- content/edge/l3/edge-car-tracker.md, written from the evidence notes
- product image of the SIM7080G Cat-M module, cropped to 640px, with a caption
naming what it is and that it is a product image
- referenced from the edge overview
- original kept in sam/CAR_GPS_TRACKER so the project repo documents its own build
The /platforms/ page (408d6cd) reads platforms/*.yml and fails the build when
they are absent. Those three files were introduced on feat/phase-4-ingestion by
the ingestion commit 1e46ec0 and nothing else on the site needed them, so the
cherry-pick of 408d6cd left the tree unbuildable.
They are data, not ingestion code: each file is config plus public copy that the
site shows verbatim so a visitor can see exactly what the assistant reads. No
corpus/, no harvester, no Prefect flow and no Qdrant code is included.
Taken verbatim from 1e46ec0.
go test ./...: pass.
The note under the loopback box ran under the edge panel at 11.5 px. Shortened
so it ends inside the runtime panel at every render size. Re-measured after the
change: performance 100, accessibility 100.
(cherry picked from commit 4fb759cf22c3609542600009536613a04e461746)
- Diagrams carry a print style block, so they print as ink on white rather than
as dark panels; the client wall and the chips go white with a hairline.
- docs/phase-7-notes.md records what batch 1 shipped, how to add a mark or an
image, and the Lighthouse measurements taken after the batch (100/100/100/100,
CLS 0, mobile and desktop).
(cherry picked from commit ae9282e5f62b8c7ee149ebf96f8d3f00067a49e2)
Phase 7 (AUD-149), cheapest first.
- internal/brand: vendored Simple Icons path data (CC0) for 50 marks, looked up
by the technology label the content already uses. Inline SVG, 16 px,
currentColor, no request and no font. tools/gen_brand_icons.sh regenerates it.
- Six hand-drawn area icons replace the two-letter badges in the nav and on the
landing cards; the evidence index gets a seventh.
- TechChips renders a mark where one exists and plain text where one does not,
so a missing icon reads as a decision rather than a broken image.
- Client wall: 45 names as one typeface, one size, one grid of equal tiles, and
the reason for no logo collage written on the page. No mark invented.
- Two diagrams, authored in-repo: homelab topology (platform-network.md) and
the RAG pipeline (ai-rag-corpus.md), each referenced from the page it explains.
- Content images now carry intrinsic width/height, loading="lazy" and
decoding="async" at render time; a site test fails the build if an image
ships without them.
- Topbar is two rows at desktop: the area names plus marks no longer overlap the
search form (Lighthouse target-size was failing at 1440 px).
- Favicon, so the browser stops requesting a missing /favicon.ico.
Satisfies: logos on every area page, area icons in nav and cards, two diagrams
referenced from their pages, images sized and lazy, Lighthouse >= 95 held.
(cherry picked from commit a473e4738a3751418c2097df2bcf0568fb6d0208)
The nav form carried hx-get="/search" hx-target="#search-results", but
#search-results exists only on /search. On every other page htmx found no
target and silently discarded the response: a search box that looks like it
works and does nothing.
- SearchForm is now a plain GET with no htmx attributes (navigation and the
landing page), so it navigates to /search?q=… from any page and with
JavaScript disabled.
- SearchFormLive keeps the in-place swap on the search page, where the target
does exist. For HX-Request the server now returns the block's *contents*
(SearchResultsBody), so the fragment no longer nests a second element with
id="search-results" inside the live region.
- Empty query shows a prompt instead of an empty list. A zero-match query says
"Nothing found" and offers the six areas plus the evidence index.
- The result count is role="status" inside the existing aria-live container, so
a screen reader reads "4 results for …" rather than a run of titles.
Tests (internal/server, internal/site):
- the nav form carries no htmx attribute on nine pages, and no built page
carries an hx-target whose element is missing (walks all of dist/);
- search submitted from /, /edge/ and /platform/ lands on a page with results
with no HX-Request header, i.e. the JavaScript-disabled path;
- empty query, zero match and the announced count;
- a content change followed by a build is visible in dist/search.db. That is
the deploy path: deploy/deploy.sh runs `make dist` before it stages a
release, and search.Build removes and rewrites the index every build.
