Milestones
List view
Clinical assurance and advanced privacy risk. Generalize intended-use and validation evidence across capabilities; add human-factors and shadow-mode safety, post-release corrective-action evidence, linkage and privacy-composition analysis, consent withdrawal, and advanced attack gates. No medical-device certification claims.
No due dateOne stable OpenMed platform contract. Introduce the deliberate breaking boundary for canonical schemas, evidence, errors, configuration, capabilities, policy receipts, lifecycle events, generated clients, migration tooling, and compatibility gates.
No due datePortable OpenMedKit and runtime parity. Align Python, Swift, Kotlin/Android, browser, desktop, and portable inference around signed offline model bundles, shared evidence and policy contracts, lifecycle safety, device qualification, and no silent cloud fallback. Release plan: #3604.
No due date•0/15 issues closedSecure exchange and recoverable service. Harden REST, GraphQL, gRPC, asynchronous jobs, webhooks, local Journey storage, model activation, upgrades, and air-gapped recovery with fail-closed privacy, security, resource, and release-evidence gates. Release plan: #3543. No new clinical capability or certification claims.
No due date•0/25 issues closedAssured agentic clinical workflows: permissioned local agents, bidirectional FHIR and OMOP operations, deterministic audit and recovery, human approval, deployment assurance, and independently governed workflow evaluation. Release plan: #3193. Entry and exit criteria, architecture, workstreams, dependencies, and release evidence are maintained there. Architecture: Extend `openmed/agent`, `openmed/interop`, and `openmed/eval`; do not create a parallel `openmed/agents` package. Target date is intentionally unset until the release epic's entry criteria pass; sequence is v3.1 -> v3.2 -> v3.3. Open scope must close or move to a named later milestone with rationale before release.
No due date•54/73 issues closedPrivate learning, synthetic data, and global languages: federated and differentially private training, secure aggregation, privacy-safe synthetic corpora, low-resource clinical languages, governance, and reproducible promotion evidence. Release plan: #3195. Entry and exit criteria, architecture, workstreams, dependencies, and release evidence are maintained there. Architecture: Extend `openmed/training/federated`, `openmed/training/synthetic`, `openmed/compliance`, `openmed/clinical`, and `openmed/eval`. Target date is intentionally unset until the release epic's entry criteria pass; sequence is v3.1 -> v3.2 -> v3.3. Open scope must close or move to a named later milestone with rationale before release.
No due date•8/92 issues closedMultimodal and ambient clinical intelligence: local document, image, ECG, waveform, streaming speech, and ambient workflows with fused provenance, privacy controls, clinical review, and measured edge safety gates. Release plan: #3194. Entry and exit criteria, architecture, workstreams, dependencies, and release evidence are maintained there. Architecture: Keep modality orchestration under `openmed/multimodal`; backend adapters remain in `mlx`, `coreml`, `onnx`, and `torch`. Target date is intentionally unset until the release epic's entry criteria pass; sequence is v3.1 -> v3.2 -> v3.3. Open scope must close or move to a named later milestone with rationale before release.
No due date•22/68 issues closedOn-device reach and ecosystem packaging. Portable runtimes, edge and desktop surfaces, document ingestion, service deployment, observability, and ecosystem adapters.
No due date•117/117 issues closed- No due date•69/70 issues closed
Guarded local clinical intelligence. Multilingual extraction, grounding, clinical timelines and relations, post-de-identification summarization, clinical NLI, SDOH, and higher-risk decision-support aids under explicit evidence and review gates.
No due date•295/297 issues closed- No due date•96/97 issues closed
Multilingual clinical extraction and structured privacy for production. Language and document routing, clinical extraction and interoperability foundations, structured de-identification, bounded processing, risk controls, drift, registry, cohort, and extension governance.
No due date•237/237 issues closed