861/tools "HTTP/1.1 200 OK"
  INFO: 10.168.6.34:41250 - "POST /chat/chat/completions HTTP/1.1" 200 OK
  2026-01-09 11:13:32,189 httpx 127295 INFO HTTP Request: POST http://10.168.6.34:7861/chat/chat/completions "HTTP/1.1 200 OK"
  2026-01-09 11:13:32,190 openai._base_client 127027 INFO Retrying request to /chat/completions in 0.439184 seconds
  2026-01-09 11:13:32,632 openai._base_client 127027 INFO Retrying request to /chat/completions in 0.810284 seconds
  2026-01-09 11:13:33.444 | ERROR | chatchat.server.api_server.openai_routes:generator:105 - openai request error: Connection error.

[dosubot]

你的 OpenAI embedding 连接错误很可能是配置或网络问题导致的。可以按以下步骤排查：

1. API base URL 检查：你现在配置的是 http://127.0.0.1:8000/v1，但如果你用的是 OpenAI 官方 API，应该填 https://api.openai.com/v1。只有本地代理（如 LocalAI）才用 localhost 地址。请确认 base URL 是否正确 参考。

2. embedding model 名称：OpenAI 官方 embedding 模型通常是 text-embedding-ada-002，而不是 text-embedding-v1。建议改为 text-embedding-ada-002，并在配置文件（如 settings.py 或 model_providers.yaml）里保持一致 参考。

3. API Key 检查：确认你的 OpenAI API key 有效且有额度。可以用 curl 测试：

   curl https://api.openai.com/v1/embeddings \
     -H "Authorization: Bearer sk-xxxx" \
     -H "Content-Type: application/json" \
     -d '{"model":"text-embedding-ada-002","input":["test"]}'

   如果能返回 embedding，说明 key 没问题。

4. 网络连通性：确保服务器能访问 OpenAI（没有被防火墙或代理阻断）。可以直接 ping 或 curl OpenAI API 测试。

5. Langchain 兼容性问题：如果配置都没问题但还是报错，有可能是 langchain 的 OpenAIEmbeddings 类兼容性问题。可以参考 这个方案 自定义 embedding 类，直接用 openai 官方 SDK 调用 embedding API。

   示例代码如下：

   from typing import List
   from langchain_core.embeddings import Embeddings
   from openai import OpenAI
   
   class CustomOpenAIEmbeddings(Embeddings):
       def __init__(self, model, openai_api_base, openai_api_key, timeout=60, batch_size=128):
           self._client = OpenAI(api_key=openai_api_key, base_url=openai_api_base, timeout=timeout)
           self.model = model
           self.batch_size = batch_size
   
       def embed_query(self, text: str) -> List[float]:
           return self.embed_documents([text])[0]
   
       def embed_documents(self, texts: list[str]) -> list[list[float]]:
           vecs: List[List[float]] = []
           bs = self.batch_size
           for i in range(0, len(texts), bs):
               batch = texts[i:i+bs]
               resp = self._client.embeddings.create(model=self.model, input=batch)
               vecs.extend([d.embedding for d in resp.data])
           return vecs

6. SSE/ASGI 错误：这些报错通常是 embedding/model API 没连通导致的二次错误，前面问题解决后一般会消失 参考。

建议优先检查 base URL 和 embedding model 名称，确认 API key 和网络没问题。如果还是不行，可以尝试自定义 embedding 类。

To reply, just mention @dosu.

---

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

   

[shengmengru]

我已经找出问题所在，现在可以上传文件并可以生成知识向量库，但是进行RAG对话的时候，模型不能从知识库中拉取信息回答，反馈结果是NONE。INFO: 10.168.6.34:41644 - "GET /knowledge_base/list_knowledge_bases HTTP/1.1" 200 OK
2026-01-09 11:33:19,729 httpx 141385 INFO HTTP Request: GET http://10.168.6.34:7861/knowledge_base/list_knowledge_bases "HTTP/1.1 200 OK"
INFO: 10.168.6.34:41646 - "GET /knowledge_base/list_knowledge_bases HTTP/1.1" 200 OK
2026-01-09 11:33:31,841 httpx 141385 INFO HTTP Request: GET http://10.168.6.34:7861/knowledge_base/list_knowledge_bases "HTTP/1.1 200 OK"
2026-01-09 11:33:31,966 openai._base_client 141385 INFO Retrying request to /chat/completions in 0.384337 seconds
2026-01-09 11:33:32,353 openai._base_client 141385 INFO Retrying request to /chat/completions in 0.938608 seconds
这个是终端输出信息

