无法启动成功，无法使用 #1354
Comments
|
hi @yuzhi-jiang thanks for reporting. As for the first subissue you mentioned here, cannot connect to postgres with password contains @paopa could you check the config yaml here ? see what's going on. |
|
Hi @yuzhi-jiang, I noticed the error is due to leaking the kwargs for llm. If you don't want specified kwargs for llm, at least leave e.g. models:
- api_base: https://api.openai.com/v1
api_key_name: LLM_OPENAI_API_KEY
kwargs: {}
model: gpt-4o-mini-2024-07-18 |
|
你好，我按要求补上了 ========== |
Right. I think you could remove |
你好，今天试了移除配置 |
|
Hi @yuzhi-jiang, can you change 127.0.0.1 to host.docker.internal for the api_base property? I think using 127.0.0.1 cannot be connected from the container. |
|
host.docker.internal?
|
Since the AI service is a container and needs to access your Ollama, I’m not sure if the way you used it works. But you can connect to the host on your machine using host.docker.internal. Check out this link for more details: https://docs.docker.com/desktop/features/networking/#i-want-to-connect-from-a-container-to-a-service-on-the-host |
我理解你的意思了，是 AI service的网络问题是吧，明天我会试试，感谢你的回答 |
最新问题: ollama 已经可以收到请求:
问题1.当我使用如下配置的embedder模型 错误信息是： 2.当我是使用 错误信息就变成了AuthenticationError： |
|
For the first question, I think the error might be because you forgot to update the pipeline setting in the config.yaml to match the current embedding model. e.g. |
|
For the second question, could you share the current config.yaml you’re using? It looks like you might be missing an API key for the OpenAI-compatible endpoint. Or, if you prefer, you can refer to my Ollama config example. I’ve made a couple of changes. I’ve updated the model with the prefix |
是的，第一个和第二个应该是同一个问题的 config.yml .env ai service log: |
@yuzhi-jiang 这个问题我已经解决，直接修改源代码 |
I noticed that the naming issue for the embedder is inconsistent. I think you can modify all instances of ‘openai’ to ‘ollama’ to fix this in the following pipeline section.
By the way, I think you should either remove the QDRANT_API_KEY or blank it out. If you leave it in, it might cause other authentication errors. |
感谢提出，不过你能否根据我的配置修改,以防我认识错误，避免多次错误修改 |
|
@yuzhi-jiang, the config I think will be like the following, and I also noticed you're using deepseek-r1:14b, and I think the model may not be good at responding to the structure output. Refer to #1121 (comment) you need to change a model from my opinion.
type: llm
provider: litellm_llm
models:
- api_base: http://192.168.177.95:11434
model: ollama/deepseek-r1:14b
timeout: 600
kwargs:
n: 1
temperature: 0
---
type: embedder
provider: litellm_embedder
models:
- model: ollama/nomic-embed-text:latest
api_base: http://192.168.177.95:11434
timeout: 600
---
type: engine
provider: wren_ui
endpoint: http://wren-ui:3000
---
type: document_store
provider: qdrant
location: http://qdrant:6333
embedding_model_dim: 768
timeout: 120
recreate_index: true
---
type: pipeline
pipes:
- name: db_schema_indexing
embedder: litellm_embedder.ollama/nomic-embed-text:latest
document_store: qdrant
- name: historical_question_indexing
embedder: litellm_embedder.ollama/nomic-embed-text:latest
document_store: qdrant
- name: table_description_indexing
embedder: litellm_embedder.ollama/nomic-embed-text:latest
document_store: qdrant
- name: db_schema_retrieval
llm: litellm_llm.ollama/deepseek-r1:14b
embedder: litellm_embedder.ollama/nomic-embed-text:latest
document_store: qdrant
- name: historical_question_retrieval
embedder: litellm_embedder.ollama/nomic-embed-text:latest
document_store: qdrant
- name: sql_generation
llm: litellm_llm.ollama/deepseek-r1:14b
engine: wren_ui
- name: sql_correction
llm: litellm_llm.ollama/deepseek-r1:14b
engine: wren_ui
- name: followup_sql_generation
llm: litellm_llm.ollama/deepseek-r1:14b
engine: wren_ui
- name: sql_summary
llm: litellm_llm.ollama/deepseek-r1:14b
- name: sql_answer
llm: litellm_llm.ollama/deepseek-r1:14b
engine: wren_ui
- name: sql_breakdown
llm: litellm_llm.ollama/deepseek-r1:14b
engine: wren_ui
- name: sql_expansion
llm: litellm_llm.ollama/deepseek-r1:14b
engine: wren_ui
- name: sql_explanation
llm: litellm_llm.ollama/deepseek-r1:14b
- name: semantics_description
llm: litellm_llm.ollama/deepseek-r1:14b
- name: relationship_recommendation
llm: litellm_llm.ollama/deepseek-r1:14b
engine: wren_ui
