[Issue]: Anybody has a workable settings.yaml for qwen GraphRAG v3.0.6 ?

[shuzi]

Do you need to file an issue?

• I have searched the existing issues and this bug is not already filed.
• My model is hosted on OpenAI or Azure. If not, please look at the "model providers" issue and don't file a new one here.
• I believe this is a legitimate bug, not just a question. If this is a question, please use the Discussions area.

Describe the issue

I can not figure it out.....

Anyone has a workable GraphRAG v3.0.6 settings.yaml for qwen model and embedding ?

Steps to reproduce

No response

GraphRAG Config Used

# Paste your config here

Logs and screenshots

No response

Additional Information

• GraphRAG Version:
• Operating System:
• Python Version:
• Related Issues:

[telunyang]

I use the Ollama service with qwen3.5:27b and qwen3-embedding. My basic configuration is shown below:

completion_models:
  default_completion_model:
    model_provider: openai
    model: qwen3.5:27b
    auth_method: api_key 
    api_base: http://localhost:11434/v1
    api_key: ${GRAPHRAG_API_KEY}
    retry:
      type: exponential_backoff

embedding_models:
  default_embedding_model:
    model_provider: openai
    model: qwen3-embedding
    auth_method: api_key
    api_base: http://localhost:11434/v1
    api_key: ${GRAPHRAG_API_KEY}
    retry:
      type: exponential_backoff

You also need to update the vector_store configuration with more detailed settings:

vector_store:
  type: lancedb
  db_uri: output/lancedb
  vector_size: 4096
  index_schema:
    community_full_content:
      index_name: community_full_content
      id_field: id
      vector_field: vector
      vector_size: 4096
    entity_description:
      index_name: entity_description
      id_field: id
      vector_field: vector
      vector_size: 4096
    text_unit_text:
      index_name: text_unit_text
      id_field: id
      vector_field: vector
      vector_size: 4096
  text_unit_text:
    vector_size: 4096
  entity_description:
    vector_size: 4096

You may be able to remove some attributes (e.g., id_field: id and vector_field: vector) depending on your requirements.

[shuzi]

I use the Ollama service with qwen3.5:27b and qwen3-embedding. My basic configuration is shown below:

completion_models:
  default_completion_model:
    model_provider: openai
    model: qwen3.5:27b
    auth_method: api_key 
    api_base: http://localhost:11434/v1
    api_key: ${GRAPHRAG_API_KEY}
    retry:
      type: exponential_backoff

embedding_models:
  default_embedding_model:
    model_provider: openai
    model: qwen3-embedding
    auth_method: api_key
    api_base: http://localhost:11434/v1
    api_key: ${GRAPHRAG_API_KEY}
    retry:
      type: exponential_backoff

You also need to update the vector_store configuration with more detailed settings:

vector_store:
  type: lancedb
  db_uri: output/lancedb
  vector_size: 4096
  index_schema:
    community_full_content:
      index_name: community_full_content
      id_field: id
      vector_field: vector
      vector_size: 4096
    entity_description:
      index_name: entity_description
      id_field: id
      vector_field: vector
      vector_size: 4096
    text_unit_text:
      index_name: text_unit_text
      id_field: id
      vector_field: vector
      vector_size: 4096
  text_unit_text:
    vector_size: 4096
  entity_description:
    vector_size: 4096

You may be able to remove some attributes (e.g., id_field: id and vector_field: vector) depending on your requirements.

Thank you!

for chunking, what is your setup, my is:

chunking:
type: tokens
size: 1200
overlap: 100
encoding_model: cl100k_base

[telunyang]

Thank you!

for chunking, what is your setup, my is:

chunking: type: tokens size: 1200 overlap: 100 encoding_model: cl100k_base

The same as yours. 😊

For chunking, it depends on the embedding model you use. If the embedding model supports max_seq_length=8192, setting the token size to around 1200 should be fine (1200 < 8192).

As for the encoding_model, it is mainly used to count the number of tokens during the chunking phase and help you calculate how many chunks are needed.

[shuzi]

Thank you!
for chunking, what is your setup, my is:
chunking: type: tokens size: 1200 overlap: 100 encoding_model: cl100k_base

The same as yours. 😊

For chunking, it depends on the embedding model you use. If the embedding model supports max_seq_length=8192, setting the token size to around 1200 should be fine (1200 < 8192).

As for the encoding_model, it is mainly used to count the number of tokens during the chunking phase and help you calculate how many chunks are needed.

now I understand, the encoding_model config is only for chunking purpose.
Because I remember qwen use its own encoding, neither cl100k_base nor o200k_base

[GoodManWEN]

Is there any way to call qwen through online service api? Thanks.