[dosubot]

你现在遇到的 RAG 返回 NONE，通常是因为知识库检索没有命中任何文档。根据代码和配置，主要排查点如下：

1. score_threshold 设置过高：score_threshold 越低，检索越严格，越高越宽松。建议设置为 0.5 左右，或者直接设为 2.0（不筛选），避免因为阈值过高导致检索不到内容。参考

2. top_k 太小：top_k 控制返回的文档数量，建议设为 5 或更大，避免遗漏相关内容。参考

3. embedding 模型和知识库不一致：切换 embedding 后必须重建知识库（重建向量库），否则检索时 embedding 不兼容会导致无法命中。参考

4. 检索内容本身问题：可以用你上传文档中的一句原文做检索测试，排除分词、embedding 失配等问题。

5. embedding 模型配置：如果你用 OpenAI，建议 embedding model 用 text-embedding-ada-002 或 text-embedding-3-small，不要用 text-embedding-v1，并确保 API base URL 是 https://api.openai.com/v1。参考

6. 知识库和向量库类型一致：知识库创建和检索时 embedding model、向量库类型要一致，否则会导致检索不到内容。参考

7. httpx 兼容性问题：如果遇到 NoneType 报错，可以尝试升级 httpx 到 0.27.2：pip install httpx==0.27.2 -U。参考

建议你先检查 score_threshold 和 top_k 配置，确认 embedding model 和知识库一致，并重建知识库后再试。如果还有问题，可以贴一下具体的检索参数和你上传的测试内容。

To reply, just mention @dosu.

---

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

   

[shengmengru]

我使用的是onepai平台，并且使用了python chatchat/cli.py kb --recreate-vs 语句，但还是反馈为NONE

[dosubot]

你遇到 RAG 检索始终返回 NONE，结合 onepai 平台和 vector store 重建流程，建议重点排查以下几个点：

1. embedding 模型和 API 配置：onepai 平台默认 embedding 模型是 text-embedding-v1 或 Embedding-V1，API base URL 通常为 http://127.0.0.1:3000/v1，请确保这两个配置与你实际 onepai 服务一致（比如端口、模型名、API key），否则 embedding 访问会失败，导致知识库检索不到内容。参考

2. embedding 模型可用性：系统会在重建向量库和检索前自动调用 check_embed_model() 检查 embedding 是否可用。如果 embedding 访问失败，知识库检索会直接返回空结果。你可以用 python chatchat/cli.py check-embed --embed-model text-embedding-v1（或你实际用的模型名）测试 embedding 是否能正常返回向量。参考

3. score_threshold 和 top_k：默认 score_threshold=2.0 表示不筛选，建议设置为 0.5 左右，top_k 建议设为 3 或更大。阈值过高或 top_k 太小都可能导致检索不到内容。参考

4. 知识库 embedding 信息一致性：知识库会记录创建时用的 embedding 模型，重建向量库时要确保用的 embedding 和检索时一致，否则会导致检索不到内容。参考

5. 实际文档是否入库：可以通过 /knowledge_base/list_files 或 /knowledge_base/search_docs（用原文片段做 query）接口确认文档是否被正确切分、嵌入并入库。

建议你：

• 检查 onepai embedding 服务和 API 配置是否可用，能否正常返回向量。
• 明确重建向量库时用的 embedding 模型和检索时一致。
• 调低 score_threshold，增大 top_k。
• 用原文片段直接检索，确认文档是否真的入库。

如果还有问题，可以贴下 embedding 配置、API base URL、score_threshold、top_k 具体值，以及重建向量库和检索时的日志片段。这样能更快定位问题。

To reply, just mention @dosu.

