openclawEntriesToModels() used the agent Name (which may contain
spaces, e.g. "Sub2API OPS") as Model.ID. This ID is passed to
openclaw via --agent, where normalizeAgentId mangles spaces into
hyphens ("sub2api-ops"), causing a lookup miss against the
registered id ("sub2api") and a "no parseable output" error.
Fix: prefer agent ID for Model.ID; use Name only for display Label.
When ID is empty, fall back to Name for backward compatibility.
Fixes#2714
* fix(daemon): disable Claude AskUserQuestion in non-interactive mode (MUL-2244)
GitHub #2588: when Claude Code calls its built-in AskUserQuestion tool
inside the daemon's stream-json runtime, the question never reaches the
user — there's no UI to render it — so the SDK returns an empty answer
and the agent silently "infers" and continues. From the issue's
perspective, execution looks stuck while the agent is actually charging
ahead on its own guess.
Two-part fix:
- `buildClaudeArgs` now passes `--disallowedTools AskUserQuestion` so
the tool is not exposed to the model at all.
- The Claude-specific runtime brief tells the agent to use a `blocked`
issue comment for genuine clarification, or to state an explicit
assumption and proceed.
Adds a regression test that pins both: AskUserQuestion is forbidden in
CLAUDE.md and is NOT mentioned in the AGENTS.md emitted for non-Claude
providers (the tool is Claude-specific).
Co-authored-by: multica-agent <github@multica.ai>
* refactor(daemon): drop CLAUDE.md AskUserQuestion guidance, rely on --disallowedTools
The --disallowedTools flag already prevents Claude from invoking
AskUserQuestion, so duplicating the rule in the runtime brief just bloats
the prompt without changing behavior. Removes the section and its
regression test; the argv-level test in pkg/agent already pins the flag.
Co-authored-by: multica-agent <github@multica.ai>
---------
Co-authored-by: multica-agent <github@multica.ai>
Gemini CLI's folder-trust feature throws FatalUntrustedWorkspaceError
(exit code 55) when the current workspace isn't in
`~/.gemini/trustedFolders.json` and the process is headless — no
interactive trust prompt is available. The daemon spawns gemini with
`-p` + `--yolo` in a freshly checked-out worktree that the user has
never trusted interactively, so every run with `security.folderTrust`
enabled fails after ~10s with exit status 55 and no useful output.
Default `GEMINI_CLI_TRUST_WORKSPACE=true` on the child env to short-
circuit `checkPathTrust` in gemini-core. This mirrors gemini-cli's
documented `--skip-trust` flag; the env var has been gemini's
documented headless escape hatch for the entire folder-trust feature
lifetime so the fix works on every gemini version that can produce
the crash. Callers that explicitly set the same key in cfg.Env win,
preserving the ability to opt back into the gate.
Co-authored-by: multica-agent <github@multica.ai>
The gemini CLI's Windows shim emits `Active code page: 65001` (from
`chcp`) to stdout before the real version reaches `--version` output.
The daemon stored the raw concatenation as the runtime version, so the
runtime detail page rendered `Active code page: 65001 0.42.0` instead
of `0.42.0`.
Scan `<cli> --version` line by line and return the first line carrying
a semver-shaped token. Full strings like `2.1.5 (Claude Code)` or
`codex-cli 0.118.0` survive unchanged; unparseable output falls back to
the trimmed raw value.
Co-authored-by: multica-agent <github@multica.ai>
The Pi backend hardcoded `--tools read,bash,edit,write,grep,find,ls` in
buildPiArgs. Pi's SDK treats --tools as a restrictive allowlist: only the
listed tools pass through `_refreshToolRegistry()`, silently filtering
out any user-installed extension tools registered via `pi.registerTool()`.
Omitting --tools makes Pi's `allowedToolNames` undefined, so the
`isAllowedTool()` filter becomes a no-op and all tools — built-in and
extension — are available. This matches Pi's standalone behavior.
Users who want to restrict tools can still pass --tools via custom_args
(it is not in piBlockedArgs).
Closes#2379
* fix(agent): expand Copilot CLI model catalog with correct dotted IDs
The Copilot CLI provider only exposed two models in the runtime
dropdown, and one of them used the dashed legacy form
`claude-sonnet-4-6` which `copilot --model` rejects with
"Model ... is not available". The CLI accepts dotted IDs
(e.g. `claude-sonnet-4.6`, `gpt-5.4`).
Sync `copilotStaticModels()` with the official supported-models
catalog so the dropdown surfaces the full set the user's account
can route to (8 OpenAI + 4 Anthropic), and add a regression test
that pins the expected IDs and bans the dashed form.
Closes MUL-1948.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: multica-agent <github@multica.ai>
* feat(agent): dynamic Copilot model discovery via ACP session/new
The previous static catalog could only ever lag behind the user's
real entitlements and what GitHub ships. Copilot CLI exposes the
live catalog through its ACP server (`copilot --acp`): the
`session/new` response includes `models.availableModels` plus
`currentModelId`, scoped to the authenticated account.
