Error occurred when using Custom OpenAI API LLM

[ki-ri]

Describe the bug I am using Custom OpenAI API LLM(llama.cpp Server) as LLM according to the manual below, when I searched something I got the error. (I tested the llama.cpp Server, and it worked well independently) https://docs.gptr.dev/docs/gpt-researcher/llms/llms#custom-openai-api-llm

• error:

  File "/Users/xxx/study/AI/tools/gpt-researcher/gpt_researcher/skills/researcher.py", line 149, in __get_context_by_search
  sub_queries.append(query)
  ^^^^^^^^^^^^^^^^^^
  AttributeError: 'str' object has no attribute 'append'
  INFO:     connection closed

• LLM config

  
  # set the custom OpenAI API url
  OPENAI_BASE_URL="http://localhost:8080"
  # set the custom OpenAI API key
  OPENAI_API_KEY="sk-no-key-required"
  TAVILY_API_KEY="my tavily key"

specify the custom OpenAI API llm model

FAST_LLM="openai:gpt-4o-mini"

specify the custom OpenAI API llm model

SMART_LLM="openai:gpt-4o" DOC_PATH=./my-docs

**To Reproduce**
- please see the information above

**Expected behavior**
- searched result displayed

**Screenshots**
- terminated unsuccessfully, and got no report.
<img width="1186" alt="image" src="https://github.com/user-attachments/assets/03132e61-8a4a-4c2a-bc69-a7e34fa91fd8">

**Desktop (please complete the following information):**
 - OS: macOS 13.5
 - Browser chrome

[ElishaKay]

Looks like the LLM is configured correctly for chat_completion (since that is a requirement for the auto-agent to be generated as seen in your screenshot), but perhaps the embedding stage is not configured correctly.

We'll need to create a separate function for testing just the embedding stage based on config variables - something similar to this for a an isolated embedding function.

In the meantime, check if your llama.cpp server has an embedding model added to it - the current GPTR version will try to use the "text-embedding-3-large" embedding model by default.