---

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

   

[shengmengru]

MODEL_PLATFORMS:

• platform_name: oneapi
  platform_type: oneapi
  api_base_url: https://dashscope.aliyuncs.com/compatible-mode/v1
  api_key: sk-
  api_proxy: ''
  api_concurrencies: 5
  auto_detect_model: false
  llm_models:
  • chatglm_pro
  • chatglm_turbo
  • chatglm_std
  • chatglm_lite
  • qwen-turbo
  • qwen-plus
  • qwen-max
  • qwen-max-longcontext
  • ERNIE-Bot
  • ERNIE-Bot-turbo
  • ERNIE-Bot-4
  • SparkDesk
    embed_models:
  • text-embedding-v1
  • Embedding-V1
    text2image_models: []
    image2text_models: []
    rerank_models: []
    speech2text_models: []
    text2speech_models: [] 这是我model_setting配置文件中的设置，key我给删了，但是是正确的，那我需要把其他文件中相关的 api_base_url也设置成这个吗？

[dosubot]

不需要在其他文件单独设置 api_base_url，只需在 model_settings.yaml（或 model_setting 配置文件）下的 MODEL_PLATFORMS 配置好即可。系统会自动根据这里的配置，把 LLM 和 embedding 的请求路由到对应的 api_base_url，无需在 settings.py、model_providers.yaml 等其他文件重复设置，也不需要重启服务，改完立即生效（v0.3.1+ 支持）参考。

如果你有多个平台（比如 LLM 和 embedding 分别走不同的 API），就在 MODEL_PLATFORMS 里分别配置不同的 platform_name、api_base_url 和模型列表即可。只要保证 DEFAULT_LLM_MODEL 和 DEFAULT_EMBEDDING_MODEL 指向的模型名和这里一致即可参考。

除非你有自定义代码或特殊部署需求，否则只需要维护 model_settings.yaml 这一个地方。

To reply, just mention @dosu.

---

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

   

[shengmengru]

kb_setting配置文件中有如下说明，是否是在页面中创建数据库时没有加相关介绍，agent没有调用，所以会显示NONE。但是我创建的时候没有介绍，系统会自动加上介绍。

每个知识库的初始化介绍，用于在初始化知识库时显示和Agent调用，没写则没有介绍，不会被Agent调用。

KB_INFO:
samples: 关于本项目issue的解答

[dosubot]

不会自动加介绍。kb_setting 里的 KB_INFO 只是为每个知识库提供一个初始化简介（比如 samples: 关于本项目issue的解答），用于页面显示和 Agent 检索知识库时的参考。如果你在页面或 API 创建知识库时没有填写简介（kb_info），系统不会自动生成介绍，kb_info 字段会是空的或用 kb_setting 里的默认值（如果有对应条目），否则就是“关于{知识库名}的知识库”这种兜底字符串参考。

没有简介只会影响 Agent 选择和调用知识库时的上下文，不会影响 RAG 检索本身。如果 RAG 返回 NONE，通常是检索参数、embedding 或向量库配置等问题，不是因为没有 kb_info参考。

如果需要让 Agent 更好地利用知识库，可以在知识库创建后通过页面或 API 编辑/补充 kb_info 简介参考。

To reply, just mention @dosu.

---

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

   

[shengmengru]

我的langchain版本是0.1，但是能运行成功，就是在RAG对话中不能成功检索出文档并回答，是否是跟langchain版本有关。我写了一个测试文件如下：import sys
sys.path.append(".")
from chatchat.server.utils import get_Embeddings

测试阿里云百炼的text-embedding-v2

try:
embed = get_Embeddings(embed_model="text-embedding-v2", local_wrap=False)
# 生成测试文本的向量
vec = embed.embed_query("介绍一下允思拓公司")
print(f"嵌入模型调用成功！向量长度：{len(vec)}")
except Exception as e:
print(f"嵌入模型调用失败：{str(e)}") 可以成功输出。分析可能出问题的文件和位置

[dosubot]