- name: question_recommendation
llm: litellm_llm.ollama/deepseek-r1:14b
- name: question_recommendation_db_schema_retrieval
llm: litellm_llm.ollama/deepseek-r1:14b
embedder: litellm_embedder.ollama/nomic-embed-text:latest
document_store: qdrant
- name: question_recommendation_sql_generation
llm: litellm_llm.ollama/deepseek-r1:14b
engine: wren_ui
- name: chart_generation
llm: litellm_llm.ollama/deepseek-r1:14b
- name: chart_adjustment
llm: litellm_llm.ollama/deepseek-r1:14b
- name: intent_classification
llm: litellm_llm.ollama/deepseek-r1:14b
embedder: litellm_embedder.ollama/nomic-embed-text:latest
document_store: qdrant
- name: data_assistance
llm: litellm_llm.ollama/deepseek-r1:14b
- name: sql_pairs_indexing
document_store: qdrant
embedder: litellm_embedder.ollama/nomic-embed-text:latest
- name: sql_pairs_deletion
document_store: qdrant
embedder: litellm_embedder.ollama/nomic-embed-text:latest
- name: sql_pairs_retrieval
document_store: qdrant
embedder: litellm_embedder.ollama/nomic-embed-text:latest
llm: litellm_llm.ollama/deepseek-r1:14b
- name: preprocess_sql_data
llm: litellm_llm.ollama/deepseek-r1:14b
- name: sql_executor
engine: wren_ui
- name: sql_question_generation
llm: litellm_llm.ollama/deepseek-r1:14b
- name: sql_generation_reasoning
llm: litellm_llm.ollama/deepseek-r1:14b
- name: sql_regeneration
llm: litellm_llm.ollama/deepseek-r1:14b
engine: wren_ui
---
settings:
column_indexing_batch_size: 50
table_retrieval_size: 10
table_column_retrieval_size: 100
allow_using_db_schemas_without_pruning: false
query_cache_maxsize: 1000
query_cache_ttl: 3600
langfuse_host: https://cloud.langfuse.com
langfuse_enable: true
logging_level: DEBUG
development: false
|
|
@paopa 好的我马上试试，不知道你认为ollama跑的什么模型合适，我可以马上pull，我目前还有空间，但是显存不太够，所以大概只能流畅运行14b左右的模型 |
现在ai service 不报错了，页面请求无反馈，所以我目前的目标是能使用，以至于性能和准确度的问题才能在申请更多的显存 |
|
@yuzhi-jiang Do you see if the background server reported this error? I'm stuck here right now |
|
Hi @yuzhi-jiang, @jonathanyb, I tried to use e.g. model: ollama_chat/phi4:14b |
是的，可能是模型返回问题? 报错: 当我使用 则是页面提出问题后报错： 当前我的config.yaml： |
|
Hey @yuzhi-jiang, I know it’s a bit of a hassle, but for your config, you need to change all here is the version i've added the alias to simplify the pipe management. |
|
@paopa 感谢你帮我修改配置，我使用了你建议的最新配置，但我依然遇到如下错误，我注意到wrenai在3天前有更新，我是否需要更新docker image 的version?
|
|
Hi @yuzhi-jiang, could you check again to make sure the config is correct? The message indicates that it cannot map to your LLM provider instance for the semantics description object. It might be a typo or something else, as I don’t have this issue with the same config. You could also try restarting the AI service to see if that helps.
I don’t think it matters much with the new release, but I’d love for you to try the latest version. If you have a chance, please let us know what you think. Your feedback is super valuable to us and helps us make this project even better. Thanks a bunch! |
@paopa 你好，经过我检查，如果不算配置错误，应该不算拼写问题，我copy以上配置的，为了方便展示，我做了diff
|
我认为config.yaml，或许可以简化，或者加一层，或许在代码中是之间读取这个config.yaml，但是对于不熟知的用户来说，却会造成一定的使用门槛，配置中写的规则(比如必填 ollama_url/v1)似乎可以不必遵守，我认为应该以用户最简配置为主，如果是customModel，则只需要填写modelname,base_url,api_key即可了 |
|
Hey @yuzhi-jiang, if the files are the same, I’m not sure why the error occurred. Maybe you could try stopping the container and restarting it again? Thanks for the suggestion! Actually, we’ve found a lot of community members on GitHub and Discord who are having trouble setting up the config. We’ve already raised an issue about it and added it to our backlog. We’ll plan how to solve it better! |
|
Hi @yuzhi-jiang, I got a same issue with you from other community member. Can you give Wren AI 0.15.4 a try? We just released it this week, and I’ve been using it and it’s working well with the config! |
|
@paopa 你好，我今天使用了 0.15.4 ，修改了env中的部分配置，其他没修改 依旧有报错 |
Hey! I can't reproduce this issue with the above config. Can you check the container version for the Wren AI service is correct or is the config.yaml correct? Thank you! |
and others
前提ifno: ollama本地部署，已有配置中相关模型,启动时docker-compose.yml相关端口未占用， wrenai version: 0.15.3
问题1：pgsql，password带有符号#时,无法识别，
问题二：无法使用
这是返回值：
以下是config.yml
.env配置：
以下是一些容器的log:
wrenai-wren-ai-service-1
The text was updated successfully, but these errors were encountered: