Using the Copilot plugin in Obsidian to interact with local data

I. Usage Scenarios#

There is a large amount of stored data, and many of them are forgotten
Need to have a conversation with the data, such as finding a sentence expressed by a leader several years ago
Personal emphasis on privacy, unwilling to feed public language models on the Internet
Local computer configuration is decent, with at least 8GB of memory

In the Obsidian note-taking software, use the copilot plugin to call the ollama service
Use ollama to call large language models, such as llama2-chinese or qwen:4b
Implement functions such as Q&A, single-question Q&A, and full-database chat

Install ollama from the official website on the computer
Open the terminal or command line mode as an administrator
Enter the command ollama run llama2-chinese
Wait for the download and installation of the llama2-chinese model to complete
Enter /bye to exit ollama
Enter the command ollama pull nomic-embed-text
Wait for the download and installation of nomic-embed-text to complete
Enter /bye to exit ollama
- Exit the running ollama from the taskbar in the lower right corner of the desktop
- Exit all ollama.exe processes from the task manager
- These two steps are crucial
- Otherwise, an error LangChain error: TypeError: Failed to fetch will occur later
Enter set OLLAMA_ORIGINS=app://obsidian.md*
Enter ollama serve to start the service
Keep the terminal or command line window running
Install the copilot plugin in Obsidian
- Set the default model to ollama(local)
- Choose the llama2-chinese model for ollama
- Set the ollama base URL to http://localhost:11434
- Choose the embedding API as nomic-embed-text
- Press ctrl+p to bring up the Obsidian command panel
- Enter copilot and select the desired command
- The copilot session box will appear on the right sidebar
- Choose the chat, single-question Q&A, or full-database Q&A mode as needed

With a well-configured computer, this may be a good choice
Under the premise of emphasizing privacy protection, this will be the best choice
Positive aspect: Index dialogue with the entire database
Shortcoming: Need to keep the terminal or command line window open at all times