* formats, mcp: add parity tests for SQLite, GPU textures, and MCP transcripts
* mcp: use BufReader in stdio_cancel_inflight_tool_call_transcript
* mcp: pass all_tools in cancel transcript test to avoid list_changed notification
* Claim the interrupted loopback call line (R5.2)
* Call a loopback Cutter session again when the call is interrupted, instead of saying nothing answered (R5.2)
* Offer Requesty by name when its key is set
Requesty is one more row in the provider table: its https base URL, REQUESTY_API_KEY and a default model, so Detect offers it when the key is set and the key goes only to router.requesty.ai.
* List the chat models of a catalogue that names its API as text
Requesty's /models gives each entry "api": "chat", which the filter for OpenCode Go's object form read as not chat, so the list came back empty. Ids with control characters are dropped too.
* Treat a null model api as unspecified and pin padded and missing ids
---------
Co-authored-by: Tanish Manku <[email protected]>
* Claim the bad plugin argument list line (R2.9)
* Say when a plugin's bracketed arguments aren't a list of quoted words, and keep Add off until they are (R2.9)
* Plan carrying a project's stated purpose to its approval card (R2.9)
* Ask each project its purpose and keep it private unless chosen (R2.9)
* Read projects made before the purpose question as unknown (R2.9)
* Show a stated purpose on the card only; it changes nothing else (R2.9)
---------
Co-authored-by: Simon Koch-Sultan <[email protected]>
* Claim the release workflow runtime audit (R5.2)
* Record the runtime audit of the shared release actions
* Refresh the release workflow runtime audit
On Fedora (and other GTK desktops), the hicolor icon cache is not
rebuilt after the AppImage copies its icons into $XDG_DATA_HOME.
The desktop keeps showing a generic icon until the user logs out
or manually runs gtk-update-icon-cache.
Add a conditional cache refresh to the AppRun integration step:
- Track whether any icon file was actually copied (changed).
- Only refresh if the cache file (icon-theme.cache) already exists,
so we never create one where the desktop hasn't built one yet.
- Skip silently when gtk-update-icon-cache is not installed.
The fix lives in the AppImage's own AppRun rather than in the
launcher because the launcher (appimagelauncher) is a separate
project with its own release cycle, and many users run AppImages
directly without it. Both fixes are complementary.
Tested on Fedora 41 (GNOME 47, Wayland) by launching from the app
grid after clearing the icon cache. The correct icon appeared on
the first launch without a logout.
Closes#1877
Adds the hardened-runtime entitlement for sending Apple Events and the usage text macOS shows in its Automation prompt. Without them macOS blocked the events and never asked.
The _headers file (for Netlify/Cloudflare Pages) already sends
X-Content-Type-Options: nosniff on all responses, but the Apache
.htaccess did not. This adds the same header for parity, so Apache
hosts also tell browsers to respect the declared MIME types for
.wasm and .js files instead of guessing.
Uses `Header always set` so the header is included on error
responses too, not only on successful ones.
Also documents the header in packaging/web/README.md under
"Required server settings" so self-hosting instructions mention it.
Verified: the .htaccess already requires mod_headers with IfModule,
so this is a no-op on servers without it. The _headers file is
unchanged.
The deb, rpm, AppImage and Flatpak all ship craft-anything-cli beside
the desktop app, but the metainfo only declares the GUI binary. Add a
second <provides><binary> element for the CLI so appstreamcli and
software centres know the package provides it.
Inside the AppImage and Flatpak the CLI is not on the host's PATH, so
this tells the package manager the binary exists in the package even
though the person's shell won't find it by name.
Add a ticked roadmap line under R5.1.
Verified: the change passes xmllint and the metainfo structure matches
the AppStream specification for <provides><binary>.
* Claim one text size in Simple and Advanced (R2.9)
* Claim one text size in Simple and Advanced, as a decision (R2.9)
* Use one text size in Simple and Advanced, so switching modes doesn't zoom (R2.9)
theme::apply no longer takes the mode. Both modes use the type scale: body
and buttons at BODY (15), small text and code at CAPTION (12), headings at
SECTION (18), controls 30 tall like the app's own buttons and switches, and
the tighter spacing. Simple was 16 pt with roomier controls and Advanced 13,
so every mode switch looked like a zoom; zoom (Cmd/Ctrl + / -) stays the way
to make everything bigger.
