Fixes an inconsistency an agent hit: it could POST/PATCH/DELETE with its token
but GET (readback) required a session and 302-redirected to the browser login.
- GET /api/artifacts and GET /api/artifacts/{id} now accept a bearer token OR
a session, owner-scoped (a token reads only its owner's artifacts). Agents
can read back what they posted with the same token.
- Unauthenticated API requests return a JSON 403, never a browser-login
redirect (resolve_request_user helper; dropped @login_required on the JSON
read views).
- GET /api/skill and GET /api/capabilities are now public (no user data), so
an agent can fetch the contract + posting instructions before it has a key.
- Docs (api.md, CLAUDE.md) updated; tests for bearer readback, owner-scoping,
and the JSON-403-not-redirect behavior. Suite: 132 passing; verified live.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Removes the "Daybook" name from everything user- and agent-facing, since no
agents are configured yet and this is the cheap moment to do it.
- UI brand, docs, README, CLAUDE.md -> AgentFeed.
- Ingest env vars DAYBOOK_URL/TOKEN/SOURCE/DATA_DIR/BIND_ADDR/HOST_DATA_DIR/
KEY_PROVISIONING_API -> AGENTFEED_* (token *values* unchanged).
- Skill: skills/daybook-posting -> skills/agentfeed-posting, daybook_post.py
-> agentfeed_post.py (+ scripts/ wrapper and its test).
- static/daybook.css -> static/agentfeed.css; localStorage daybook.* ->
agentfeed.*; templatetag daybook_extras -> agentfeed_extras; multipart
boundaries; json/csv viewer element ids; daybook_exception_handler ->
agentfeed_exception_handler; compose service/container -> agentfeed.
- Key-provisioning API now OFF by default (independent of DEBUG); enable with
AGENTFEED_KEY_PROVISIONING_API=1 on test installs only.
- Posting helper now points at Settings -> API Keys.
Deferred on purpose (a later, separate step): the Django package
`daybook_project`, the `/opt/daybook` install path, and the Forgejo repo name.
The sibling AGENT_DAYBOOK__LLM_TEST name is intentionally untouched.
Full suite: 129 passing; live server verified serving under the new name.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Add a top-level README (repo had none): overview, local quick start,
posting via API keys / the skill endpoint, a docs index, and the admin
split.
- deployment.md: new "Django Admin (/admin/)" section covering how to log in,
the is_superuser vs is_staff distinction, how to grant /admin/ access, and
the raw-edit-bypasses-app-logic caution.
- Refresh constraint wording to per-user multiple API keys.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Replace the gear/username button with a round initials avatar (per-user
deterministic color via a new daybook_extras template filter); dropdown
now shows the username and keeps Settings/About/theme/logout. Sets up an
optional profile photo later.
- Regroup the settings nav into "You" (Account, API Keys, Storage, Demo
Data) and a superuser-only "Admin" (Users), About separate — the split
now reflects who each setting affects, not a flat list.
- Register ApiKey (key read-only) and TagColor in Django admin so /admin/
is complete as the operator break-glass tool, distinct from the themed
Settings product surface.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Replaces the single per-user DRF token with named, individually-revocable
API keys, so you can run one key per agent and cut off a leaked one without
disrupting the others.
- New ApiKey model (user, name, dbk_-prefixed key, created_at, last_used_at).
Data migration copies every existing authtoken 1:1 into a "default" named
key with the same value, so no client (incl. the sibling test project's
hardcoded test-user key) needs reconfiguring.
- Auth (resolve_token_user) resolves against ApiKey and stamps last_used_at,
throttled to one write/minute.
- Settings -> API Keys: table with name/key/created/last-used + per-row
revoke and a create form; keys viewable on demand. Settings -> Account now
links here; Settings -> Users issues a named initial key.
- Gated provisioning API for test installations: GET/POST /api/keys and
DELETE /api/keys/{id}, controlled by DAYBOOK_KEY_PROVISIONING_API (unset ->
follows DEBUG, so off in deployment). Advertised via key_provisioning_api
in /api/capabilities; 403 when disabled.
