feat(ui): node tapper, built from whatever fleet the server reports - #1825
Open
clonea1 wants to merge 3 commits into
Open
feat(ui): node tapper, built from whatever fleet the server reports#1825clonea1 wants to merge 3 commits into
clonea1 wants to merge 3 commits into
Conversation
Phase is the channel every published CSI technique leans on -- Doppler, direction of arrival, breathing extraction -- and there is no way to see what the phase channel on a given deployment actually contains. When a phase-based method fails, nothing distinguishes "the algorithm is wrong" from "the phase on this hardware is unusable". Adds phase_diag.rs (600 lines, 8 tests): an opt-in sink recording raw and sanitised per-frame phase to CSV for offline analysis, behind --phase-diagnostics <dir>. Off unless the flag is given. Single-antenna ESP32 CSI carries a packet-detection timing offset appearing as a phase slope across subcarrier index, and it is large -- on this hardware roughly 758 radians per ADC sample of jitter. That is not noise to be averaged away; it swamps the physical phase entirely, and the sanitised channel is what remains after detrending it. Comparing the two on real captures is the only way to judge whether a phase-based method has signal to work with on a given board. Co-Authored-By: claude-flow <ruv@ruv.net>
Ground truth recalled after the fact is soft. The only high-provenance label of an overnight run is the one reported AS IT HAPPENS -- everything recalled afterwards drifted by roughly half an hour, and matching a detection to a remembered time is post-hoc fitting: during an active stretch the detector bumps every ~1.5 min, so any estimate "matches" something. Adds three pages that can be driven one-handed from a phone while walking: mark.html stamp a labelled instant into the recording survey.html record a position observation at a known point walk.html capture a timed walk path and GET /api/v1/diag/mark/:label, which stamps the label into the phase diagnostics stream and returns the recording-relative timestamp so the client can show what it captured. Marks are written twice, to the server sink and to localStorage, because losing ground truth means re-walking everything. The endpoint reports plainly when diagnostics are not enabled rather than accepting a mark that silently goes nowhere. Co-Authored-By: claude-flow <ruv@ruv.net>
Adds the per-location tapper: stand near a node, tap its button, and the label is stamped into the recording as it happens rather than recalled afterwards. Built entirely from /api/v1/nodes at load. One button per node the server actually knows about, sorted by id and labelled "Node 1", "Node 2" and so on, so the page suits a fleet of three or thirty with no editing. Node identity is the only spatial reference: no floor plan, no coordinates, nothing tied to one building. Two generic places sit alongside the nodes. A fleet never covers every space, and the uncovered ones need ground truth arguably more than the rest, precisely because nothing there is measuring. An unreachable server or an empty roster is reported as such and leaves the generic places usable. It never guesses at node ids -- invented ones would attach marks to nodes that do not exist, which is worse than no mark. Occupancy is recorded by ROLE, never by person: adult1, adult2, child1, children-together, everyone. Ground truth outlives the run it was captured in, so it should not carry anyone's name. Marks are written twice, to the server sink and to localStorage, because losing ground truth means re-walking everything. Co-Authored-By: claude-flow <ruv@ruv.net>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Adds the per-location tapper: stand near a node, tap its button, and the label
is stamped into the recording as it happens rather than recalled afterwards.
Built entirely from /api/v1/nodes at load. One button per node the server
actually knows about, sorted by id and labelled "Node 1", "Node 2" and so on,
so the page suits a fleet of three or thirty with no editing. Node identity is
the only spatial reference: no floor plan, no coordinates, nothing tied to one
building.
Two generic places sit alongside the nodes. A fleet never covers every space,
and the uncovered ones need ground truth arguably more than the rest, precisely
because nothing there is measuring.
An unreachable server or an empty roster is reported as such and leaves the
generic places usable. It never guesses at node ids -- invented ones would
attach marks to nodes that do not exist, which is worse than no mark.
Occupancy is recorded by ROLE, never by person: adult1, adult2, child1,
children-together, everyone. Ground truth outlives the run it was captured in,
so it should not carry anyone's name.
Marks are written twice, to the server sink and to localStorage, because losing
ground truth means re-walking everything.
Rebased onto current
mainbefore opening: staged before today's seven merges, so it needed replaying to avoid reading as a revert of them. Clean rebase, no files deleted.