How hitomi-team/shimeji works
Project type and purpose
• This is a Python library (“shimeji”) for building open-domain chatbots in a platform- and API-agnostic way.
• Key indicators:
• shimeji/ package with modules like preprocessor, postprocessor, model_provider, memory, etc.
• Alembic migration config (alembic.ini, alembic/versions/…) for a memory table.
• README describes “A platform and API Agnostic library for writing Open Domain and General Chatbots.”
File structure
Top-level directories/files:
• shimeji/ – core Python package
– preprocessor.py, postprocessor.py: pipeline stages for cleaning or shaping user input/output
– model_provider.py: plugin interface to LLM APIs or local models
– memory.py + memorystore_provider.py + sqlcrud.py: in-process memory and memory persistence layer
– util.py: helper routines
– shimeji.py: orchestrator that ties all modules into a chatbot pipeline
• alembic/ + alembic.ini: database migration setup for persistent memory (SQLAlchemy + Alembic)
• docs/: Sphinx documentation source
• requirements.txt + setup.py: packaging and dependency declarations
• test.py, test_async.py: example or test scripts
Architecture insights from README
• Persistence (SQL) is optional but pluggable via memorystore_provider.