- Docs (api/deployment/quickstart/posting-helper), SKILL.md, and env.example
updated from "token / Settings -> Account" to "key / Settings -> API Keys".
- Tests: new API + UI suites for keys; existing token-based tests moved to
ApiKey. Full suite 129 passing.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Fixes the deploy blocker found deploying v0.1.0 to cloud-vm-1: the vendored
gridjs.umd.js ended with a //# sourceMappingURL=gridjs.umd.js.map trailer,
but that .map isn't in the repo. The strict CompressedManifestStaticFilesStorage
backend follows the reference and raises MissingFileError, aborting the
entrypoint's collectstatic so the container never served. Stripped the
debug-only trailer (no runtime effect); collectstatic now succeeds from a
pristine checkout (169 copied, 481 post-processed).
Also rolls up this release's already-committed work: GET /api/skill and the
location-agnostic posting-skill rewrite.
- Bump __version__ to 0.1.1.
- Runbook: note that a post-build crash-loop is most likely a collectstatic
MissingFileError (check docker logs), and how to clear on-server tracked
drift before checking out a tag.
- CHANGELOG: 0.1.1 entry.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Serves skills/daybook-posting/SKILL.md so users/agents can fetch the
posting instructions from the running instance instead of hunting for the
file. The server rewrites the DAYBOOK_URL config line to the request's own
origin, so the text handed to an agent already points at the right server;
the recipient only supplies their token.
- Same auth as /api/capabilities (session or bearer token); no user data.
- Raw text/markdown by default, ?format=json returns {version, url, skill}.
- Advertised via skill_url + a /api/skill methods entry in capabilities.
- Ships in the Docker image (COPY . /app, skills/ not dockerignored).
- Tests for auth, origin rewrite, JSON mode, session auth, capabilities ad.
- docs/api.md + CHANGELOG updated.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The skill was authored in this dev workspace, where the Daybook server,
the skill, and the posting agent all share one directory — so it read as
if the agent always runs from a Daybook source checkout on localhost.
- State up front that you post to a running instance that may be anywhere,
need no source checkout, and are tied to it only via DAYBOOK_URL/TOKEN.
- Drop "from the project root"; reference the bundled helper by absolute
path (DAYBOOK_POST=.../this-skill/scripts/daybook_post.py).
- Reframe DAYBOOK_URL as set-this-to-your-instance; label 127.0.0.1:8001
as the local dev fallback only.
- Genericize the hardcoded localhost URL in the image-embed example.
- Add a raw curl fallback for environments without the Python helper.
- Keep frontmatter to name+description (the Skills-API-portable subset).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- CHANGELOG.md started (Keep a Changelog style), with the v0.1.0 release
notes; versions track the git tags and __version__.
- Runbook §7 now deploys a specific tag (git fetch --tags && checkout
vX.Y.Z) and rolls back to the previous tag, with a note that rollback
reverts code not data (restore the data backup if a migration ran).
(The deploy kickoff brief lives in planning/, which is gitignored — it's a
local hand-to-the-agent note, not shipped in the repo.)
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Single source of truth: __version__ in daybook_project/__init__.py, starting
at 0.1.0. Surfaced in two places from that constant:
- Settings -> About page (logo, name, version, description), in the settings
nav and linked from the gear menu.
- GET /api/capabilities now includes "version", so the deploy agent / tester
can confirm which version is actually running.
Release flow: bump __version__, commit, `git tag vX.Y.Z`, push tag, deploy.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Real bug: the container healthcheck (and the docs' verify step) hit
/api/capabilities, which returns 403 unauthenticated; curl --fail treats
403 as an error, so a healthy container would report "unhealthy" and the
verify step would look like a failure. Healthcheck now hits /accounts/login/
(200, no auth). Verify commands now print the status code and note not to
use curl -f on the 403.
