Encounter a bug：TypeError: list indices must be integers or slices, not str

[chenyihang1993]

Excute：deepsearcher --query "小组成员有哪些？"
Result:

<think> Select agent [ChainOfRAG] to answer the query [小组成员有哪些？] </think>

>> Iteration: 1

<think> Perform search [谁是小组的成员？] on the vector DB collections: ['deepsearcher'] </think>

<search> Search [谁是小组的成员？] in [deepsearcher]...  </search>

supported_doc_indices:  ['\\Document 0']
Traceback (most recent call last):
  File "/opt/conda/bin/deepsearcher", line 8, in <module>
    sys.exit(main())
  File "/root/deep-searcher/deepsearcher/cli.py", line 61, in main
    final_answer, refs, consumed_tokens = query(args.query, max_iter=args.max_iter)
  File "/root/deep-searcher/deepsearcher/online_query.py", line 10, in query
    return default_searcher.query(original_query, max_iter=max_iter)
  File "/root/deep-searcher/deepsearcher/agent/rag_router.py", line 69, in query
    answer, retrieved_results, n_token_retrieval = agent.query(query, **kwargs)
  File "/root/deep-searcher/deepsearcher/agent/chain_of_rag.py", line 197, in query
    all_retrieved_results, n_token_retrieval, additional_info = self.retrieve(query, **kwargs)
  File "/root/deep-searcher/deepsearcher/agent/chain_of_rag.py", line 182, in retrieve
    supported_retrieved_results, n_token2 = self._get_supported_docs(
  File "/root/deep-searcher/deepsearcher/agent/chain_of_rag.py", line 167, in _get_supported_docs
    supported_retrieved_results = [retrieved_results[i] for i in supported_doc_indices]
  File "/root/deep-searcher/deepsearcher/agent/chain_of_rag.py", line 167, in <listcomp>
    supported_retrieved_results = [retrieved_results[i] for i in supported_doc_indices]
TypeError: list indices must be integers or slices, not str

config.yaml:

provide_settings:
  llm:
    # provider: "OpenAI"
    # config:
    #   model: "gpt-4o-mini"  # "gpt-o1-mini"
#      api_key: "sk-xxxx"  # Uncomment to override the `OPENAI_API_KEY` set in the environment variable
#      base_url: ""

#    provider: "DeepSeek"
#    config:
#      model: "deepseek-chat"  # "deepseek-reasoner"
##      api_key: "sk-xxxx"  # Uncomment to override the `DEEPSEEK_API_KEY` set in the environment variable
##      base_url: ""

#    provider: "SiliconFlow"
#    config:
#      model: "deepseek-ai/DeepSeek-V3"
##      api_key: "xxxx"  # Uncomment to override the `SILICONFLOW_API_KEY` set in the environment variable
##      base_url: ""

#    provider: "PPIO"
#    config:
#      model: "deepseek/deepseek-v3/community"
##      api_key: "sk_xxxxxx"  # Uncomment to override the `PPIO_API_KEY` set in the environment variable
##      base_url: ""

#    provider: "TogetherAI"
#    config:
#      model: "deepseek-ai/DeepSeek-V3"
##      api_key: "xxxx"  # Uncomment to override the `TOGETHER_API_KEY` set in the environment variable

#    provider: "AzureOpenAI"
#    config:
#      model: ""
#      api_version: ""
##      azure_endpoint: "xxxx"  # Uncomment to override the `AZURE_OPENAI_ENDPOINT` set in the environment variable
##      api_key: "xxxx"  # Uncomment to override the `AZURE_OPENAI_KEY` set in the environment variable

   provider: "Ollama"
   config:
     model: "qwen2.5:3b"
#      base_url: ""

  embedding:
  #   provider: "OpenAIEmbedding"
  #   config:
  #     model: "text-embedding-ada-002"
#      api_key: ""  # Uncomment to override the `OPENAI_API_KEY` set in the environment variable
#      base_url: "" # Uncomment to override the `OPENAI_BASE_URL` set in the environment variable
#      dimension: 1536 # Uncomment to customize the embedding dimension 


   provider: "MilvusEmbedding"
   config:
     model: "BAAI/bge-large-zh-v1.5"

#    provider: "VoyageEmbedding"
#    config:
#      model: "voyage-3"
##      api_key: ""  # Uncomment to override the `VOYAGE_API_KEY` set in the environment variable

#    provider: "BedrockEmbedding"
#    config:
#      model: "amazon.titan-embed-text-v2:0"
##      aws_access_key_id: ""  # Uncomment to override the `AWS_ACCESS_KEY_ID` set in the environment variable
##      aws_secret_access_key: ""  # Uncomment to override the `AWS_SECRET_ACCESS_KEY` set in the environment variable
    
#    provider: "SiliconflowEmbedding"
#    config:
#      model: "BAAI/bge-m3"
# .    api_key: ""   # Uncomment to override the `SILICONFLOW_API_KEY` set in the environment variable   

  file_loader:
    provider: "PDFLoader"
    config: {}

#    provider: "JsonFileLoader"
#    config:
#      text_key: ""

#    provider: "TextLoader"
#    config: {}

#    provider: "UnstructuredLoader"
#    config: {}

  web_crawler:
    provider: "FireCrawlCrawler"
    config: {}

    # provider: "Crawl4AICrawler"
    # config: # Uncomment to custom browser configuration for Crawl4AI
    #   browser_config:
    #     headless: false
    #     proxy: "http://127.0.0.1:7890"
    #     chrome_channel: "chrome"
    #     verbose: true
    #     viewport_width: 800
    #     viewport_height: 600
    
#    provider: "JinaCrawler"
#    config: {}

  vector_db:
    provider: "Milvus"
    config:
      default_collection: "deepsearcher"
      uri: "./milvus.db"
      token: "root:Milvus"
      db: "default"

query_settings:
  max_iter: 3

load_settings:
  chunk_size: 1500
  chunk_overlap: 100

[zc277584121]

@chenyihang1993 the llm is too small to follow the format parsing prompt, ref this: #99 (comment)