Wire copilot through the existing discoverACPModels helper —
already used by hermes/kimi/kiro — so the dropdown reflects the
account's real catalog, including the `auto` entry and per-tier
model availability (Pro / Pro+ / Enterprise / evaluation models).
The Copilot CLI puts itself into ACP server mode via the `--acp`
flag instead of an `acp` subcommand, so acpDiscoveryProvider now
takes an optional acpArgs override.
Copilot's ACP payload omits the vendor name, so a small
prefix-based inferCopilotProvider keeps the UI's openai /
anthropic / google grouping working.
When the binary is missing or auth fails, fall back to
copilotStaticModels() so self-hosted runtimes without a copilot
install still see a populated dropdown.
Verified against `copilot 1.0.44`: live discovery returns 13
models with gpt-5.5 marked Default. Closes MUL-1948.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: multica-agent <github@multica.ai>
* fix(agent): drop no-op COPILOT_ALLOW_ALL env and generalize OpenAI o-series prefix check
- discoverCopilotModels: remove COPILOT_ALLOW_ALL=1 (not a real
Copilot CLI env var; copy-pasta from HERMES_YOLO_MODE=1).
Discovery only drives initialize + session/new which never
trigger tool-permission prompts, so no extra env is needed.
- inferCopilotProvider: replace the o1/o3/o4 prefix chain with a
generic o<digit>+ check via isOpenAIReasoningSeriesID, so future
o5/o6/… reasoning models are tagged as openai automatically.
Guards against false positives like 'opus-…' or bare 'o'.
- Extend TestInferCopilotProvider with o5/o6 forward-compat cases
and negative cases (opus-fake, omni, o).
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: multica-agent <github@multica.ai>
---------
Co-authored-by: Eve <eve@multica-ai.local>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: multica-agent <github@multica.ai>
`hermes`, `kimi`, and `kiro` all wired stderr through
`cmd.Stderr = io.MultiWriter(logWriter, providerErrSniffer)`.
The OS-pipe → MultiWriter copy goroutine that exec spawns for
that form is only joined by `cmd.Wait()`, which the lifecycle
goroutine fires in deferred cleanup — *after*
`promoteACPResultOnProviderError` already consulted the sniffer.
When stopReason=end_turn (success) raced ahead of the stderr
drain, the sniffer's `lines` slice was empty, the helper fell
through to the synthetic agent-text fallback ("hermes provider
error: API call failed after 3 retries"), and the actionable
upstream signal (HTTP 429 / usage limit) was lost.
This was visible as a flaky
`TestHermesBackendPromotesProviderErrorWithNonEmptyOutput` in CI
under high parallelism — a real prod bug, not a test issue: live
runs hit the same race when an upstream LLM returns 429 and
hermes' synthetic agent turn beats the stderr drain to the
parent.
Replace the MultiWriter wiring with `cmd.StderrPipe()` + an
explicit copier goroutine that signals on `stderrDone`. The
lifecycle goroutine already awaits `<-readerDone` for stdout;
add `<-stderrDone` next to it before `promoteACPResultOnProviderError`
runs. The deferred `cmd.Wait()` ordering is unchanged — it just
becomes a cheap reap by the time it fires.
Verified: `go test ./pkg/agent/ -run "TestHermes|TestKimi|TestKiro"
-count=10 -race`, then full package `-count=3 -race`, all green.
Co-authored-by: multica-agent <github@multica.ai>
ACP backends (Kiro, Hermes, Kimi) put the actionable reason for
code=-32603 'Internal error' in the JSON-RPC `data` field, e.g.
"No session found with id". The wrapped Go error only carried
`code` and `message`, leaving operators staring at a bare
"kiro session/prompt failed: session/prompt: Internal error
(code=-32603)" with no way to tell apart session expiry, model
unavailability, lost auth, or quota.
Parse `data` too. Strings render unquoted; objects/arrays render
as raw JSON; null/missing keeps the previous format unchanged.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* fix(daemon): mark provider 429 / out-of-credit runs as failed, not completed
Two bugs combined to silently report failed agent runs as
"Completed" in the UI when the upstream LLM returned a 4xx (e.g.
HTTP 429 rate-limit / no credit on the account).
1. ACP backends (hermes, kimi, kiro) only promoted the run status to
"failed" when their stderr sniffer fired AND the agent output
buffer was empty. But hermes injects a synthetic agent text turn
("API call failed after 3 retries: HTTP 429...") on retry
exhaustion, so the buffer was never empty in the rate-limit
case and the promotion never ran. Drop the empty-output
precondition: the sniffer's regex (HTTP-status markers, named
error types) is specific enough to trust on its own.