Also reframed cloud_agent_deploy.md's intro for an Opus-class deploy agent:
it's a guide + Daybook-specific gotchas, not a rigid script — points at the
authoritative sources (repo incl. the git-SSH :2222 note, deploy/ files,
requirements.txt deps auto-installed, entrypoint runs migrate+collectstatic,
docs/deployment.md) and trusts the agent's judgment. Dropped the now-
redundant Source Of Truth section.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Final regression sign-off before deployment. Since #018 the only changes
are cosmetic (table/body fonts), so no new API surface — this round is a
broad regression sweep: full CRUD across all kinds, auth boundaries,
confirmation that every closed issue (#003/#004/#006/#008/#010/#014/#015/
#016) stays fixed, and a sharing re-check. A clean pass is the deploy gate.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
report used Georgia (serif) for body prose, which looked dated/out of place
(it's the default for markdown and html_fragment, so most text artifacts).
Switched it to the system sans stack and pinned that same stack explicitly
in tasks too, so all three document themes (base/report/tasks) share one
font family rather than matching only by inheritance. Headings, table
cells (monospace), and table headers (Arial) are unchanged.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Table cell data (td) now uses a fixed-width stack (Consolas, SF Mono, Menlo,
...) at 0.92em so columns — especially numbers — align and scan cleanly;
headers (th) use Arial/sans for clear contrast against the mono data.
Applies to markdown and html_fragment tables via the shared base.css.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Covers the #014-#016 fixes (sanitization+CSP, nested-JSON 400, ragged-CSV
normalization) and flags the sanitizer as the top round-4 adversarial
target (mutation XSS, attribute/scheme bypasses, CSP presence). Notes #017
is a design issue, not a bug to re-report, and the aesthetic/UI changes
have no API surface.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The document themes (report/tasks/base) now share the unified off-white
background, so they differ in typography/width, not background color.
Updated the --theme description accordingly (was still claiming report has a
distinct "warm background").
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
When a single source is selected the row's source column is hidden, so the
sidebar now renders 112px (7rem) narrower and the reclaimed space goes to
the preview pane — instead of the sidebar staying the same width and just
widening the title. One stored "base" width drives both modes (wide =
stored, source-selected = stored - 112, floored at 150px); dragging the
splitter in either mode adjusts that base directly, so the width stays
user-settable in whichever mode and the modes track each other by the fixed
offset. Avoids having to re-resize every time you switch modes.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Design issue capturing the request to let users customize document-artifact
colors and switch between theme variants, instead of everyone getting one
forced theme. Notes the two separable pieces (structure/skin split via CSS
variables as a prerequisite — the themes currently hardcode colors — and a
switching mechanism), the preset-vs-raw-CSS security tradeoff, dark-mode
interaction, and deck as the concrete trigger. References the existing
planning docs on theme variants/palettes.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
markdown/html_fragment (report, tasks, base themes) and the json/csv viewers
each used their own near-white background (cream #fffaf0, light-blue
#f7fbff, or plain white). They now all use #faf9f6 — the same off-white the
preview area uses for image/binary artifacts — so browsing across kinds is
visually consistent.
Left as-is on purpose: deck (intentional dark slide background; switching it
would require reworking its light text colors), html_full (served verbatim,
poster owns styling), and pdf (the browser's own viewer chrome). Existing
json/csv artifacts keep their baked-in #fffaf0, which is visually
indistinguishable from #faf9f6; only new ones get the exact value.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Bumps the brand logo a notch; still fits within the existing top-bar
height (the filter toolbar remains the tallest element, so no layout
growth).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Regenerated logo/favicon/apple-touch from daybook_icon_2.png, a genuinely
transparent PNG (vs the first icon, keyed out of a JPEG checkerboard). The
new source has a subtle neon glow rim: the top-bar logo uses a tight
alpha>=140 crop so it stays crisp on the light chrome, while the square
favicon/apple-touch keep the thin glow (looks good on dark tab bars/tiles).
Placement unchanged — logo sits in the top-left brand corner next to
"AgentFeed".
