The work is inspired by JayZeeDesign researcher-gpt

• app.py is the original file
• app2.py is the file created by me based on langchain v0.2

Key points

1. To test the FastAPI service
   • run uvicorn app2:api --reload
   • go to the swagger file under http://127.0.0.1/8000/docs
2. To run as streamlit webapp
   • uncomment the app function
   • run streamlit run app2.py
3. Import multiple versions of pydrantic
4. Use StructuredTool to create custom tools for agent
5. Create custom agent with custom tools and memory
6. Use Apify because
   • browserless costs too much
   • Apify actors are easier to integrate
   • But Apify webcrawler actor already summarizes the web content to some extent
7. To make it work on Windows 11,
   • Set langsmith version to be 0.0.87 required by langchiain-core.
   • Modify .venv\lib\site-packages\langchain_community\document_loaders\pebblo.py
   • Windows has no 'pwd' module. (Enjoy the open source)
   • Comment out 'import pwd' at line 5
   • Assign "unknown" to the document owner at line 264