2. The daemon's task-result switch only routed "blocked" through
FailTask; every other status — including "cancelled", and any
future status we forget to enumerate — fell through to
CompleteTask. Invert it so only an explicit "completed" status
reports success, and extract the switch into reportTaskResult
for direct testing. Cancelled now defaults to failure_reason
"cancelled" instead of being silently completed.
Closes GitHub multica#1952.
Co-authored-by: multica-agent <github@multica.ai>
* fix(agent): only promote ACP run to failed on terminal provider error
Address GPT-Boy's review on the multica#1952 fix. The previous
promotion rule ("any sniffer line → fail") was too broad: the
existing sniffer also captures transient per-attempt warnings
("API call failed (attempt 1/3): RateLimitError [HTTP 429]"), and
those lines stay in the buffer for the rest of the run. A retry
sequence whose first attempt blipped but whose third attempt
succeeded would have been wrongly reported as failed.
Tighten the criteria with two additional signals, both defined on
the existing acpProviderErrorSniffer / output buffer:
- acpTerminalErrorRe — sticky `terminal` flag set when stderr shows
an exhausted/non-retryable marker (❌, [ERROR], "after N retries",
Non-retryable, BadRequestError, AuthenticationError). Per-attempt
warnings deliberately don't match.
- acpAgentOutputTerminalRe — matches the synthetic "API call failed
after N retries..." turn that hermes-style adapters inject into
the agent text stream when they give up; this catches multica#1952
even if hermes' stderr only logged transient attempts.
Promotion logic becomes a shared helper, promoteACPResultOnProviderError,
called from hermes / kimi / kiro. Promotes when (a) terminalMessage
is non-empty, (b) output contains the synthetic give-up turn, or
(c) output is empty and the sniffer captured anything at all
(preserves the original empty-output safety net for transient-only
sequences with no real result to fall back on).
Tests:
- TestHermesProviderErrorSnifferTerminalVsTransient — transient
attempt 1/3 alone returns terminalMessage="" but message!="";
a follow-on terminal marker flips terminal on.
- TestHermesProviderErrorSnifferTerminalNonRetryable — confirms
BadRequest / Authentication / Non-retryable / ❌ / [ERROR] are
classified terminal even on the very first attempt.
- TestHermesBackendDoesNotPromoteOnTransientRetry — fake hermes
emits attempt 1/3 to stderr then a normal agent text turn and
end_turn; resulting Status must stay "completed".
Co-authored-by: multica-agent <github@multica.ai>
---------
Co-authored-by: multica-agent <github@multica.ai>
PR #2281 added table-format support to parsePiModels but kept the
unconditional `strings.Replace(":", "/", 1)`, which would silently
rewrite a `:` inside a model name read from column 1 of the table
output (e.g. `claude-sonnet-4-6:exp` would become
`claude-sonnet-4-6/exp`). Move the replace into the legacy
`provider:model` branch so only the colon-as-separator case is
normalized, and restore a short doc comment describing the dual-
format contract. Test extended with a colon-bearing table row.
Co-authored-by: multica-agent <github@multica.ai>
The pi CLI changed its --list-models output from a single-field
'provider:model' format to a multi-column table with separate
'provider' and 'model' columns. The existing parser only looked
at the first whitespace-delimited field (the provider name) and
skipped lines without ':' or '/' — discarding every model entry.
Update parsePiModels to handle both formats:
- New table format: combine fields[0] (provider) + fields[1] (model)
- Legacy format: single field with ':' or '/' separator
Add regression test for the table format using real pi output.
PR #2101 swapped the openclaw runtime adapter from reading --json on
stderr to stdout. That fixed openclaw 2026.5+ but inverted the breakage
for pre-2026.5 builds — those still write JSON to stderr, so the
adapter now sees an empty stdout and falls through to the same
"openclaw returned no parseable output" failure that 2026.5+ users
saw before #2101.
Add a per-task version gate inside openclawBackend.Execute that runs
`openclaw --version`, parses the dotted version, and rejects anything
below 2026.5.5 with a hardcoded upgrade hint:
openclaw <detected> is below the minimum supported version 2026.5.5.
Run `openclaw update` to upgrade and try again.
The check is intentionally per-task and uncached so users who upgrade
do not need to restart the daemon — the next task automatically
re-checks. ~20ms per task is negligible vs. the typical run.
Co-authored-by: multica-agent <github@multica.ai>
Multica's openclaw runtime adapter has been reading agent output from
stderr since the early openclaw integration days. Current openclaw
(2026.5.5, c37871e) writes its --json blob exclusively to stdout:
$ openclaw agent --local --json --agent main --message 'say hi' >stdout 2>stderr
STDOUT bytes: 27401
STDERR bytes: 0
Result: every successful turn was followed by a daemon-generated system
comment 'openclaw returned no parseable output', visible to users,
looked like the agent broke when it didn't. Reproduced live on WOR-2,
turn at 2026-05-05 16:35 UTC; daemon log confirmed the full result JSON
arrived on the [openclaw:stdout] debug channel and was discarded while
the empty stderr pipe hit the no-events fallback.