DoD: search from home, /edge and /platform lands on results; works with
JavaScript disabled; zero-match query shows a message; content change followed
by a deploy is reflected in search; no htmx request targets a missing element.
(cherry picked from commit a56c795768e1e9b10204aae04c3612d7a96775a7)
Point 7 of the page-wording review. The pages no longer show their data
structure as headings, and they no longer explain their own conventions.
- Level 1 is one screen: claim, status badge, three figures, the technology
row, and one sentence carrying one link. Measured in headless Chrome at
1440x900: 1404-1845 px tall before, exactly 900 px (no scroll) after, on
all six area pages.
- DocPage: no kicker line, no visible level label, no "Evidence for this
page" heading or instruction note, no "Drill-downs" list, no
"Status <word> - sentence" footer. The artefact table starts at level 2,
under the heading "Artefacts"; the history timeline is level 2 only.
- Landing page: "Start here" link text gone, "How to read this site" panel
gone, the footer's status-label explainer gone.
- Content, presentation only, no claim softened: the leading H1 that
duplicated the page title removed from all 41 documents; level 1 bodies
reduced to the constraint as ordinary prose plus the way in;
"**Status: ...**" prefixes removed from the artefact stubs; "## Drill-downs"
and "## Period and confidence" removed from the level 2 pages, with the
cross-area links moved into the prose.
- The six level 1 claims are now sentences with a subject, and use commas
instead of middots. Same facts, no new ones.
Not in this commit: the 35 level 2 and 3 claims still use the compact CV
style. That is a voice decision for Sam, not a wording tidy-up.
make validate: 41 documents, 0 violations. go test ./...: pass.
(cherry picked from commit 03009f80fc7815ad7719412e125df9297a6e653f)
127.0.0.1:8096 is unreachable from .35, which is where Caddy runs. Every other
service on .13 is published this way (8095, 8092, 3909, 4200) and proxied by the
same Caddy. Port is LAN-only; Caddy remains the sole public listener.
Once the port is published on 192.168.20.13 the host no longer listens on
127.0.0.1:8096, so wait_healthy(), do_status() and the rollback path would all
get connection-refused and every deploy would roll back. They now probe
$PUBLISH_IP (RESUME_PUBLISH_IP, default 192.168.20.13). assert_shape() asserts
the published host IP is that address instead of loopback; the mem_limit ==
memswap_limit checks are unchanged. check_port stays address-agnostic on
`sport = :8096`.
Records the amendment in reference/stack.md (ports table + the bind rule) and
docs/phase-3-deploy.md (§3.2 bullet, flow diagram, DoD line), citing Sam's
decision of 2026-10-26, interaction 54aeb1a2. AGENTS.md is Sam's to change and
is untouched.
- docs/phase-4-corpus.md rewritten: Gitea (all repos), vault via Gitea, Graphify
graph, curated content as tier 1. Platform guides are config and public copy.
- docs/phase-5-ai-overlay.md: citations to originals, tier preference, live
source readout.
- Board: AUD-144 (index content/ only) retired; AUD-147 replaces it, blocked by
AUD-141 and AUD-142. AUD-145 relinked to AUD-143 and AUD-147.
Chunks, deterministic ids, payload filter, FTS5 keyword index, incremental mode,
freshness audit, Qdrant container shape, measured memory. Deploying Qdrant to
the runtime host and the Prefect deployment are still outstanding and belong to
the runtime host, not this repository.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Satisfies docs/phase-4-corpus.md. content/ stays the single source of truth:
one pass over it writes both indexes, so the static pages and the retrieval
corpus cannot drift.
- corpus/: stdlib-only pipeline. validate -> chunk -> enrich -> embed -> upsert.
Chunking is block-based, so a boundary cannot fall inside a fenced block, a
table or front-matter (verified structurally, not by inspection).
- Deterministic ids: sha256(source_path + chunk_index + content_hash), rendered
as a UUID. Re-running updates points; re-chunking deletes the old ones by
source filter, so duplicates and orphans are both impossible.
- Incremental by content hash against a SQLite manifest. A provider, dimension
or chunker change is refused instead of mixed: it needs a rebuild.
