How anvithdommata/cerebellum works
Cerebellum is an investment-research platform: users ingest market and reference data, run predictions across five signal engines, inspect forecasts and weights, then evaluate matured forecasts to adapt engine weights. The README describes backfill, settings, and data-source capabilities, but the supplied tree contains only the server entry point and a partial dashboard; most backend paths are absent. Those documented subsystems are shown without invented source paths, and their internal wiring is limited to the documented workflow.
Main entry points: server.ts HTTP server · DashboardPanel.tsx Research dashboard
Components 22
Application entry
- HTTP server
server.ts - Research API
- Key settings
Data ingestion
- Ingestion scheduler
- Ingestion jobs
- Price sources
- Fundamental data
- News data
- Government data
- Social stream
Prediction
- Signal engines
- Weighted ensemble
- Forecasts
- Historical backtest
Reflection
- Outcome scoring
- Weight updates
- Mistake journal
User interface
- Research dashboard
DashboardPanel.tsx
Other components
- Research user
- Market and data APIs
- Gemini
- Local settings file
Connections 31
- Research user → HTTP server: requests
- HTTP server → Research API: routes API
- HTTP server → Key settings: mounts settings
- HTTP server → Ingestion scheduler: starts jobs
- Ingestion scheduler → Ingestion jobs: schedules
- Research user → Ingestion jobs: triggers imports
- Ingestion jobs → Price sources: fetches prices
- Ingestion jobs → Fundamental data: fetches fundamentals
- Ingestion jobs → News data: fetches news
- Ingestion jobs → Government data: fetches policy data
- Price sources → Market and data APIs: reads prices
- Fundamental data → Market and data APIs: reads fundamentals
- News data → Gemini: scores sentiment
- Research user → Research API: runs pipeline
- Research API → Signal engines: runs inference
- Signal engines → Weighted ensemble: emits scores
- Weighted ensemble → Forecasts: stores forecasts
- Research API → Historical backtest: starts backtest
- Historical backtest → Signal engines: simulates predictions
- Historical backtest → Outcome scoring: evaluates history
- Research API → Outcome scoring: runs reflection
- Outcome scoring → Forecasts: reads matured forecasts
- Outcome scoring → Weight updates: scores outcomes
- Weight updates → Weighted ensemble: updates weights
- Outcome scoring → Mistake journal: records large misses
- Research user → Research dashboard: views research
- Research dashboard → Research API: loads dashboard data
- Research dashboard → HTTP server: fetches candles
- Research user → Key settings: enters API keys
- Key settings → Local settings file: writes keys
- Signal engines → Social stream: reads sentiment stream