Changes
- server/pkg/agent/openclaw.go: swap pipes, StdoutPipe() for the JSON
stream, cmd.Stderr = newLogWriter(...) for log overflow. Cleanup
goroutine now closes stdout on cancel. Comments and the read-error
errMsg updated to reflect the new pipe.
- server/pkg/agent/openclaw_test.go: TestOpenclawProcessOutputReadError
asserts on 'read stdout' (was 'read stderr'), string-only fix,
no behavior change. New TestOpenclawProcessOutputStdoutFixture feeds
a recorded openclaw 2026.5.5 --json blob through processOutput and
asserts result + messages parse cleanly.
- server/pkg/agent/testdata/openclaw-2026.5.5-stdout.json: 27401-byte
fixture captured fresh from the openclaw CLI for the regression test.
Side effects (net positive)
- Log lines openclaw writes to stderr (security warnings, tool errors)
now show up under [openclaw:stderr] instead of being silently consumed
by the JSON parser.
- Daemon's success_pattern heuristic (empty-output -> 'blocked')
becomes meaningful again because result.Output actually populates.
Closes WOR-10.
When the local state.db of an ACP backend (hermes, kimi, kiro) is wiped
— crash, config change, manual kill, container reset — the backend's
session/resume (or session/load, in kiro's case) silently creates a
brand-new session rather than failing, and returns the new id in the
response. Today the daemon ignores the response and stamps
sessionID = opts.ResumeSessionID across all three backends, so every
subsequent session/prompt is addressed to a session id the backend has
no record of. The task fails with JSON-RPC -32603 (Internal error) on
the very first turn, with no operator-visible signal that the problem
is a session-id mismatch one layer down.
The behavior is invisible: agent shows "started", then "failed" with a
generic Internal error. Reproducing in production took repeated runs
because nothing in the logs pointed at the silent reset.
Fix: route all three ACP backends through a small `resolveResumedSessionID`
helper that:
- prefers the id the backend returned in its response (the canonical
id; the one the backend will accept on the next call)
- falls back to the requested id when the response is malformed,
empty, or omits sessionId — defensive fallback so older / non-
conforming backends (notably kiro's current session/load shape)
behave identically to today
- signals (via a bool) when the id changed, so the caller logs a Warn
with `backend=<hermes|kimi|kiro>` and operators can grep for silent
state resets to correlate them with task failures
Why this is at the backend layer rather than the daemon's existing
session-resume fallback: server/internal/daemon/daemon.go:1554-1566
already retries with a fresh session when resume fails, but it gates
on `result.Status == "failed" && result.SessionID == ""`. The backend
WILL hand back a result.SessionID — just the new one it silently
committed to — so the daemon-level fallback never fires for this
failure mode.
The helper is also what session/new already uses (extractACPSessionID,
documented in code as "Shared by all ACP backends"). session/new
extracts the canonical id from the response; session/resume just
didn't, until now.
Coverage:
- hermes.go: confirmed bug, root cause of -32603 in production
- kimi.go: same code shape, same protocol method, same response
schema as hermes (per extractACPSessionID's comment) — same bug
- kiro.go: same code shape, different method (session/load). Current
observed response doesn't include sessionId, so the defensive
fallback means today's behavior is preserved. Routing through the
same helper means a future kiro release that DOES return a sessionId
on silent reset works the same way as hermes/kimi without another
diff.
Tests (server/pkg/agent/hermes_test.go — helper covers all three
backends, no per-backend duplication):
- TestResolveResumedSessionIDMatching — backend confirms requested id
- TestResolveResumedSessionIDDifferent — backend returned a new id;
caller is told to switch
- TestResolveResumedSessionIDEmptyResponse — older / malformed body;
defensive fallback to requested id (covers kiro's current shape)
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* refactor(quick-create): remove daemon CLI version gate
Local-source daemons report dev-suffixed versions (e.g.
v0.2.15-235-gdaf0e935) that the picker pre-check and server gate both
treat as too old, blocking quick-create during local testing.
Drops the gate end-to-end: removes MinQuickCreateCLIVersion +
CheckMinCLIVersion in pkg/agent, the checkQuickCreateDaemonVersion
handler and readRuntimeCLIVersion helper in handler/issue.go, and the
mirrored cli-version.ts plus the modal's pre-check, blocked-state UI,
and daemon_version_unsupported error branch.
Co-authored-by: multica-agent <github@multica.ai>
* refactor(quick-create): skip daemon CLI version gate in dev
Restores the gate (reverts the full-removal commit) and bypasses it in
non-production environments instead. The motivation for the original
removal — local source-built daemons report a `git describe` version
like v0.2.15-N-gHASH that parses below 0.2.20 and blocks dev testing —
is now handled by checking APP_ENV on the server and NODE_ENV on the
client. Production keeps the original "needs upgrade" UX.
