How ali-ncibi/wassla-app works
Wassla ingests RSS articles, processes them through a multi-agent translation and quality pipeline, stores multilingual content and vectors, then serves a trilingual reader and semantic search with cited answers. The graph also preserves article, search, and stats API branches plus tracing and offline evaluation. Relationships among sampled pipeline modules are grounded in excerpts; frontend and some service connections rely on README documentation, since their source excerpts were not provided.
Main entry points: pipeline.py Article Graph · [...].ts API Proxy · repo.py Repository · __init__.py LLM Providers
Components 26
RSS Ingestion
- Ingest API
pipeline/app/routes/ingest.py - Feed Fetcher
pipeline/app/feeds.py - Ingest Graph
pipeline/app/graph/pipeline.py
Article Processing
- Article Graph
pipeline/app/graph/pipeline.py - Translation
pipeline/app/graph/nodes.py - Categorization
pipeline/app/graph/nodes.py - Summarization
pipeline/app/graph/nodes.py - Quality Evaluation
pipeline/app/graph/nodes.py - Embedding
pipeline/app/graph/nodes.py - Task Prompts
pipeline/app/graph/prompts.py - LLM Providers
pipeline/app/providers/__init__.py - Embedding Provider
pipeline/app/providers/embeddings.py
Storage and API
- Article API
pipeline/app/routes/articles.py - Search API
pipeline/app/routes/search.py - Metadata API
pipeline/app/routes/meta.py - RAG Search
pipeline/app/rag.py - Repository
pipeline/app/repo.py - PostgreSQL + pgvector
Reader and Search
- Reader Pages
web/app/pages/index.vue - Article Page
web/app/pages/article/[id].vue - Search Page
web/app/pages/search.vue - Stats Page
web/app/pages/stats.vue - API Proxy
web/server/api/[...].ts
Tracing and Evaluation
- Phoenix
- Offline Evaluations
pipeline/evals/run_evals.py
Other components
- Reader
Connections 39
- Reader → Reader Pages: browses
README.md - Reader → Search Page: searches
README.md - Reader Pages → API Proxy: requests data
- Article Page → API Proxy: requests article
- Search Page → API Proxy: submits query
- Stats Page → API Proxy: requests stats
- API Proxy → Article API: proxies requests
README.md - API Proxy → Search API: proxies requests
README.md - API Proxy → Metadata API: proxies requests
README.md - Reader → Ingest API: triggers ingest
- Ingest API → Ingest Graph: invokes
pipeline/app/routes/ingest.py - Ingest Graph → Feed Fetcher: fetches items
pipeline/app/graph/pipeline.py - Ingest Graph → Article Graph: dispatches articles
pipeline/app/graph/pipeline.py - Article Graph → Translation: runs first
pipeline/app/graph/pipeline.py - Article Graph → Categorization: runs next
pipeline/app/graph/pipeline.py - Article Graph → Summarization: runs next
pipeline/app/graph/pipeline.py - Article Graph → Quality Evaluation: runs next
pipeline/app/graph/pipeline.py - Article Graph → Embedding: runs next
pipeline/app/graph/pipeline.py - Translation → Task Prompts: builds messages
pipeline/app/graph/nodes.py - Categorization → Task Prompts: builds messages
pipeline/app/graph/nodes.py - Summarization → Task Prompts: builds messages
pipeline/app/graph/nodes.py - Quality Evaluation → Task Prompts: builds messages
pipeline/app/graph/nodes.py - Translation → LLM Providers: calls model
pipeline/app/graph/nodes.py - Categorization → LLM Providers: calls model
pipeline/app/graph/nodes.py - Summarization → LLM Providers: calls model
pipeline/app/graph/nodes.py - Quality Evaluation → LLM Providers: calls model
pipeline/app/graph/nodes.py - Embedding → Embedding Provider: creates vectors
pipeline/app/graph/nodes.py - Article Graph → Repository: persists results
- Repository → PostgreSQL + pgvector: reads and writes
pipeline/app/repo.py - Feed Fetcher → PostgreSQL + pgvector: reads sources
pipeline/app/feeds.py - Article API → Repository: reads articles
pipeline/app/routes/articles.py - Metadata API → Repository: reads metadata
pipeline/app/routes/meta.py - Search API → RAG Search: runs search
pipeline/app/routes/search.py - RAG Search → Repository: retrieves matches
pipeline/app/rag.py - RAG Search → Embedding Provider: embeds query
pipeline/app/rag.py - RAG Search → LLM Providers: generates answer
pipeline/app/rag.py - Offline Evaluations → Article Graph: replays articles
pipeline/evals/run_evals.py - Offline Evaluations → Phoenix: stores experiments
pipeline/evals/run_evals.py - Ingest Graph → Phoenix: traces runs
README.md