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Generated a transparent PNG set from the source icon (a JPEG with a baked-in
checkerboard) by keying out the background on saturation — the arrow is
strongly saturated (0.47-0.68) while the grey/white background is 0.00, so
the cutout is clean. Produced static/logo.png (top-bar, natural aspect),
static/favicon.png (32px), and static/apple-touch-icon.png (180px).
Wired favicon + apple-touch links into all three page heads (feed, settings,
login) and placed the logo next to the "AgentFeed" title in the top bar.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Establishes a clearer background hierarchy across the app chrome:
- Top bar: its own slightly darker grey (--topbar-bg).
- Both sidebar headers (pane-bars) and the content control bar: grey
(--pane-bar-bg now matches the sidebar grey instead of white).
- Artifact sidebar: grey background like the source sidebar; the rows
themselves stay white so they read as cards floating on grey.
- Preview/content area: a light off-white (--content-bg), distinct from
both the grey chrome and the white rows.
New --topbar-bg and --content-bg vars added to both light and dark
palettes. Also fixes a pre-existing bug from the dark-mode commit where
--scrollbar-thumb and --focus-ring had become self-referential (a
too-broad replace_all had clobbered their own definitions), so scrollbar
and focus-ring colors resolve correctly again. Verified both modes render
coherently.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The filter form sat at the left edge of its (wide, 1fr) grid cell, jammed
up against the brand. justify-content:center groups it in the middle of the
top bar, above the content area and roughly screen-centered.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
When a single source is selected, every row in the middle sidebar repeats
that same source, so the source column is pure waste — the pane header
already names the source. The column now only shows under "All sources"
(where it's needed to tell rows apart); selecting a source hides it and
gives the width back to the artifact title. Also drops the now-dangling
row-main border-right in that state.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
#015 — deeply nested JSON raised RecursionError (a RuntimeError, not
ValueError) and became an unhandled 500 on two paths:
- json-kind content validation now catches RecursionError -> 400.
- Added a custom DRF EXCEPTION_HANDLER
(artifacts/exception_handlers.py, wired via REST_FRAMEWORK) mapping
otherwise-unhandled RecursionError -> 400, covering the parser-level
crash on deeply nested request bodies (metadata etc.) too.
#016 — wrap_csv_viewer handed Grid.js rows without reconciling width to the
header, so over-long rows silently dropped cells. Now normalizes every row
to the max width: long rows widen the columns (auto-named "Column N"),
short rows pad with "". No cell is dropped.
Both verified live; 4 regression tests added. 106 tests pass.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Records the decision to keep html_full fully flexible: no sanitization, no
CSP, and no HTML/CSS validation — malformed markup is kept as-is. The
themed kinds (markdown/html_fragment/deck) remain sanitized, which also
normalizes them to well-formed HTML.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
markdown/deck/html_fragment rendered author-influenced content verbatim,
so <script>/<img onerror>/etc. executed on the Daybook origin — self-XSS
against the owner's session, and (post-sharing) unauthenticated via /s/.
Two layers, both in artifacts/views.py:
- Create-time allowlist sanitization via nh3 (maintained ammonia/Rust
sanitizer, not deprecated bleach). Keeps formatting/tables/links/images
(incl. relative /assets URLs) and codehilite markup; drops script, on*
handlers, style, unknown tags, and javascript:/data: URLs. html_full is
deliberately NOT sanitized (documented verbatim escape hatch).
- Serve-time CSP (script-src 'none') for the sanitized kinds, on both
/a/{id} and the unauthenticated /s/{token} route. Being a response
header, it also neutralizes scripts in artifacts stored before this fix.
Not applied to html_full or json/csv (json/csv data is already escaped;
their viewers need their own JS).
Verified live across all kinds; 7 regression tests added. 102 tests pass.
Not addressed (noted in the issue): CSV formula injection,
CSRF_COOKIE_HTTPONLY (moot with XSS closed), nonce CSP for json/csv/full.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Moves the 11 Fixed issues (#001-006, #008-012) into issues/_archive_/,
which is gitignored — they stay on local disk and in git history but drop
out of the committed tree, keeping the issues/ dir focused on what's open.