Co-authored-by: multica-agent <github@multica.ai>
* refactor(quick-create): exempt git-describe daemons instead of env bypass
Replaces the per-environment bypass added in the previous commit with a
shared daemon-version signal. CheckMinCLIVersion / checkQuickCreateCliVersion
now treat any daemon whose CLI version matches the
`vX.Y.Z-N-gHASH[-dirty]` git-describe shape as OK; tagged releases keep
going through the normal min-version comparison.
Why: Emacs flagged that (a) NODE_ENV !== "production" also disables the
gate on staging and other non-prod deployments, undoing the protection
for the case the gate was originally written for, and (b) NODE_ENV (web
client) and APP_ENV (server) are not equivalent, so the modal pre-check
and server gate could disagree on the same request. Both go away when
the signal is intrinsic to the daemon's version string.
Co-authored-by: multica-agent <github@multica.ai>
---------
Co-authored-by: multica-agent <github@multica.ai>
* fix(codex): handle MCP elicitation server requests correctly
Fixes#1942.
handleServerRequest responded with {} to unrecognized Codex server
requests including mcpServer/elicitation/request. Codex 0.125+ expects
{action, content, _meta} for elicitation — the empty object causes a
deserialization error and the MCP tool call is reported as user-rejected.
Changes:
- Add mcpServer/elicitation/request case with correct response schema
- Add respondError helper for JSON-RPC error responses
- Return proper JSON-RPC method-not-found error for unknown server
requests instead of silent empty object
- Add tests for MCP elicitation and unknown method handling
* fix: use cfg.Logger instead of global slog in codex handleServerRequest
Switch the unhandled-server-request warning from global slog.Warn to
c.cfg.Logger.Warn for consistency with all other log calls in codex.go.
This ensures the warning appears in daemon run-logs and per-task
pipelines where operators look during triage.
Hermes ACP can flush queued session updates from the previous turn
before the current turn actually starts — both as session/resume
history replay and as chunks queued before our session/prompt response
streams. Without a gate those updates were appended to output and
re-emitted to the UI, so the previous answer appeared duplicated next
to the new one. Closes#1997.
PR #1789 added the acceptNotification hook field to hermesClient and
the call site in handleNotification, but never assigned it for Hermes,
so the guard short-circuited and every notification was processed.
This change mirrors the working Kiro pattern (kiro.go:87/97/240):
- declare a streamingCurrentTurn atomic.Bool in the backend.
- assign acceptNotification, onMessage, onPromptDone gates that all
return early when the flag is false.
- flip the flag to true immediately before c.request("session/prompt").
Adds TestHermesClientAcceptNotificationGate as a regression test that
exercises the gate directly on hermesClient.
Verified with `go test ./pkg/agent`.
Co-authored-by: multica-agent <github@multica.ai>
Latest Codex CLI ships with GPT-5.5 / GPT-5.5 mini, but the static
catalog still topped out at GPT-5.4 so users couldn't pick the new
model from the agent picker.
Add gpt-5.5 + gpt-5.5-mini to codexStaticModels and promote 5.5 as
the default badge. Keep the older 5.4 / 5.3-codex / gpt-5 / o3
entries for users on older Codex CLI builds. Add a regression test
mirroring TestGeminiStaticModelsExposesAliasesAndGemini3 so the
next OpenAI release isn't a silent miss.
Co-authored-by: multica-agent <github@multica.ai>
* fix(quick-create): bound dialog height + scroll editor when content overflows
Pasting a screenshot into the agent-create prompt expanded the editor
unbounded, which dragged DialogContent past the viewport since the agent
mode className had no max-height. Manual mode was unaffected because
manualDialogContentClass pins `!h-96`.
- Cap agent-mode DialogContent at `!max-h-[80vh]` (width stays
`!max-w-xl`); short prompts still render compact, tall content stops
at 80% of the viewport.
- Switch the editor wrapper to `flex-1 min-h-[140px] overflow-y-auto`
so it absorbs the remaining vertical space inside the now-bounded
DialogContent and scrolls internally instead of pushing the dialog.
* feat(quick-create): gate on daemon CLI version with pre-check + server enforcement
The agent-create flow depends on multica CLI behavior introduced in
v0.2.20 (URL attachment handling, no-retry semantics on
`multica issue create` failure — see PR #1851 / MUL-1496). Older
daemons either double-create issues on partial CLI failures or
mishandle pasted screenshot URLs. Per J's review on MUL-1496, gate
the flow at two layers — frontend pre-check for fast feedback,
server re-check as the trust boundary, both fail-closed on
missing/unparsable versions.
Server:
- New MinQuickCreateCLIVersion + CheckMinCLIVersion helper in
pkg/agent (with sentinel errors for missing vs too-old).
- QuickCreateIssue handler reads runtime metadata.cli_version and
returns a stable 422 { code: "daemon_version_unsupported",
current_version, min_version, runtime_id } before enqueuing.
