When the host kills a container (absolute-ceiling, claim-stuck, or crashed),
resetStuckProcessingRows reset messages_in but left orphan rows in
processing_ack. The next sweep tick spawned a fresh container and, on the
same tick, ran enforceRunningContainerSla against outbound.db that still
contained the previous container's claim with a hours-old status_changed
timestamp — instant kill-claim, before the agent-runner could open
outbound.db to run its own clearStaleProcessingAcks(). Loop until tries
hit MAX_TRIES.
Add deleteOrphanProcessingClaims() in session-db and call it at the end of
resetStuckProcessingRows. Safe to write outbound.db here because the host
only enters this path after killContainer (or when no container is running).
Tests in host-sweep.test.ts cover the helper plus the regression: orphan
claim from a 2h-old kill is now removed atomically with the messages_in
reset, so the next sweep tick sees an empty claims list and the freshly
respawned container survives long enough to start its agent-runner.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Move the MIME/type-to-extension maps and derivation helpers out of
session-manager.ts into a dedicated attachment-naming module — keeps
session-manager focused on session lifecycle and gives the helpers
a natural home for unit tests alongside the existing attachment-safety
module.
Two small fixes alongside the extraction:
- extForMime now guards `typeof mime !== 'string'` before .split, so a
buggy bridge passing `mimeType: { ... }` (object) no longer crashes
the inbound write loop.
- deriveAttachmentName computes Date.now() once per call instead of
twice, and tightens the explicit-name check to a string-and-truthy
guard so non-string values fall through to derivation.
Adds attachment-naming.test.ts with 11 cases covering MIME normalization
(case + parameters), Telegram type fallback, the non-string defensive
guard, and the bare-timestamp fallback.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Removing the "Not logged in · Please run /login" detection and
substitution from this PR — narrowing scope to just the OneCLI
gateway transient-retry change. The login-message handling will be
addressed separately.
Reverts:
- AgentProvider.isAuthRequired / authRequiredMessage
- ClaudeProvider auth-required regex, classifier, and remediation text
- poll-loop writeAuthRequiredMessage helper + call sites
- claude.test.ts (auth-only test file)
OneCLI/wakeContainer changes (the remaining content of the PR) are
unaffected.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Closes#1820.
The container agent-runner sets CLAUDE_CODE_AUTO_COMPACT_WINDOW
unconditionally on the container process env, with no way to override
it per-deployment without editing source. Read process.env first and
fall back to the existing 165000 literal when unset.
Default behavior is unchanged for installs that do not set the env
var. Operators running 1M-context models or emergency-tuning a live
deployment can now raise or lower the threshold from the host env.
Adds a paired `authRequiredMessage()` method to AgentProvider so
per-provider auth-failure remediation can differ. Claude returns the
Anthropic/`claude` instruction; future providers (Codex, OpenCode, …)
can return their own remediation text. The poll-loop calls
`provider.authRequiredMessage?.()` and falls back to a generic message
if a provider implements `isAuthRequired` without supplying its own
remediation.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
When a channel bridge passes an attachment without an explicit `name`,
extractAttachmentFiles fell back to `attachment-<ts>` with no extension.
Agents could not tell whether the file was a JPEG, PDF, or audio clip,
and tools keyed on extension (image viewers, exiftool, etc.) misbehaved.
Two cases are now covered:
1. Channels that set `mimeType` but no `name` (Discord/Slack documents,
Telegram document uploads). A small MIME-to-extension table covers
the common content types — image/*, audio/*, video/*, pdf, zip,
txt, json. Unknown MIMEs fall back to the unsuffixed name.
2. Channels that set `att.type` but no `mimeType` (Telegram photos,
stickers, voice, animations). The chat-sdk bridge sets a coarse
media-class (`photo` / `sticker` / `voice` / `video` /
`animation`) which is reliable enough to derive a canonical
extension. Telegram GIFs are MP4 under the hood.
The existing isSafeAttachmentName security guard is preserved — the
derived name still passes through it before disk I/O. The new lookup
tables emit static values from internal maps and cannot construct a
path-traversal payload; attacker-controlled att.name continues to flow
through the same validator.
- wakeContainer now never throws — returns Promise<boolean>, catches
internally. Closes the regression risk for the 5 awaited callers in
agent-to-agent, interactive, and approvals/response-handler that the
previous version left unwrapped. Router uses the boolean to stop the
typing indicator on transient failure; host-sweep just awaits.
- Tighten AUTH_REQUIRED_RE: anchor to start-of-string with the specific
`·` (U+00B7) separator the CLI uses, so an agent that quotes the
banner mid-sentence in a normal reply doesn't trip the classifier.
- Log a one-line note from writeAuthRequiredMessage so substitutions
are visible when debugging "user got the credentials message but I
don't see why."