The sibling test project reads this dir by local path, so it still sees
the archived files. Adds the three new open issues found in round-3
testing: #014 (stored XSS in markdown/deck/html_fragment — High), #015
(RecursionError -> 500 on deeply nested JSON), #016 (ragged CSV rows lose
data in the rendered table). Documents the archive convention in the README.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Covers everything shipped since the #007 handoff: the json/csv kinds,
.xlsx conversion, public sharing, syntax highlighting/table styling, and
the #008-#012 fixes. Prioritizes round-3 probe suggestions by risk, with
sharing (the new unauthenticated surface) and the #010 quota race (no
automated coverage) called out as the highest-value targets.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
The UI stylesheet already centralized its palette in a :root block, so this
mostly meant (1) migrating the ~30 colors still hardcoded in rules into
variables so they flip in one place, and (2) adding a [data-theme="dark"]
override palette. Artifact rendering is untouched — those use the separate
static/themes/*.css and the preview backdrop stays white by design, since
artifacts (charts, reports) are authored light.
- Toggle lives in the gear-menu dropdown (Dark mode / Light mode), persists
to localStorage.
- A tiny inline script in each page <head> (feed, settings, login) applies
the stored theme before first paint, so dark-mode users never see a white
flash on load.
- color-scheme: dark on the dark palette makes native form controls
(inputs, buttons) render dark too.
- Semantic accents (kept=green, shared=blue, today, danger) got
dark-appropriate variants; status/data-viz colors (usage bar, tag chips)
and the white artifact backdrop are intentionally left as-is.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Same pass as the SKILL.md review. Fixed factually-wrong content and filled
gaps against the current code.
docs/api.md:
- Capabilities example response was stale: artifact_kinds missing json/csv,
raw_supported_for missing json/csv. Corrected to match live output, with
a note that the endpoint is the authoritative source.
- "Office document conversion is not implemented" is now false - replaced
with the accurate scope (.xlsx -> csv only; .docx/.pptx/.xls not).
- Source Fields overview table omitted source_prompt entirely; added it,
plus a note that source_name/source_id/source_prompt must be strings
(400, not coerced) and metadata must be an object.
- Kinds table: noted tables now render styled in markdown/html_fragment,
and made explicit that html_full is served verbatim with no styling.
- MVP constraints: noted the public-share-link exception to "feeds private
per user" and clarified PATCH edits metadata only.
docs/llm_posting_helper.md:
- Added .xlsx upload, --upload-image image embedding, a Source & Theme
Flags table (all six source-*/theme flags were undocumented here), and a
note that --meta/--meta-json merge and are distinct from tags.
- Added a scope note that the helper only creates - update/delete need the
API, sharing is feed-UI-only.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Found six real flags with zero mention anywhere in the doc:
--source-type, --source-name, --source-id, --source-prompt, --meta-json,
--dry-run. Added a new "Source & Provenance Fields" section for the
source-* flags and covered the rest inline/in Rules.
Fixed:
- html_full's cheat-sheet entry overstated the requirement as needing
doctype/html/head/body; the actual check only requires <html>.
- The "Images Inside Content" section had grown to cover four unrelated
topics (images, CSV/JSON, xlsx, code-as-markdown) under a header
describing only the first one. Split into properly-titled sections:
Images Inside Content, CSV/JSON/Spreadsheets, Other Code.
- An unclosed parenthesis in the xlsx limitations paragraph, rewritten as
a bulleted list for clarity while fixing it.
- Nothing anywhere mentioned that this script only creates artifacts -
update/delete need the raw API, sharing is feed-UI-only with no API
path at all. Added an explicit scope note.
- --source-name's actual default (the resolved source, which can come
from DAYBOOK_SOURCE, not just --source) was previously undocumented
entirely, so "same as --source" would have been imprecise once added.