- The check runs after the existing online + ownership validation,
so all rejections surface uniformly through the modal's existing
error path.
Frontend:
- New @multica/core/runtimes/cli-version with the min version
constant, parser, and runtime-metadata reader (tiny semver, no
new lib dep).
- AgentCreatePanel resolves the selected agent's runtime, runs the
same check, shows an inline amber notice below the agent picker
when missing/too old, and disables the Create button.
- Submit handler also catches the server's 422 (defensive race —
runtime can re-register between pre-check and submit) and
surfaces the same wording in the error row.
Switching to manual create remains a clean escape hatch — manual
mode doesn't talk to a daemon at all, so an outdated CLI doesn't
block the user from filing the issue.
On Windows the official cursor-agent installer ships cursor-agent.cmd whose
body is `powershell ... -File cursor-agent.ps1 %*`. CreateProcess for a .cmd
file goes through cmd.exe, and `%*` in a batch file is expanded by
re-tokenising the original command line, which mangles arguments containing
newlines or other whitespace - most notably a long, multi-line `-p <prompt>`.
The agent then only sees a truncated prompt and fails with "Workspace Trust
Required" or exits 1 immediately.
When LookPath resolves cursor-agent to a .cmd/.bat launcher and a sibling
cursor-agent.ps1 exists, invoke PowerShell directly with `-File <ps1>` so
Go's os/exec passes each argv as a discrete token. This is exactly what the
.cmd does internally; we just skip the cmd.exe re-tokenisation step.
PowerShell host resolution prefers pwsh.exe (PS 7) on PATH, then
powershell.exe on PATH, and finally falls back to
%SystemRoot%\System32\WindowsPowerShell\v1.0.
Platform-specific code is split via build tags
(cursor_invocation_windows.go / cursor_invocation_other.go) so non-Windows
builds carry no Windows-only dependencies. The lookup is exposed as a
package variable to make the Windows path fully unit-testable without
spawning real PowerShell. Five unit tests cover: passthrough on non-launcher
targets, successful rewrite with a multi-line prompt, .exe direct launch
(skip), missing .ps1 (skip), and missing PowerShell host (skip).
The change leaves macOS / Linux behaviour entirely untouched and stays on
the official cursor-agent launch chain - no node.exe direct invocation, no
prompt mutation, no extra flags.
Closes#1297
Made-with: Cursor
Ignore Kiro ACP session/load history replay before the active prompt starts; keep task messages, usage, and tool state scoped to the current Kiro turn. Verified with go test ./pkg/agent -run TestKiro, go test ./pkg/agent, and git diff --check origin/main...HEAD.
* fix(agent/opencode): bypass npm .cmd shim on Windows to preserve multi-line prompts
The npm-generated `opencode.cmd` shim forwards argv via Windows batch `%*`,
which silently truncates positional arguments at the first newline. The
daemon spawns OpenCode with a multi-line prompt (system prompt + user
message), so on Windows the agent only ever sees the first line and
responds generically as if it never received the user's message
(reported in #1717 with native-binary repro confirming the same prompt
arrives intact when cmd.exe is skipped).
When `runtime.GOOS == "windows"` and `exec.LookPath` returns a `.cmd`
shim, walk to the native binary that npm bundles next to the shim:
<prefix>\opencode.cmd
<prefix>\node_modules\opencode-ai\node_modules\opencode-windows-x64\bin\opencode.exe
If the native binary is missing (unusual install layout), keep the
original shim path so PATH lookup still wins. The resolver is a pure
function with an injectable `statFn`, so layout assertions are testable
on Linux:
- shim resolves to the bundled native binary
- missing native returns "" (caller keeps original path)
- non-cmd paths (Linux/Mac binary, opencode.exe direct, empty) skip resolution
- uppercase `.CMD` is accepted (PATHEXT entries can be either case)
Closes the user-facing failure mode without restructuring exec resolution
across the rest of the agent backends — the other shim-aware fixes can
follow the same shape if/when they land in similar repros.
* fix(agent/opencode): cover x64-baseline and arm64 npm package variants
`npm install -g opencode-ai` ships three Windows platform packages
(opencode-windows-x64, opencode-windows-x64-baseline for older CPUs
without AVX2, opencode-windows-arm64 for Surface / Copilot+ PC) and
installs whichever matches the host. The previous resolver only knew
about opencode-windows-x64, so baseline-x64 and arm64 hosts would fall
back to the .cmd shim and hit the multi-line prompt truncation again.
Iterate the three package candidates in GOARCH-preferred order. ARM64
hosts try arm64 first; everything else tries x64, then baseline, then
arm64 as a last resort. Cost is one extra statFn call per miss when
the GOARCH-preferred package isn't installed.
Surfaced by review on #1718.
* test(agent): add Windows counterpart to writeTestExecutable
writeTestExecutable in exec_fixture_unix_test.go is referenced by
claude_test.go / codex_test.go / kimi_test.go, but the //go:build unix
constraint meant `go test ./pkg/agent` failed to build on Windows.