- Payloads per section 4.2 plus provenance, validated before any vector is
written; a missing field fails the run and names the document and chunk.
- SQLite FTS5 over chunks (the keyword half of hybrid retrieval, decision D10)
with area/confidence filters and a `terms` column so front-matter tech and
tags are findable.
- Freshness audit: stale documents, due soon, index drift, contract drift;
optional Apprise alert; exit code for Prefect.
- Embeddings behind one interface: omniroute (OpenAI-compatible, retries, key
from the environment only) and local-hashing (offline placeholder). The
provider decision is still Sam's; the pipeline does not depend on it.
- deploy/qdrant.yml: mem_limit = memswap_limit = 400m, loopback only.
- docs/phase-4-notes.md: operator manual, DoD evidence, measured memory.
- 76 unit tests, no container and no network.
Measured, Qdrant 1.20.0 under a 400 m cap at 1024 dimensions: 500 / 5,000 /
20,000 chunks settle at 128 / 138 / 151 MiB, peaking at 149 / 305 / 292 MiB
during ingest. Qdrant's default settings are OOM-killed between 5,000 and
20,000, so the collection is created with HNSW and payloads on disk; recorded in
reference/stack.md with the ceiling and the escalation order.
Note: the `make corpus*` targets were swept into the Phase 3 deploy commit
24590cc by concurrent work in this tree; the Makefile content is correct, and
this commit contains the pipeline they call.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Two failures from the first real deploy, both repo-side:
1. The container logged "exec /usr/bin/resume-server: no such file or
directory" and restarted until the health gate failed. The build host has
cgo enabled, so the binary carried the nix glibc interpreter, and the
distroless/static image has no dynamic loader. Every SQLite path already
uses pure-Go modernc.org/sqlite, so build with CGO_ENABLED=0 and ship a
static binary.
2. Compose warned about orphan containers ai-resume-backend-1 and
ai-resume-db-1: /home/sam/deployment/ai-resume on .13 already owns the
`ai-resume` compose project (on the ai-mesh network). Sharing the name makes
each stack visible as orphans to the other, so a `compose down
--remove-orphans` in either could stop the other. Run the site as project
`ai-resume-site`; the image tag stays ai-resume:local and the port binding,
mem_limit and memswap_limit are untouched.
The first deploy failed before the health gate:
Network ai-resume_default Error Error response from daemon:
all predefined address pools have been fully subnetted
.13's docker daemon has 31 bridge networks already, which exhausts its
default-address-pools, so any new implicit compose network fails to create.
Pinning 10.89.0.0/28 (no route, bridge or LAN overlap on .13) makes this
project independent of that host condition without touching another project,
container or the daemon config. mem_limit, memswap_limit and the 127.0.0.1
port binding are unchanged.
The release directory under /home/sam/deployment/resume is the rollback unit;
deploy.sh stages releases/<timestamp>-<sha>/, flips `current` atomically,
builds the image from that directory and gates on /healthz, restoring the
previous release on any failure.
The rehearsal recipe in the header is kept as comments only; the pasted spec
lost the comment prefixes on the multi-line printf, which would have made the
script exit 1 on startup. It now uses \n escapes on one commented line.
DoD satisfied (evidence in AUD-142): A1 with no AI services, one binary,
deterministic templ generate, 6 L1 + 6 L2 + artefact pages, FTS5 search,
Lighthouse 100/100 mobile and desktop, <100ms first byte, stage panel driven
by route or stream, print legible.
One Go binary serves a pre-rendered site. No Node, no bundler, no node_modules:
htmx and the SSE extension are vendored files, Markdown and code highlighting
run at build time (goldmark + chroma), and the server only writes files.