你的问题大概率不是由 langchain 0.1 版本导致的。Langchain-Chatchat 的向量检索和 RAG 逻辑已经适配了 langchain 0.1 的新 API（比如 search_type/score_threshold），理论上是兼容的，RAG 返回 NONE 更常见的原因是配置或数据问题，而不是版本不兼容参考。

你 embedding 能正常生成向量，说明模型和 API 没问题。RAG 检索不到内容，常见原因有：

1. score_threshold 设置过高：建议设置为 0.5 左右，或者 2.0（不筛选），否则容易检索不到内容。
2. top_k 太小：建议设为 5 或更大，避免遗漏相关内容。可以在 retrievers/ensemble.py 和 retrievers/vectorstore.py 里调整默认值参考。
3. 知识库向量库未正确重建：切换 embedding 后一定要用 python chatchat/cli.py kb --recreate-vs，否则新旧 embedding 不兼容会导致检索不到内容参考。
4. API/环境变量配置问题：如 API_SERVER、no_proxy、端口等配置错误，会导致 API 请求失败，RAG 也会返回 NONE。no_proxy 建议为 127.0.0.1,localhost,[::1]，端口必须纯数字参考。
5. httpx 版本冲突：务必保证 chatchat 环境下 httpx 版本为 0.27.2，且没有其他推理框架（如 Ollama）依赖冲突参考。

建议你：

• 检查 score_threshold 和 top_k 配置。
• 用上传文档中的原文做一次检索测试，确认文档是否真的入库。
• 检查 API 端口、no_proxy、httpx 版本等环境变量。
• 如果用的是自定义 embedding，确保 model_settings.yaml 里的 DEFAULT_EMBEDDING_MODEL 和实际一致。

如果还有问题，可以贴下你的检索参数、环境变量和重建向量库时的日志片段，这样能更快定位问题。

To reply, just mention @dosu.

---

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

   

[shengmengru]

我的问题还没解决，一下是我的model_setting文件，

默认选用的 LLM 名称

DEFAULT_LLM_MODEL: qwen-plus

默认选用的 Embedding 名称

DEFAULT_EMBEDDING_MODEL: text-embedding-v2

AgentLM模型的名称 (可以不指定，指定之后就锁定进入Agent之后的Chain的模型，不指定就是 DEFAULT_LLM_MODEL)

Agent_MODEL: ''

默认历史对话轮数

HISTORY_LEN: 3

大模型最长支持的长度，如果不填写，则使用模型默认的最大长度，如果填写，则为用户设定的最大长度

MAX_TOKENS:

LLM通用对话参数

TEMPERATURE: 0.7

支持的Agent模型

SUPPORT_AGENT_MODELS:

• chatglm3-6b
• glm-4
• openai-api
• Qwen-2
• qwen2-instruct
• gpt-3.5-turbo
• gpt-4o

LLM模型配置，包括了不同模态初始化参数。

model 如果留空则自动使用 DEFAULT_LLM_MODEL

LLM_MODEL_CONFIG:
preprocess_model:
model: ''
temperature: 0.05
max_tokens: 4096
history_len: 10
prompt_name: default
callbacks: false
llm_model:
model: ''
temperature: 0.9
max_tokens: 4096
history_len: 10
prompt_name: default
callbacks: true
action_model:
model: ''
temperature: 0.01
max_tokens: 4096
history_len: 10
prompt_name: ChatGLM3
callbacks: true
postprocess_model:
model: ''
temperature: 0.01
max_tokens: 4096
history_len: 10
prompt_name: default
callbacks: true
image_model:
model: sd-turbo
size: 256*256
MODEL_PLATFORMS:

• platform_name: aliyun
  platform_type: oneapi
  api_base_url: https://dashscope.aliyuncs.com/compatible-mode/v1
  api_key: sk-省略
  api_proxy: ''
  api_concurrencies: 5
  auto_detect_model: false
  llm_models:
  embed_models:
  • text-embedding-v2
    text2image_models: []
    image2text_models: []
    rerank_models: []
    speech2text_models: []
    text2speech_models: []