ETXTBSY is a Linux/Unix fork-exec race; Windows doesn't have that
pathology, so a plain os.WriteFile is sufficient.
Lifted from #1719 (Codex) with attribution. Surfaced by review on #1718.
#1674 wired claude's post-handshake error path through withAgentStderr but
left the writeClaudeInput failure branch returning a bare "broken pipe"
error. That branch fires precisely when claude crashes during startup —
exactly when the stderr tail is most useful for root-causing V8 aborts,
Bun panics, or missing native modules. cmd.Wait() before sampling Tail()
flushes os/exec's internal stderr copy goroutine, matching the
Wait→Tail synchronization contract spelled out in stderr_tail.go.
Adds TestClaudeExecuteSurfacesStderrWhenChildExitsEarly mirroring the
codex test: a fake claude binary drains stdin, writes a V8-abort line to
stderr, and exits 3. Locks in the contract that Result.Error carries the
stderr tail in the post-handshake failure path on the claude backend too.
Hoist the existing stderrTail ring-buffer (previously codex-only) into
a shared pkg/agent helper so every Backend that supervises a child CLI
can include the last ~2 KB of that CLI's stderr in Result.Error. Wire
the claude backend through the same path.
Motivation: claude on Windows occasionally exits with a non-zero status
after ~5–8 minutes of a single long-running tool_use, and right now the
daemon only reports "claude exited with error: exit status 3" /
"exit status 0x80000003" — useless for root-causing V8 aborts, Bun
panics, native-module OOMs, or any other CLI-side crash. With the tail
attached, the failure message carries the real signal (panic line, V8
assertion, stderr-printed HTTP error) all the way into the task row's
error field that users see in the API.
Renames withCodexStderr to withAgentStderr(msg, label, tail) so the
helper is self-documenting across providers.
Follow-up to #1686. Locks in two nits flagged during review:
1. agent.Result.Status doc comment now lists "cancelled" alongside the
existing values, so the enum surface matches actual usage.
2. New TestExecuteAndDrain_ContextCancelled_ReportsCancelled exercises
the path added in #1686: when the parent context is cancelled before
the backend produces a Result, executeAndDrain must return
Status="cancelled" (not "timeout"). A regression here would silently
restore the misleading log line we just fixed.
* fix(daemon): use CREATE_NEW_CONSOLE to stop grandchild console popups on Windows (#1521)
CREATE_NO_WINDOW strips the console entirely. When the agent CLI then
spawns a console-subsystem grandchild (bash, cmd, netstat, findstr,
timeout) without itself passing CREATE_NO_WINDOW, Windows allocates a
brand-new visible console window per invocation — trading one popup per
agent run for N popups per tool call.
Switch to CREATE_NEW_CONSOLE + HideWindow=true so the agent gets a
hidden console that grandchildren inherit. Stdio pipes still work via
STARTF_USESTDHANDLES; no changes needed at the 17 hideAgentWindow call
sites.
Add a Windows-only regression test asserting CREATE_NEW_CONSOLE is set
and CREATE_NO_WINDOW is not, per the #1474 Windows-test follow-up.
Root-cause diagnosis by @matrenitski (verified against the shipped
multica.exe and the Claude Code CLI it spawns) in issue #1521.
* test(agent): use CREATE_NEW_CONSOLE-compatible flag in preservation test
CREATE_NEW_PROCESS_GROUP is silently ignored by Windows when combined
with CREATE_NEW_CONSOLE, so asserting it 'survives' was only bitwise-true,
not semantically meaningful. Switch the example to
CREATE_UNICODE_ENVIRONMENT (documented compatible) and also assert a
non-flag field (NoInheritHandles) survives to exercise full struct
preservation.
* fix(daemon): suppress agent terminal windows on Windows (#1471)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix: add hideAgentWindow to detectCLIVersion and avoid SysProcAttr overwrite
- Add missing hideAgentWindow(cmd) call in detectCLIVersion (claude.go:554)
so --version checks don't flash console windows on Windows.
- Refactor hideAgentWindow to preserve existing SysProcAttr fields
instead of overwriting the entire struct.
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
* fix: pass model to Hermes ACP session/new and add hermes to InjectRuntimeConfig
- hermes.go: include opts.Model in session/new params so Hermes uses
the configured model instead of its default (fixes local LLM failures)
- runtime_config.go: add "hermes" to the AGENTS.md provider list so
Hermes receives the Multica runtime instructions and skill discovery
Fixes: https://github.com/multica-ai/multica/issues/1195
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* fix(hermes): drop false native-skill claim and add regression tests
The previous change added 'hermes' to the 'skills discovered automatically'
branch of buildMetaSkillContent, but resolveSkillsDir has no Hermes case so
skills still land in the .agent_context/skills/ fallback. AGENTS.md ended up
claiming native discovery while the files were somewhere else, which would
mislead Hermes (and future debuggers).