Implements docs/phase-2-static-site.md:
2.1 service skeleton cmd/build (content -> dist), cmd/server (dist -> HTTP),
env-configured, 127.0.0.1:8096, /healthz with separate
static / ai / search readiness, --health-check probe,
deploy/compose.yml + deploy/Dockerfile at 120m
2.2 rendering templ components (Layout, Nav, PillarCard, Stage,
EvidenceTable, CodeBlock, Timeline, Footer);
`make generate` is a build step and is byte-stable
2.3 navigation six-area rail, <details> disclosure under 900px,
breadcrumbs, aria-current, L1/L2/L3 with artefact pages
2.4 stage panel internal/stage owns the contract; RouteDriver (fragment
swap) and SSEDriver (event stream) differ only in the
Binding, never in layout. fragment ids live in the
contract so route files, links and streamed frames agree
2.5 search SQLite FTS5 over titles and body, rebuilt by the build;
one handler returns a page without HTMX, a fragment with
it. Visitors' FTS operators are never honoured
2.6 quality floor print stylesheet, JS-off navigation, focus states,
semantic landmarks, dark default; security headers
Definition of Done evidence (see issue AUD-142):
A1 all pages 200 in a network namespace with no external route
one binary serves the site; make dist writes 45 pages, 86 fragments, 40 docs
templ generate is deterministic (md5 identical across runs)
6 L1 + 6 L2 pages, artefact pages reachable in <= 3 clicks
FTS5 returns sensible hits for zigbee / signage / routing (tested)
Lighthouse 100 performance and 100 accessibility, mobile and desktop
first byte well inside 100ms over loopback (test)
panel renders from a route driver and re-points at the stream untouched
print legible: /edge/ renders to a 3-page PDF with the print stylesheet
Also fixes found while finishing the build:
- reject "." / ".." path segments before ServeMux redirects them into a file
outside the asset directory (/static/../build.json now 404, was 307->200)
- make test is the real gate: go test ./... plus the validator self-test
'The spine' records 50+ containers across 3 hosts and a 4-host estate; the
pages asserted containers 'across four hosts'. Reworded so the container
count and the host count are each stated without implying a distribution
the sources do not give. No metric value changed.
DoD: schema valid (make validate clean), metrics still trace to the
approved chip set.
Satisfies the Phase 1 Definition of Done:
- content/ validates cleanly against the frozen schema. `make validate`
reports OK: 40 document(s) validated, 0 violations.
- Six areas exist, each with an L1 entry: ai-overview, platform-overview,
edge-overview, products-overview, history-overview, about-overview.
- All six areas have an L2 entry: ai-systems, platform-estate, edge-fleet,
products-business, history-engagements, about-narrative.
- 28 L3 artefact stubs under content/<area>/l3/. Every artefacts[].url in
every document resolves to one of them; the validator proves there are no
dangling references (and a deliberately broken copy exits 1).
- Every metric value traces to a source line in the evidence notes or the
approved metric set in reference/content-map.md (per-metric citations are
recorded in the AUD-131 completion comment).
- Every document carries confidence and last_verified=2026-10-26; the
validator rejects a missing or non-canonical value.
- docs/content-redaction.md records the redaction pass: identifiers, third
party personal data, credentials and infrastructure detail excluded, with
reasons, plus the two numbers that are reported rather than smoothed.
Honesty positions held: 3 businesses active (12 adoptions over time),
Where-Woof deployed and pre-market, failures kept in (edge-failures), no
promotion of in-progress work to delivered.
No site code, no templates, no CSS, no rendering (Phase 2).
Implements AGENTS.md section 7 front-matter schema as a single-file,
standard-library-only Python validator; no third-party YAML dependency
and no framework. Adds the `make validate` gate, a 20-case fixture set
and docs/content-schema.md.
DoD satisfied:
- make validate runs and exits 0 on an empty/clean content/ tree
(missing dir and empty dir both exit 0; fixture ok-empty covers it).
- Every violation class has a fixture under test/fixtures/fail-*; each
exits non-zero with one line per violation (file:line: field: rule).
`make test` runs all 20 cases: 20/20 pass.
- Dangling artefact reference is fatal (fail-dangling-artefact;
artefacts[].url must resolve to an existing in-repo L3 stub or asset).
- Unknown front-matter field is fatal (fail-unknown-field), including
unknown keys nested inside artefacts[]/metrics[].
- docs/content-schema.md committed; states the frozen schema, links to
AGENTS.md section 7, and records that any later schema change is a
breaking change requiring Sam's approval.
- Scope respected: no content authored, no rendering, no site code.
Files: Makefile, tools/validate_content.py, test/fixtures/**,
docs/content-schema.md, .gitignore (__pycache__).
