libs/phoebe/* is the workspace for the Phoebe domain: the science and
medicine of the brain and mind. See DOMAINS/phoebe/ for architecture,
features, and specifications, and TODO Phase 181 (TODOS/phase-181.md) for
the implementation roadmap.
Phoebe is a standalone domain that depends on existing assets via contracts (Iris BCI device layer, Aphrodite biosignals, Psyche affective computing, the Phase 178 autonomous-research substrate, Kalika research agents). It does not move or absorb their code.
Package map (planned — built incrementally per Phase 181)#
| Package | Owns | Status |
|---|---|---|
@phoebe/core |
Domain primitives, RDoC/HiTOP dimensional model, state machines | ✅ built |
@phoebe/neural-io |
Multimodal acquisition + LSL/XDF sub-ms time-sync | planned |
@phoebe/signal |
Real-time DSP (ICA/CSP/xDAWN/Riemannian) | planned |
@phoebe/decoding |
Decoder runtime, neural foundation models, cross-subject adaptation | planned |
@phoebe/neuromodulation |
Closed-loop aDBS/RNS/tFUS/TI, read–write, safety governor | planned |
@phoebe/connectomics |
FFN segmentation, CAVE proofreading, petascale graph serving | planned |
@phoebe/simulation |
NEST/NEURON/Arbor/Brian2 over SONATA, multi-scale co-sim | planned |
@phoebe/brain-models |
Brain foundation models + circuit/region digital twins | planned |
@phoebe/neuro-analysis |
SpikeInterface/Kilosort, LFADS/CEBRA, manifolds | planned |
@phoebe/assessment |
PHQ-9/GAD-7/PCL-5/C-SSRS + DSM-5-TR/ICD-11 coding | planned |
@phoebe/mbc |
Measurement-based + collaborative care | planned |
@phoebe/phenotyping |
Passive sensing, EMA, voice biomarkers, JITAI | planned |
@phoebe/comp-psychiatry |
RDoC/HiTOP models, biotypes, treatment matching | planned |
@phoebe/therapeutics |
CBT/DBT/ACT/MI content, anti-sycophancy, clinician copilot | planned |
@phoebe/hypothesis |
KG-grounded generate/debate/rank + novelty/feasibility | planned |
@phoebe/design |
BOED/DAD/ADO, Bayesian adaptive trials, SMART/JITAI | planned |
@phoebe/experiment-runtime |
Frame-precise authoring/runtime + closed-loop/robotic execution | planned |
@phoebe/subjects |
Panel connectors, data-quality firewall, eligibility matching | planned |
@phoebe/analysis |
BIDS-App/MNE pipelines, causal inference | planned |
@phoebe/reproducibility |
Auto-preregistration, multiverse, provenance ledger | planned |
@phoebe/knowledge-graph |
Biolink KG, ontology service, RDoC OWL + DSM↔ICD crosswalk | planned |
@phoebe/literature |
PubMed/OpenAlex/S2 ingestion, SciFact claim extraction | planned |
@phoebe/evidence-synthesis |
Living systematic reviews, GraphRAG + KGARevion | planned |
@phoebe/datasets |
FAIR catalog, NWB/BIDS/DICOM I/O | planned |
@phoebe/governance |
Neurorights/HIPAA/GDPR, defacing, federated + DP compute | planned |
@phoebe/regulatory |
FDA SaMD + PCCP, bias/drift testing, RCT evidence | planned |
@phoebe/safety |
988 crisis routing, age-gating, jurisdiction-aware HITL | planned |
@phoebe/ai |
Autonomous closed-loop neuroscientist + fail-closed AI-IRB | planned |
@phoebe/integration |
Cross-domain hub | planned |
@phoebe/contracts |
Zod contracts + cross-domain DTOs | planned |
Hard requirements (enforced across every package)#
- Fail-loud uncertainty — every inference/decode/prediction carries calibrated uncertainty; out-of-scope requests return an explicit refusal.
- Human-in-the-loop — no autonomous diagnosis, prescribing, or unsupervised therapy.
- Append-only audit — stimulation events, data access, and autonomous actions are immutably logged.
- Jurisdiction-aware compliance — feature availability gated by the user's jurisdiction (AI-therapy law; neurorights).
- Standards-native data — recordings are born NWB/BIDS/XDF-conformant with W3C-PROV provenance.
- Real-time safety — the decode→stimulate loop has bounded deadlines, a thermal/charge governor, watchdogs, and a fail-safe stop.
Local development#
Because the monorepo runs many worktrees, prefer direct tooling over Nx:
bash
cd libs/phoebe/core
npx tsc --noEmit # type check
npx vitest run # tests
npx tsc -p tsconfig.lib.json # build