[Health] Allow users to import lab results from unsupported providers into longitudinal lab history

Feature request

ChatGPT Health currently creates useful longitudinal views from connected providers, but many users receive labs from concierge practices, independent labs, or providers that are not supported as Health connections.

Users can upload those reports to ChatGPT, but extracted results cannot be saved into Health's persistent lab history. This means, for example, an uploaded HDL result cannot appear alongside HDL values synced from connected providers in the user's longitudinal chart.

Requested behavior

Allow a user to upload a lab report (PDF/image/CSV), have ChatGPT extract the observations, show the user a confirmation/reconciliation screen, and save approved observations into a clearly labeled User-imported source in Health.

Imported observations should retain:

• collection date/time
• test name and standardized mapping (e.g. LOINC when available)
• value and units
• reference range and abnormal flag
• source lab/provider
• fasting status when present
• provenance linking back to the uploaded source document
• an explicit marker that the result was user-supplied rather than provider-synced

Why this matters

A user's medical history is often fragmented across systems. Concierge practices and independent labs frequently use portals that are not available as Health connectors. Without manual import, Health cannot become a complete longitudinal record even when the user has the original lab report.

Preserving provenance solves the trust problem: user-imported data should be visibly distinct from provider-synced data while still participating in charts, trends, and analysis.

Related product gap

This is adjacent to the broader issue where uploaded medical documents and connected Health records live in separate knowledge stores. A supported manual-lab import path would make uploaded reports first-class longitudinal health data rather than one-off chat context.

No repository files, tests, or entry points are named in the issue. Start by locating the Health document-upload, observation extraction, reconciliation, and longitudinal-history entry points, then clarify the data model and persistence boundaries. Done means approved imported observations retain the listed metadata and provenance while remaining visibly distinct from provider-synced results.