The roadmap line, dropped from this branch by a merge of main made in
GitHub's editor, is back. A test checks both modes draw at the same sizes;
the Settings test that adds an AI by its address scrolls to its form, now
just below the fold.
* Say in Settings' Mode row how to make text bigger (R2.9)
With one text size in both modes, zoom is the way to larger text: the Mode
row now names its keys (Ctrl/⌘ + and −, Ctrl/⌘ 0 to reset), in every catalog.
* Check every text style and all spacing match in both modes, also after a theme and zoom change (R2.9)
Review of #1236: the test compared Body, Small, Heading and the control height only. It now
compares all five text styles and egui's whole spacing (padding, gaps, control sizes, margins)
between Simple and Advanced, then changes the theme and the zoom in Advanced, switches back to
Simple, and checks the sizes still match and the zoom is left as it was.
---------
Co-authored-by: Tanish Manku <[email protected]>
* Claim carrying pictures to API models (R5.4)
Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Let a message to an API model carry pictures
Adds Block::Image: PNG, JPEG, GIF or WebP read from the bytes, up to 3.75 MB. OpenAI-style servers get image_url data URIs in the message's content parts, Anthropic an image block with a base64 source; messages without a picture are sent as before. The bytes stay in memory only: transcripts keep the picture's name and a resumed conversation sends a note instead. Pictures count as 1,600 tokens and old ones are trimmed to their note.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
* State only the limit when a picture is too large to show
image_file reads one byte past the limit, so the size it read isn't the file's, and dividing it as an integer printed "3 MB, more than the 3.75 MB". Found by CodeRabbit.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Add the picture follow-ups from #1175's review to the roadmap (R5.4)
Before pictures reach a builder: a budget for a request's pictures as a
whole and each provider's size and dimension limits; tests with real,
decodable pictures against a provider that takes them, and with several
live and resumed pictures interleaved with text.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
---------
Co-authored-by: Claude Opus 5.5 <[email protected]>
Co-authored-by: Tanish Manku <[email protected]>
* Claim R2.9 sign-in button on Make
* Sign in to a website from Make (R2.9)
A website address typed on Make now offers "Sign in to this site", which
runs web_site_sign_in for that address, so a site behind a login can be
signed in to from the app, not only from the CLI, MCP or the agent.
One window at a time; while it's open the row says so and offers "Stop
waiting", and Start (and every other way a start is reached) waits.
The result is said in a message: kept, still asking to sign in, check
failed, timed out or stopped.
* Let the i18n check see the sign-in messages
The sign-in toasts and the button's hover text were looked up through a
variable or inside format!, so i18n-coverage counted their awaiting.txt
entries as unused and the gate failed. Look each one up with tr() on the
literal.
* Name the site in sign-in messages and wait for project creation
The sign-in button stays disabled while a new project is being made, so its build
can't study the site before the sign-in ends. Each sign-in message starts with the
address the sign-in was started for, which may no longer be the one on Make. A closed
window without a password field is reported as closed, not as signed in. The tool is
matched by its constant.
* Keep credentials out of sign-in toasts and show the i18n check every sign-in string
The site named in a sign-in toast drops userinfo, query and fragment, as the engine does for URLs it shows. "Stopped the sign-in." is looked up outside format! and the i18n check now reads Button::new labels, so neither awaiting entry is reported as unshown.
* Hold an update's install while a sign-in window is open
Installing closes the app, and the sign-in's job is invisible, so jobs.busy() didn't see it: closing then could lose the sign-in before the browser profile is saved. The install gate is now safe_to_install(), which also waits for signing_in.
---------
Co-authored-by: Mansur Akhmadi <[email protected]>
Co-authored-by: Tanish Manku <[email protected]>
* docs(formats): classify single-app formats and add ASAR/AXML/ARSC boundary tests (storytold/craft-libs#778)
* fix(formats): accept Truncated or Malformed for short ARSC header
* test(formats): use < 8 byte header in arsc short header test
* Cut over MCP line and JSON-RPC envelope framing to craft-libs (#778)
* mcp: clone request id before worker thread spawn
* chore: update Cargo.lock for craft-jsonrpc
* mcp: add deterministic stdio transcript parity and framing bounds tests
* chore: repin craft-jsonrpc and craft-mcp-server to craft-libs main
* Plan honest usage counts and a cost-limit design in the roadmap (R5.3)
* Say a limit that holds has to treat the turn's price as an estimate (R5.3)
---------
Co-authored-by: Simon Koch-Sultan <[email protected]>
Co-authored-by: Tanish Manku <[email protected]>
* Claim R1.5: mcp-http-cleartext-guard
* Validate MCP HTTP URLs and refuse cleartext or loopback redirects (R1.5)
- Parse `url_allowed` inputs with `ureq::http::Uri`, reject userinfo (`@`), validate `https://` hosts/IPv6 brackets and `u16` ports, and delegate `http://` loopback checks to `craft_anything_platform::standin_server::host_is_loopback` so malformed or spoofed authorities (`http://[::1].evil.example.com`, `http://evil.example.com\@127.0.0.1`, `https:///...`) cannot pass.
- Trim `ServerUrl::url` in `Http::new` to match `url_allowed`.
- Enforce `.https_only(true)` on `https://` MCP requests (reporting `RequireHttpsOnly` without echoing the redirect target URI) and `.max_redirects(0)` on loopback `http://` requests so plugin headers and payloads never follow redirects off loopback or across cleartext HTTP.
* Reject mid-host brackets and signed ports, and drain POST bodies in tests (R1.5)
- Reject mid-host square brackets (`!host.contains(['[', ']'])`) and leading `+` signs on ports across `https://` and loopback `http://` authorities in `url_allowed`.
- Drain incoming HTTP `POST` headers and `Content-Length` bodies (and clear inherited socket non-blocking mode) before sending `302` replies in the redirect regression fixtures.
* Follow loopback MCP HTTP redirects that stay on this computer (R1.5)
* Keep MCP redirect hops in one deadline, cancellable, and without configured headers off the configured origin (R1.5)
* Lengthen the per-hop delay in the MCP redirect deadline test and update STATUS (R1.5)
* Keep MCP redirect hops within the caller's deadline and the configured origin (R1.5)
Pass the caller's deadline into post without a one-second minimum, compare origins by scheme, host and effective port, refuse a session started by another origin, keep the path for query-only redirects, give the cancellation notice 2 s, and report a late answer body as a timeout.
* fix(engines): keep malformed MCP redirect targets out of errors and send cancel notices only for sent calls (R1.5)
* Discard the cached MCP client after a send-side protocol error (R1.5)
---------
Co-authored-by: Tanish Manku <[email protected]>
* Check both comparison backends before building either (R2.9)
* Keep the API refusal test independent of provider keys (R2.9)
---------
Co-authored-by: Tanish Manku <[email protected]>
* Check which list lookups run at the summary cap (R5.7)
* Record the list lookup boundary regression (R5.7)
* Use the workspace minimum Rust API for the lookup test (R5.7)
---------
Co-authored-by: Tanish Manku <[email protected]>
* Claim R2.2 Codex asset tool approval (#1464)
Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Let a Codex builder call the paid asset tools within the project's budget (#1464)
codex exec can't ask for approval, so it refused assets_image_generate and
assets_model_generate before our server saw the call. Approve them on Codex's
side too, like rebuild_start: our own limits still apply (the project's
generated-assets switch, its budget, and the question before going over it).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
---------
Co-authored-by: Claude Opus 5.5 <[email protected]>
Co-authored-by: Tanish Manku <[email protected]>
* Find a build only Cargo places by reading files, starting no process
The file lookup (the Project screen's Play and "Pin this build", a saved Live
pin) now also reads CARGO_BUILD_TARGET_DIR, build.target-dir in Cargo's home
config, and the closest workspace's target/. Every folder outside the app still
gives only a program whose build record names the app's sources.
* Say which of Cargo's rules the file lookup approximates; read a Cargo.toml only when it is a file under 1 MiB
A relative CARGO_HOME means no home config is read (not ~/.cargo).
* Bound the read of Cargo's config too; a decoy for CARGO_BUILD_TARGET_DIR; STATUS says what a Live pin asks and which rules aren't followed
* Test that Play looks in the folder Cargo names; say exactly which settings files are capped and what asks Cargo
---------
Co-authored-by: Tanish Manku <[email protected]>
* Claim: keep development builds from filling the disk with debug information (R5.2)
Co-Authored-By: Claude Fable 5.2 <[email protected]>
* Build development code with line tables only and dependencies without debug information (R5.2)
Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Say how to bring back dependencies' debug information too
Co-Authored-By: Claude Opus 5.5 <[email protected]>
---------
Co-authored-by: Claude Fable 5.2 <[email protected]>
Co-authored-by: Tanish Manku <[email protected]>
* Claim the live Go project checks (R2.7)
* Exercise rebuild checks with a real Go project (R2.7)
* Record live Go project check evidence (R2.7)
---------
Co-authored-by: Tanish Manku <[email protected]>
Every job that changes an AI hands back a snapshot of the AI list and its options, and the app showed whichever finished last. A snapshot read before a save but finished after it put the old values back on screen until the next refresh, which is also why claude_codes_account_folder_is_set_in_settings failed about one CI run in ten. Snapshots are now numbered as they're read, and one older than the snapshot already shown is dropped.
Co-authored-by: Tanish Manku <[email protected]>
* Claim formatting in batches that fit Windows' command line (R5.2)
* Format in batches that fit Windows' command line: cargo xtask fmt (R5.2)
* Say what cargo xtask fmt formats, and run the rustfmt RUSTFMT names
---------
Co-authored-by: Tanish Manku <[email protected]>
* Claim unreadable folder entry names (R3.11)
* Report unreadable entry names without merging file identities (R3.11)
* Format unreadable entry regression iterator
* Keep invalid filename fixtures off Apple targets
APFS rejects the raw filename bytes used by the two disk-backed regressions.
Keep the iterator-only limit coverage enabled on Unix and record the fixture
constraint in the contributor pitfalls.
Signed-off-by: Afloat <[email protected]>
---------
Signed-off-by: Afloat <[email protected]>
* Claim R2.9: start a new Claude Code conversation when the saved one is too long (#2042)
* Start a new Claude Code conversation when the saved one has grown too long (R2.9)
A conversation past what the model accepts made Claude Code refuse every later turn with "Prompt is too long", and Try again resumed the same conversation, so the project was stuck (#2042). That refusal is now read like a saved conversation that isn't on this computer: the saved session is dropped and the next message starts a new one from what the project already has.
---------
Co-authored-by: Tanish Manku <[email protected]>
egui-wgpu requests textures of 8192 px at most whatever the GPU supports, and the window's surface can't be larger, so on a display wider than that the app aborted before its window showed (#1639). The device request now keeps egui-wgpu's settings and raises the texture limit to the GPU's own, never lower than before.
Co-authored-by: Tanish Manku <[email protected]>
* Correct what Arch's Ghidra package installs, and what a Ghidra file setting accepts
Arch's extra repository ships ghidra (12.2), with /usr/bin/ghidra and
/usr/bin/ghidra-analyzeHeadless as links into /opt/ghidra; the earlier
text said it had no package and described a start script that was
written for the test. That real layout is found with each setting.
A Ghidra file setting is kept when it is a short text file naming the
launcher, which is not checked to run it; the wording said it was.
* Say that any file in a Ghidra folder gives its launcher, and that only Arch's own package was tried
A readable sqlite_schema row could still hold a partial value (an overflow chain that ends
early, loops or leaves the file, or text cut at a schema budget), so a table's column list could
be short while the schema was reported complete. Such records now set schema_truncated and add
a counted integrity note; the object's readable name and metadata are kept.