How manan025/evslm works
Project type and purpose
• From the README (“evsLM”) and the presence of two Jupyter notebooks (RAG.ipynb, main.ipynb) plus a small utility module (utils/helper.py), this is primarily a lightweight Python-based tool or library, explored and driven via notebooks. It appears to implement retrieval-augmented workflows or PDF processing (poppler is required).
File structure
• Top level:
• RAG.ipynb, main.ipynb – user-facing notebooks (UI/experimentation)
• pyproject.toml & uv.lock – Python packaging/config & locked dependencies
• utils/ – your core Python logic (helper.py)
• external requirement: poppler (PDF rendering tools)
• There is no separate frontend/backend or database service—this is a single-process, script-based tool.
README for additional insights
• The focus is on running code inside Jupyter, loading/processing data via helper.py, and leveraging poppler for PDF operations.
Components
a. Jupyter Notebook Clients (RAG.ipynb, main.ipynb)
b. Core Library (utils/helper.py, and any other modules installed via pyproject.toml)
c. External Dependencies:
– poppler CLI / library (PDF rendering)
– any other Python packages listed in pyproject.toml
d. Local File System (data input/output – e.g. PDFs, intermediate files)
Relationships & data flow
• Notebook → imports → Core Library (helper.py)
• Core Library → invokes → External Dependencies (poppler)
• Core Library ↔ reads/writes ↔ Local File System
• Notebook ↔ displays ↔ Data/results (plots, text)
Architectural patterns & principles
• Separation of concerns:
• “UI” layer in notebooks
• “Business logic” in utils/helper.py
• “Infrastructure” calls out to poppler and file system
• Modularity via Python packaging (pyproject.toml)
• Reproducibility via locked dependencies (uv.lock)
Detail
• Even though this is a small codebase, break out each piece: the two notebooks, the helper module, the poppler dependency, and the file system.
• Every flow: notebook → Python function → poppler → back to Python → notebook display.