Added: available --theme values (report/tasks/base) and which kinds
ignore it, and a note that markdown/html_fragment tables now render
styled automatically (from this session's earlier CSS pass).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Uploading a .xlsx file no longer becomes a generic downloadable binary -
it's read server-side with openpyxl and converted to a csv artifact,
rendering in the same interactive grid as a native CSV post. Detected by
content type or .xlsx extension (some clients send generic/incorrect
content types for multipart uploads). Only the active sheet converts;
formula cells use whatever cached result the spreadsheet software last
saved (openpyxl doesn't evaluate formulas itself). Corrupt/unreadable
files return 400.
Since the existing posting script already forwards any non-text file via
multipart with no special-casing, `daybook_post.py balance_sheet.xlsx`
already works with zero script changes - verified live.
Two real bugs found and fixed while building this, both now covered by
regression tests:
- {**a_querydict, ...} silently wrapped every value (including title) in
a list - Python's ** spread on a dict subclass bypasses QueryDict's
__getitem__ override and exposes its internal one-list-per-key storage
directly. Fixed by building the delegated dict explicitly via .get().
- Pre-setting None for absent optional fields (e.g. source_prompt) defeated
create_text_artifact's own data.get(field, default) calls, since .get()
only falls through to its default when the key is *missing*, not when
the value is None - caused a NOT NULL constraint failure. Fixed by only
setting keys that were actually provided.
- csv.writer's default "\r\n" terminator was getting doubled by
Path.write_text()'s own newline translation on Windows, producing a
blank line between every row. Fixed with lineterminator="\n".
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Grid.js and renderjson (vendored for the csv/json kinds) are reachable as
plain static files from any artifact, including html_full - a better
default than an external CDN for table/JSON rendering specifically, since
it works offline and has no third-party dependency.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Any artifact can now be made publicly viewable via a share link, no login
required - opt-in per artifact, toggled from the feed UI next to Keep.
- Artifact.share_token: nullable, unique, high-entropy (secrets.token_urlsafe(32)).
NULL = not shared; SQL treats each NULL as distinct so uniqueness still
holds for the many not-shared rows.
- New public route GET /s/<token> (shared_artifact_view), completely
unauthenticated, serving the same rendered/raw content as the owner
route via a shared serve_artifact_file() helper.
- toggle_artifact_share: generates/clears the token: mirrors
toggle_artifact_keep's pattern.
- Fixed a real gap this surfaced: embedded images in a shared artifact's
content would otherwise break for an anonymous visitor, since asset_view
was owner-only. It now also allows access when the asset's parent
artifact is currently shared - unattached (orphaned) assets stay private
regardless.
- serialize_artifact now reports is_shared/share_url so the API reflects
sharing status (read-only for now; toggling stays UI-only).
- Feed UI: Share button (mirrors Keep's toggle interaction), a copyable
share-link box that appears when shared, and a "shared" badge on the
artifact row - distinct blue accent from Keep's green so both can be
visible on the same row without confusion.
Verified live end-to-end (toggle on/off, public access, fake-token 404,
old-link-dies-on-unshare, embedded-asset access before/after sharing)
before writing the 12 automated tests in tests/views/test_sharing.py.
Also fixed two unrelated stale spots noticed in docs/api.md while adding
the Sharing section: the example response was missing is_kept/size_bytes,
and the raw-content table didn't list json/csv (added earlier this
session but never documented there).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
TEXT_KINDS and TEXT_SUFFIXES now include json/csv, with matching
sniff_text_kind branches so both explicit --kind and positional auto-mode
work. Added validate_payload's JSON shape check mirroring the server.
Updated SKILL.md/docs/llm_posting_helper.md with real usage examples,
replacing the stale --as-markdown workaround guidance (kept only for
genuinely unstructured code, which still has no dedicated kind). Also
fixed an unrelated stale doc line found in the same table: html_full's
validation was documented as requiring <!doctype AND <html, but only
<html is actually required.
Verified via dry-run (explicit --kind, auto-mode, invalid-JSON rejection)
and one real end-to-end post through the script against the live dev
instance.
