How xlr-aex/kyrria works
Project type and purpose
• KYRRIA is an open-source, Python/Streamlit application for RSS/Atom feed aggregation, enriched reading, local storage of articles, relational analysis (NER via Google Gemini), and interactive visualizations (charts + D3.js graphs).
• It targets end users who want to collect, browse, store, and analyze news/information streams.
File structure
• Top level:
• Kyrria.py – single entry point (Streamlit app)
• requirements.txt – Python dependencies
• .devcontainer/ – configuration for VS Code Dev Containers
• LICENSE, README.md
• No separate frontend/backend directories: the project follows the typical Streamlit pattern where the “frontend” UI and “backend” logic live in the same Python process.
• Storage layer (probably SQLite or local file caching) is invoked from within Kyrria.py or its imported modules.
README for additional insights
• Describes five major UI pages (Home, Feed Manager, RSS Reader, Nodes, Dashboard, Entities & Relations).
• Mentions external services & libraries: feed RSS/Atom, caching, local DB, D3.js, Google Gemini (NER), various plotting modules.
• Deployment target: local Python 3.12 or Streamlit Cloud.
Main components & how to reflect them
a. Main components
• Streamlit Application (Python runtime)
• UI Pages / Controllers (Home, Feed Manager, RSS Reader, Dashboard, Nodes, Entities & Relations)
• Feed Fetcher & Parser (RSS/Atom ingestion)
• Cache Layer (in-memory or file)
• Local Storage (SQLite or flat files)
• Analytics / NER Pipeline (calls to Google Gemini API)
• Visualization Modules (Plotting library for bar/treemap/timeline + D3.js graph)
• External Services: RSS/Atom sources, Google Gemini
• Deployment Environment: Local VM/Container or Streamlit Cloud
b. Relationships and interactions
1. User’s browser → HTTP request → Streamlit server
2. UI controller triggers Feed Fetcher → retrieves RSS/Atom feeds → forwards data to Cache
3. On “Save” action → Cache writes to Local Storage
4. Analytics page → reads from Local Storage → invokes NER pipeline → queries Google Gemini → returns entity/relation data
5. Visualization modules pull from Cache or Local Storage → render graphs (D3.js, matplotlib/plotly) → streamed back to UI
c. Architectural patterns or design principles
• Single-process, event-driven UI + business logic (Streamlit).
• Separation of concerns by page/controller within the single Python app.
• Layered architecture: Ingestion → Caching → Storage → Analysis → Visualization → Presentation.
• Use of external API for heavy tasks (NER), keeping core app lightweight.
• Stateless UI with stateful back-end components (cache, DB).
d. Relevant technologies, frameworks, or libraries
• Python 3.12, Streamlit
• feedparser or RSS libs
• SQLite (via Python’s sqlite3) or file-based cache
• Google Gemini APIs for NER
• D3.js for interactive graphs, plus Python plotting libs for other charts
• Devcontainer.json for uniform dev environment
• Streamlit Cloud for deployment
Detail without overthinking
• Logical components even if they run in one Python process.
• Every UI page’s dependencies.
• Role of caching vs. permanent storage.
• External NER service and RSS sources.