How probablyapigeon/ellie-s-voice-lab works
Bird Voice Lab is a local research notebook with a separate synthetic-agent sandbox. The principal workflow is to define an immutable study protocol, commit predictions before a trial, record the observation, then review descriptive comparisons and export evidence. The browser UI calls the local Python server, which dispatches requests to the lab logic and stores study events in SQLite. The README documents major analysis and privacy behavior; several internal implementations and routes were not fully sampled, so those links are omitted where their callers are not shown.
Main entry points: app.py Local HTTP server · lab.py Observation record · app.js Notebook interface
Components 14
Browser experience
- Notebook interface
web/app.js - Local HTTP server
app.py
Observation workflow
- Study protocol
lab.py - Trial predictions
lab.py - Observation record
lab.py - Study stopping
lab.py
Analysis and simulation
- Comparison controls
lab.py - Descriptive scoring
lab.py - Agent simulation
lab.py
Local records
- Hash-chained events
lab.py - SQLite store
lab.py - Evidence exports
app.py - Export verification
lab.py
Other components
- Researcher
Connections 16
- Researcher → Notebook interface: uses
README.md - Notebook interface → Local HTTP server: sends API requests
web/app.js - Local HTTP server → Notebook interface: serves interface
app.py - Local HTTP server → Study protocol: creates studies
app.py - Local HTTP server → Trial predictions: commits trials
app.py - Local HTTP server → Observation record: records observations
app.py - Local HTTP server → Agent simulation: runs simulation
app.py - Local HTTP server → Export verification: verifies exports
app.py - Study protocol → Hash-chained events: appends study
lab.py - Trial predictions → Comparison controls: locks predictions
README.md - Observation record → Descriptive scoring: scores completed trials
README.md - Observation record → Study stopping: applies end rules
README.md - Hash-chained events → SQLite store: writes events
lab.py - Local HTTP server → Evidence exports: exports JSON and CSV
app.py - Evidence exports → Export verification: checks receipt
README.md - Researcher → Agent simulation: configures sandbox
README.md