Added tests/scripts/test_daybook_post.py - 17 tests, plain unittest, zero
new dependencies, loaded via importlib since the script isn't a package.
Closes the test-coverage question the issue raised.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
#009: html_fragment's content-shape check is substring-based, so it can
false-positive on inert text inside <script>/<pre> (e.g. a JS string
literal that happens to contain "<html>"). Documented as a known
limitation in docs/api.md rather than implementing real HTML parsing - a
proper fix is a much bigger change than the narrow trigger justifies, and
substring matching is the same approach the client helper already uses.
#011: added a curated CDN Snippets section to SKILL.md for html_full
(Tailwind, Chart.js, with an offline/air-gapped caveat), scoped explicitly
to that kind only. Fixed daybook_post.py's --theme help text, which named
a non-existent "bootstrap" theme (written ahead of a vendored-Bootstrap
feature that was planned but never built, not a typo).
Also updated SKILL.md's kind cheat sheet, which never mentioned the
json/csv kinds and still pointed at the old --as-markdown workaround for
them - and filed issue #012 for a real gap that surfaced doing so: the
posting script itself doesn't support --kind json/csv yet, only the raw API
does.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
These accepted any JSON type on both POST and PATCH, silently mangled
into Python repr() strings at save time (a dict became the literal text
"{'a': 1}", unparseable as JSON) - and the immediate response echoed the
raw pre-save value, while a later GET returned the mangled one, an
internally inconsistent API.
Added validate_string_fields(), rejecting non-string values with 400
before they reach save(), matching #005's approach for tags. Verified live
and with new tests.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
make_room_for() read usage before any write lock was held, so concurrent
requests could each read stale data and both conclude "this fits" -
confirmed reproducible (measured 616 bytes over a 1551-byte quota with 5
truly concurrent posts).
Fix: force SQLite to acquire its write lock immediately via a no-op UPDATE
on the user's UserProfile row, before reading `used`, wrapped in
transaction.atomic() through the subsequent Artifact.objects.create() so
the lock is held across the whole eviction+insert sequence. Added a 20s
DB timeout as headroom for legitimate contention.
Deliberately no retry/backoff logic - this project's actual load (a
low-volume self-hosted tool) doesn't need it, and SQLite has no row-level
locking regardless, so quota-affecting writes serialize globally either
way. Did fix a real bug found while verifying this: exception handling
around artifact creation only caught QuotaExceeded, so a genuine lock
timeout left the just-written file orphaned on disk - both call sites now
clean up on any exception.
Verified live against the real dev instance (twice, with 5 truly
simultaneous posts against a tight quota) rather than via an automated
test - Django's in-memory shared-cache test database uses a different,
non-retryable SQLite locking model that fails for reasons unrelated to
this fix's correctness.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Markdown pipe-tables and hand-written html_fragment <table> markup already
rendered - the only existing rule was 'max-width: 100%', so they came out
as bare, unbordered browser-default tables. Adds borders, header
background, cell padding, and zebra striping, plus a display:block +
overflow-x:auto pattern so wide tables scroll horizontally instead of
squeezing columns or breaking page layout.
Lives in base.css, which report.css and tasks.css both @import, so it
applies to every existing and future markdown/html_fragment artifact
immediately - no agent-side change needed, since this is rendered live
from CSS rather than baked into stored HTML.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Gap 1 from planning/2026-09-02-artifact_kind_gaps_for_frequent_agent_output.md.
Chose new first-class kinds over riding on html_fragment, resolving that
doc's open design question in favor of filtering support, raw-source
preservation (like markdown's .md sidecar), and server-owned validation.
- json: validated with json.loads at post time (400 on invalid JSON),
rendered with renderjson (collapsible tree), vendored locally.
- csv: parsed server-side with the stdlib csv module into columns/rows and
handed to Grid.js as plain data - no client-side CSV-parsing plugin
needed, keeping normalization server-owned. Vendored locally (MIT).
- Both preserve their original text as a sidecar file, generalizing what
was previously markdown-only handling (storage.py's SIDECAR_SUFFIXES) so
disk-size accounting, eviction cleanup, and ?raw=1 all cover json/csv too.
- Raw JSON is embedded in a <script type="application/json"> block with
`</script` occurrences escaped (safe because JSON's \/ escape round-trips
correctly) - verified with a live </script>-injection payload against the
actual running server, not just unit tests.
New migration (0012) for the KIND_CHOICES change - SQLite doesn't enforce
choices at the schema level, but Django's migration state tracks it.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Enables codehilite on the markdown renderer (Pygments, 'friendly' style,
vendored as static/themes/pygments.css) so fenced code blocks get real
syntax highlighting with zero client-side JS - matching the "themes are
CSS-only" stance in framework_artifact_rendering.md. Verified live against
the running dev instance: a fenced python block renders with highlighted
keyword/string/function spans.
This was Gap 2 from planning/2026-09-02-artifact_kind_gaps_for_frequent_agent_output.md,
picked as the first fix since it needed no new kind or design decision -
just enabling an existing markdown extension.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Confirms all five are live on the local dev instance without a restart
(autoreload picked them up), and suggests where to probe next - boundary
cases on the new length/content-shape validation, other wrong-typed PATCH
fields beyond tags, and eviction race conditions #006 doesn't cover.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
- #006 (high, data loss): make_room_for() no longer deletes unkept
artifacts on a doomed request - checks whether eviction can possibly
succeed before deleting anything. Also fixed a related latent bug found
while testing: the function returned evicted Artifact objects after
calling .delete() on them, so their .id was always None (Django clears
pk on delete); now returns ids captured beforehand.
- #003: added server-side field-length validation (source/title/
source_name/source_id/theme) on POST and PATCH, matching the model's
own max_length limits, which were previously unenforced on the ingest
path entirely.
- #004: ported daybook_post.py's content-shape validation
(html_fragment/html_full/deck/markdown) to the server, closing the gap
for any client that doesn't go through the helper script.
- #005: PATCH now rejects wrong-typed `tags` (previously silently wiped
existing tags) and rejects immutable fields (content/source/kind/day)
with 400 instead of silently no-op'ing them.
- #002: deleted the misleading skills/daybook-posting/.env.example (a
zero-dependency script that never reads .env files) and clarified
SKILL.md/docs/llm_posting_helper.md that these must be real
process/shell env vars.
Adds tests/unit/ for storage-layer tests. 46 tests passing (was 25).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
AGENT_DAYBOOK__LLM_TEST (a sibling project) always uses this account to
test the API. Documenting it here so a future reset of the local dev
database gets it recreated instead of silently breaking that project.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Peter needed a second user for throwaway testing without going through
Django admin each time. Superusers now get a "Users" tab in Settings
(hidden for regular users) with a create-user form and a list of existing
accounts. New users are regular (non-staff, non-superuser) by default, with
an optional "Grant admin access" checkbox; creating one immediately issues
their API token and displays it once, since that's the credential an agent
account needs and a human account can otherwise regenerate for themselves
later from their own Settings -> Account.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
The app-side fix (CSRF_TRUSTED_ORIGINS, SECURE_PROXY_SSL_HEADER in
settings.py) was already committed as d458cd3 after the first Cloud-VM-1
deployment hit "CSRF verification failed" on login behind Apache. This
finishes the rest of that issue's checklist: documents
DJANGO_CSRF_TRUSTED_ORIGINS in deploy/env.example, adds the required
X-Forwarded-Proto header to the Apache/Caddy examples in
cloud_agent_deploy.md (plus a checklist item that verifies via an actual
login, not just an unauthenticated capabilities curl), and adds the same
proxy-agnostic guidance to docs/deployment.md.
Also introduces issues/, a tracked one-file-per-issue log for the project
going forward.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Django had no CSRF_TRUSTED_ORIGINS and no SECURE_PROXY_SSL_HEADER, so
it always treated requests as plain HTTP. Login over HTTPS behind
Apache/Nginx failed CSRF verification as a result.
