Skip to main content

​
System Architecture

The Newsreader is composed of three modules: 
elata-vsm-system-4/
├── scraper/         # AI-powered news aggregation engine
├── web/             # Next.js frontend
└── shared-types/    # Zod schemas and shared TypeScript types

​
Scraper Pipeline

The scraper runs as a scheduled process via PM2. Each cycle executes the following pipeline:

​
Phase 1: Aggregation

  • NewsAPI queries for neuroscience, psychiatry, BCI, DeSci, and related keywords
  • Web scraper fetches from 400+ curated sources: journals, preprint servers, pharma outlets, neurotech companies, regulatory agencies, and community platforms

​
Phase 2: Analysis

  • GPT processor evaluates each article against the Elata mission and assigns a relevance score (0-1)
  • GPT validator filters out low-quality or irrelevant content
  • Tag assignment applies from a controlled vocabulary of 86 domain-specific tags across categories: neuroscience, hardware, AI/ML, pharmacology, biohacking, biomarkers, DeSci, and industry

​
Phase 3: Enrichment and Output

  • Content scraper extracts full text, word count, and reading time
  • Embedding generation via OpenAI for semantic search
  • Podcast generation produces AI-narrated audio summaries using ElevenLabs with two narrator personas (Nova and Dr. Renn)
  • Discord notifications alert the community to high-relevance articles
  • Results are written to current.json and served via the REST API

​
Data Model

Articles follow a Zod-validated schema: 
const ArticleSchema = z.object({
  id: z.string(),
  title: z.string(),
  url: z.string(),
  source: z.string(),
  description: z.string(),
  summary: z.string().optional(),
  tags: z.array(NewsTagSchema).default([]),
  relevanceScore: z.number(),
  author: z.string().optional(),
  publishedAt: z.string().optional(),
  scrapedAt: z.string(),
  sourceType: z.enum(["newsapi", "scrape", "twitter", "reddit", "community"]).optional(),
  embedding: z.array(z.number()).optional(),
  // ... additional fields
});
Summary metadata tracks aggregate statistics: 
const SummaryMetadataSchema = z.object({
  totalArticles: z.number(),
  dateRange: z.object({ from: z.string(), to: z.string() }),
  tagCounts: z.record(z.string(), z.number()),
  sourceCounts: z.record(z.string(), z.number()),
});

​
Tag Taxonomy

The 86 tags are organized into domains:
DomainTag countExamples
Neuroscience & Research15eeg, fmri, neuroimaging, synaptic-plasticity
Hardware & Devices16bci, neural-interfaces, openbci, muse, eeg-headsets
Computational & AI12precision-psychiatry, machine-learning, neural-decoding
Neuro-Pharmacology8neuropharmacology, clinical-trials, drug-discovery
Biohacking & Experimental12nootropics, psychedelics, neurofeedback, meditation
Biomarkers & Mechanisms10biomarkers, neuroinflammation, gut-brain-axis
DeSci & Web37desci, daos, tokenomics, on-chain-governance
Industry & Business6pharma, biotech, startups, regulatory

​
Running Locally

git clone https://github.com/Elata-Biosciences/elata-vsm-system-4.git
cd elata-vsm-system-4
See the individual module READMEs for setup: