eb1324cb167e79656f725606d1ca85475b3a8800
2
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
  shadow-testandClaude Opus 5
|
cc5f691677 |
Add llama.cpp backend, model gateway, URL relay and browser view
Four features, plus a latent bug fix.
llama.cpp backend. Claude Code only ever speaks the Anthropic Messages
API — confirmed empirically by pointing it at a logging server, which
received POST /v1/messages?beta=true. llama-server implements that
natively (verified in its README, alongside --port default 8080), so
this is a plain base-URL backend with no translation shim, the same
shape as Ollama. Its --api-key defaults to none, so the auth token is a
placeholder Claude Code requires and llama-server ignores.
Model alias fix. ANTHROPIC_DEFAULT_HAIKU_MODEL is documented as "also
used for background functionality", and Triple-C set none of the alias
vars. So on every custom-endpoint backend, Claude Code resolved `haiku`
to an Anthropic model id and sent it to a local server that does not
have it — background features failed silently. All four
ANTHROPIC_DEFAULT_{OPUS,SONNET,HAIKU,FABLE}_MODEL vars are now pinned to
the backend's configured model, with an optional Haiku override, and
blanked for Anthropic and Bedrock so those keep Claude Code's defaults.
The deprecated ANTHROPIC_SMALL_FAST_MODEL is never emitted. Existing
Ollama and OpenAI-Compatible containers are recreated once so the new
env reaches them; the snapshot is preserved.
Model gateway. Optional LiteLLM sibling container, off by default,
mirroring stt.rs — this is what makes real OpenAI usable, since
api.openai.com has no /v1/messages. Pinned to v1.96.0 by tag and digest:
the 1.82.7/1.82.8 malware was PyPI-only and never affected the official
images, which is precisely why this builds FROM the image rather than
pip-installing, but 1.84.0 is still the floor for proxy CVEs (API-key
SQLi, Host-header auth bypass, MCP auth bypass). Binds 0.0.0.0 because
project containers consume it, and therefore always sets a master_key —
LiteLLM without one accepts any key. The provider key lives in the OS
keychain and is uploaded into a volume, never an image layer or label.
URL relay. A container-side xdg-open/BROWSER shim opens URLs in the
host's browser. Uses an OSC sequence to /dev/tty rather than a printed
sentinel, because the shim usually runs as a grandchild of a process
capturing its children's output. Degrades to printing the URL when no
terminal is attached, so scheduled tasks do not hang. Only http/https,
with control characters rejected before new URL() — which strips
newlines, so java\nscript: would otherwise parse as javascript:. Nothing
auto-opens; the user confirms. The web terminal shows a tap-to-open
banner instead, since that browser may be a phone across a tunnel.
Browser view. A Project Home tab that watches and takes over the browser
Claude drives with Playwright, using Playwright's own dashboard. Zero
image cost — Playwright stays user-installed. It does not reuse the auth
bridge's PortForward, which binds an unauthenticated port: correct for a
throwaway OAuth listener, wrong for mouse and keyboard control of a
browser in a passwordless-sudo container. Instead a token-gated loopback
proxy checks Host, then token or a forbidden-header origin signal,
before a byte reaches the container. Host ports are confined to
47820..=47827 so CSP frame-src can enumerate them rather than widening
to a wildcard, with a test asserting the two agree.
188 frontend tests, 107 Rust tests, both builds clean.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
||
|
shadow-testandClaude Opus 5
|
01a2f6aec8 |
Add Project Home, Auth Bridge, shared auth token, and Tier-1 polish
Project Home (DESIGN-REVIEW §B2): the project is promoted from a 280px
sidebar card to a first-class main-area view. ProjectCard.tsx (1,257
lines) is replaced by a select-only ProjectRow plus tabs for Overview,
Sessions, Automation, Config and Files. The PortMappings, FileManager
and ContainerProgress modals are absorbed rather than reimplemented.
Config gains a Saved/Saving/Failed indicator — save-on-blur failures
previously reached only console.error.
Tier-1 polish (DESIGN-REVIEW §A): new elevation, muted-accent, disabled
and focus-ring tokens; a global :focus-visible ring with every
focus:outline-none removed; filled buttons moved to --accent-emphasis
and white-on-success toggles retired, fixing three WCAG AA failures
(2.1:1, 2.5:1, 2.4:1); a shared Modal primitive with role="dialog",
focus trap and restore, adopted by all remaining modals; status
indicators that carry a glyph and word rather than colour alone.
Ctrl+Shift+W closes a tab, deliberately not Ctrl+W — that is readline's
kill-word, used constantly in the terminal this app is built around.
Auth Bridge: a general loopback-callback bridge so browser logins run
inside a container (aws sso login, Concourse fly login, claude login)
can complete against the host browser. Listeners are discovered from
/proc/net/tcp{,6} — ss/netstat/lsof are absent from the image — bound on
host 127.0.0.1 only, and tunnelled in over the Docker API via socat,
which keeps working on Docker Desktop where container IPs are not
routable. Falls back to [::1] because Node resolves localhost to IPv6
first, so claude login often binds ::1 alone. Opt-in per project.
This extracts create_attached_exec() and moves the existing terminal
session path onto it, so there is one attached-exec implementation
rather than two.
Shared auth token: `claude setup-token` is run in a container, the token
is stored in the OS keychain and injected as CLAUDE_CODE_OAUTH_TOKEN
into Anthropic-backend projects. Contrary to the initial design note,
setup-token uses an Anthropic-hosted redirect and blocks on a stdin
paste prompt rather than a loopback callback, so a stdin command is
required for the flow to complete.
The token is never logged, never returned to the frontend, and is
redacted from the streamed output with a stateful matcher that withholds
any tail that could still grow into a secret. Change detection uses a
random rotation id rather than a hash, since a hash in a docker-inspect
readable label would be an offline verification oracle.
Frontend 33 -> 51 tests; Rust 34 tests. Both builds clean.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|