feat(peat-ffi): ADR-074 schema alignment — shrink FFI types to proto-backed fields

[phornstein]

Summary

• ADR-074 establishes peat-schema protobuf definitions as the single source of truth for all message types. This PR eliminates schema drift in peat-ffi by shrinking FFI document types (NodeInfo, TrackInfo, CellInfo) to only contain fields that exist in their corresponding proto messages.
• Removes CommandInfo/CommandStatus, NodeStatus, CellStatus, MarkerInfo, TrackCategory — types and enums with no proto backing or that belong at the consumer level.
• Adds peat-schema dependency and From<proto> for FfiType bridges (From<Node> for NodeInfo, From<Track> for TrackInfo, From<CellState> for CellInfo, From<i32> for HealthStatus).
• Net -1,849 lines (200 added, 2,049 removed across 9 files).

Field changes

Type	Dropped	Renamed
NodeInfo	name, readiness, battery_percent, heart_rate, last_heartbeat	status→health (HealthStatus enum), hae→altitude
TrackInfo	cell_id, formation_id	hae→altitude
CellInfo	name, status, center_lat, center_lon, scenario_command	formation_id→cohort_id
CommandInfo	(entire type removed)	—

Breaking changes for consumers

• Consumers using MarkerInfo, TrackCategory, CommandInfo, or any dropped field must define their own types. See peat-atak-plugin/docs/SCHEMA-MIGRATION-NOVEL-TYPES.md for migration guidance.
• NodeStatus enum replaced by HealthStatus (Unspecified/Nominal/Degraded/Critical/Failed) matching node.proto HealthStatus.

Test plan

• cargo check -p peat-ffi (lib, tests, examples) — clean
• cargo test -p peat-ffi --lib — 84 tests pass
• Grep-clean: no references to NodeStatus, CommandInfo, CommandStatus, CellStatus, MarkerInfo, TrackCategory, battery_percent, heart_rate, last_heartbeat in peat-ffi/src/, examples/, tests/
• CI passes
• Consumer plugin (peat-atak-plugin) defines its own MarkerInfo/TrackCategory before pulling this dependency

🤖 Generated with Claude Code

[peat-bot]

Peat QA Review (SHA: ad64fe7)

Scope: ADR-074 schema alignment for peat-ffi — shrinks NodeInfo/TrackInfo/CellInfo to proto-backed field sets, removes MarkerInfo/CommandInfo/NodeStatus/CellStatus/TrackCategory, adds From<proto> bridges from peat-schema, and drops the corresponding JNI/UniFFI/Dart FFI symbols. Net −1,849 lines.

Assessment: Field-set alignment is defensible and the JSON codec paths (parse_node_json / serialize_node_json / parse_node_publish_json / serialize_nodes_get_json) still have solid round-trip coverage in the rewritten node_tests module. Removal of MarkerInfo / CommandInfo / TrackCategory / NodeStatus / dropped fields is a breaking change; the author has already listed this under Breaking changes for consumers in the PR body and marked the peat-atak-plugin migration as an unchecked pre-merge item — so consumer coordination is tracked, not a fresh finding. The hae→altitude Option-to-scalar collapse (also cep) matches peat-schema/proto/track.proto:68 (float altitude) and :71 (float cep_m) proto3 scalars, so semantic alignment with the proto is intentional under ADR-074 — not a regression to flag.

FIPS posture, Automerge / Iroh transport semantics, OTA safety, and BLE platform bindings are unaffected — this PR touches only FFI type-shape and JSON codecs.

Findings

[WARNING] From<peat_schema::*> for FfiType bridges are unused and untested — peat-ffi/src/lib.rs:2784 (From<Node> for NodeInfo), :2806 (From<Track> for TrackInfo), :2839 (From<CellState> for CellInfo). A repo-wide grep for NodeInfo::from, TrackInfo::from, and CellInfo::from returns zero call sites — the only invocation is HealthStatus::from(state.health) at :2792, internal to the From<Node> impl itself. The actual FFI data path is still parse_node_json / serialize_node_json (the ~600 lines of manual JSON codec ADR-074 §Consequences promised would collapse). Compile-time field-rename drift protection does fire (the impls reference proto field names by identifier), but semantic-mapping bugs (swapped lat/lon, wrong enum variant mapping, off-by-one on HealthStatus::try_from) cannot be caught by the compiler and would ship undetected because nothing exercises the bridges. Action: either add unit tests that construct a representative peat_schema::node::v1::Node / Track / CellState and assert the derived FFI value field-for-field, or wire the bridges into the actual data path (proto→FfiType) so cargo test covers them via the existing codec suites. Without one of those, the PR delivers ADR-074's letter (field sets align with protos) but not its spirit (the "single auditable conversion layer" isn't audited by CI).

[peat-bot]

Peat QA Review (SHA: c684f80)

Scope: Incremental review over the CI-fix push on the ADR-074 schema-alignment PR. Base branch state (ad64fe7) already reviewed. The new commit is c684f807 fix(ci): resolve schema-alignment CI failures — four small changes: stale node.name reference removed from publishNodeJni android_log (now logs node.node_type), p.name → p.id in dual_test_peer and iroh_test_peer println paths, JNI native-method-table count assertion updated from 30 → 27 to match the removal of getCommandsJni / getMarkersJni / publishMarkerJni, and cargo fmt fixups. Total delta: 4 files, +14 / −21.

Assessment: The CI fix is mechanical and consistent — the node.name field no longer exists on the shrunk NodeInfo, so the log-line update is correct; the JNI count arithmetic (30 − 3 removed methods = 27) tracks the diff exactly. cargo fmt and log-string edits don't touch protocol, transport, FFI ABI shape, or crypto surface. FIPS posture, Automerge / Iroh semantics, OTA safety, and BLE bindings remain unaffected by this push.

The prior QA review at ad64fe7 raised one [WARNING] finding: the From<peat_schema::*> for FfiType bridges (From<Node> for NodeInfo at peat-ffi/src/lib.rs:2784, From<Track> for TrackInfo at :2802, From<CellState> for CellInfo at :2835) are unused and untested — the actual FFI data path still runs through parse_node_json / serialize_node_json, so a semantic bug in the derived conversions would ship undetected. This push does not modify the bridge code, so per incremental-review rules the finding is considered known-to-author (the CI-fix scope narrows the intent of this push, and the author can address the coverage gap in a follow-up push or a subsequent PR under ADR-074's ongoing migration).

No findings.