- Add unit tests for ClaudeProvider.isAuthRequired covering both banner
variants, trailing content, mid-sentence quoting, leading-prose
quoting, alternate separators, and unrelated text.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two related fixes for the case where credentials aren't usable:
1. Replace Claude Code's "Not logged in / Invalid API key · Please run
/login" output with a host-aware message. The user can't run /login
from chat, so the raw text is unhelpful. Provider gains an optional
isAuthRequired() classifier; the poll-loop substitutes the message
on both result-text and error paths.
2. Treat OneCLI gateway failure as a transient hard error instead of
spawning a credential-less container. The catch in container-runner
now propagates; router and host-sweep wrap wakeContainer to log and
leave the inbound row pending so the next 60s sweep tick retries.
Router also stops the typing indicator on failure.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Wraps the word "assistant" in `accentGreen` (#3fba50, added in #2103)
across the six channel adapters that ask "What should your assistant
be called?" — Discord, iMessage, Signal, Slack, Telegram, WhatsApp.
Mirrors the green emphasis on "you" in the display-name prompt: the
green word names the subject of the question (assistant vs operator)
so the operator parses it at a glance.
Adds an `accentGreen` helper (#3fba50) with the same TTY/NO_COLOR/
truecolor gating as the rest of the palette, then wraps the word
"you" in the "What should your assistant call you?" prompt so the
operator parses at a glance who the question is about — the user,
not the assistant. The mirror prompt that asks for the assistant's
name ("What should your assistant be called?") is left for a
follow-up.
Customize `brightSelect`'s render function so the focused option's
label paints in brand cyan during selection and the submitted answer
paints in dim cyan after the user moves on. Inactive options keep
their default rendering — only the cursor and submitted state pick
up the color, matching the body-text emphasis added in #2101.
Also migrate the one remaining `p.select` call site (the "What next?"
prompt after the first chat) to `brightSelect` so every menu in the
setup flow goes through the same render path. The shape of the call
matches what `brightSelect` already supports — message + options
with value/label/hint — so no feature is lost in the swap.
Reuses `brandBody` from #2101 for the cyan, so the prompt highlight
and the body prose share one definition of the brand body color.
Adds a `brandBody` helper in setup/lib/theme.ts that wraps prose in
brand cyan (#2BB7CE), with the same TTY/NO_COLOR/truecolor gating used
by `brand`/`brandBold`/`brandChip`. The helper splits multi-line input
and colors each line independently so the SGR sequence doesn't bleed
across clack's gutter prefix.
Routing:
- `note()` (the un-dim card wrapper from #2095) now passes
`brandBody` as its `format` callback, so card bodies render
cyan line-by-line.
- Every prose `p.log.{message,info,success,step,warn}` call in the
setup flow wraps its body argument in `brandBody`. Calls whose
body is explicitly `k.dim(...)` (failure transcript tails, log
paths, claude-assist response previews) are left alone — those
are the "preview/debug" cases the dim-policy comment in
theme.ts already carves out.
- Spinner-finish lines in windowed-runner / claude-assist color
only the message portion; the `(5s)` elapsed suffix stays dim.
Brand cyan accents (chips, wordmark, inline emphasis) are unchanged.
This PR only adds the body color.
A follow-up will add OSC 11 dark/light detection so light-mode
terminals get a brand blue (#2b6fdc) variant — opt-in upgrade with
no regression for the dark-mode default.
When pasting an invalid token, the old value stayed in the input
field. Pasting a new token appended to the old one instead of
replacing it, causing repeated validation failures.
Add clearOnError: true to all 8 password prompts across setup.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
When re-running setup on a machine that already has a .env with
channel tokens or OneCLI config, detect them early and offer to
reuse instead of prompting the user to paste everything again.
- Add detectExistingEnv() to parse .env and group known keys
- Add detectExistingDisplayName() to read display name from v2.db
- Defer display name prompt until actually needed (cli-agent or channel)
- Skip cli-agent and first-chat when groups are already wired
- Add token reuse checks to Telegram, Discord, Slack, Teams, iMessage
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Clack's `p.note` defaults to `format: e => styleText("dim", e)`, which
fades note bodies regardless of the project's stated readability stance
(see comment on `dimWrap` in setup/lib/theme.ts: "prose renders at the
terminal's regular weight"). The dim styling makes body copy hard to
read on dark terminals and visibly washes out brand-colored segments
embedded in cards (e.g. the chip + bold heading rows).
Add a `note()` helper in setup/lib/theme.ts that wraps `p.note` with a
pass-through formatter, and route every setup-flow `p.note` call site
through it: setup/auto.ts, every setup/channels/*.ts adapter, and the
two setup/lib/claude-* helpers.
Pre-styled segments (brandBold, brandChip, formatPairingCard,
formatCodeCard) now render at full strength instead of being faded
alongside surrounding prose.
Use script(1) to capture PTY output and extract OAuth token when
browser-based auth isn't available, with fallback code-paste flow.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
When setup fails and claude-assist kicks in, instead of silently
skipping when the CLI is missing or unauthenticated, interactively
offer to install it (via install-claude.sh) and sign in (via
claude setup-token) so the user can get diagnostic help immediately.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Ethan