- Move 'hermes' to the fallback branch alongside 'gemini' so AGENTS.md points
Hermes at .agent_context/skills/ — matching where writeContextFiles actually
writes them.
- Extract buildHermesSessionParams so the session/new payload is unit-testable.
- Add regression tests covering:
* buildHermesSessionParams includes/omits 'model' correctly
* InjectRuntimeConfig('hermes') writes AGENTS.md with the fallback hint
* writeContextFiles('hermes') writes skills to .agent_context/skills/
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
---------
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: CC-Girl <cc-girl@multica.ai>
* fix(landing): scope landing route to always-light palette
The landing page sections use hardcoded light colors (bg-white / #0a0d12),
but shared components rendered inside — notably CloudWaitlistExpand on
/download — use semantic tokens that flip to dark values under next-themes'
`.dark` class, producing a mismatched dark card on an otherwise light page
when the user's OS is in dark mode.
Add a `.landing-light` class on the landing layout wrapper that re-declares
all color tokens to their light values for the subtree, so nested
token-driven components stay in lockstep with the hardcoded palette.
* test(agent): serialize fake-executable writes to avoid ETXTBSY on CI
TestKimiBackendInvokesACPSubcommand (and its Kimi/Codex siblings) write a
shell script to a per-test TempDir and then fork/exec it. With t.Parallel()
enabled across the package, a concurrent goroutine's fork can inherit the
still-open write fd to another test's new executable; Linux then rejects
the subsequent exec with ETXTBSY (seen as
fork/exec /tmp/.../kimi: text file busy
on GitHub Actions).
Introduce writeTestExecutable, which holds syscall.ForkLock.RLock across
OpenFile→Write→Close. Fork (which takes ForkLock.Lock) cannot run while we
hold RLock, so no sibling fork inherits our write fd. Ran the three callers
with -count=10 under -p=1 and the full package with no failures.
'discoverOpenclawAgents' runs several 'openclaw' subprocesses under one
context; 5s was too short on cold starts or under load, causing empty
listings in the model picker. Increase the per-discovery cap to 30s.
Gemini CLI has no `models list` subcommand, so Multica can't do real
dynamic discovery. Instead, swap the static catalog from fixed version
names (2.0/2.5 only) to the CLI's own aliases (`auto`, `pro`, `flash`,
`flash-lite`, `auto-gemini-2.5`) plus explicit pins for Gemini 3
preview and 2.5 variants. Aliases are resolved inside the Gemini CLI
per user entitlement + quota, so new model releases light up without
a Multica redeploy. Default is `auto`, matching Google's recommended
selection.
Fixesmultica-ai/multica#1503.
* feat(server): orphan-task recovery + auto-retry + manual rerun (MUL-1128)
When the daemon process crashed mid-task the issue was stuck at
in_progress for up to 2.5h: the in-flight task timeout was the only
mechanism that ever moved the row, and the runtime heartbeat sweeper
only fires after the runtime stays offline for 45s — a quick restart
beats both windows.
This change implements the A+B plan from the issue thread:
A. lifecycle hygiene
- migration 055 adds attempt / max_attempts / parent_task_id /
failure_reason / last_heartbeat_at to agent_task_queue
- new daemon-auth endpoint POST /runtimes/{id}/recover-orphans:
daemon calls it on every register so the server fails any
dispatched/running tasks the previous process left behind
- new daemon-auth endpoint POST /tasks/{id}/session: persists the
agent's session_id + work_dir mid-flight so a crash doesn't
lose the resume pointer (claude+codex emit MessageStatus with
SessionID; daemon forwards on the first one it sees)
- FailAgentTask / FailStaleTasks / FailTasksForOfflineRuntimes
now set failure_reason ('agent_error' / 'timeout' /
'runtime_offline')
B. auto-retry with resume context
- TaskService.MaybeRetryFailedTask spawns a fresh queued attempt
carrying parent's session_id/work_dir when the failure reason
is infrastructure-shaped (timeout, runtime_offline,
runtime_recovery) and attempt < max_attempts; skips autopilot
- wired into the runtime sweeper paths and TaskService.FailTask
so the user transparently sees a new in_progress run instead of
a stuck row
- new user-auth POST /api/issues/{id}/rerun + multica issue rerun
CLI for the manual escape hatch
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* fix(server): address PR review for orphan-task recovery (MUL-1128)
Three review-must-fix items on top of the A+B implementation:
1. recover-orphans now funnels through TaskService.HandleFailedTasks,
the same shared post-failure pipeline used by the runtime sweeper.
This guarantees task:failed events are emitted, agent status is
reconciled, and issues stuck in_progress with no remaining active
task are reset to todo even when no auto-retry is created
(max_attempts exhausted, autopilot, non-retryable reason).
2. RerunIssue now uses CancelAgentTasksByIssueAndAgent, scoped to the
issue's current assignee. The previous implementation called
CancelAgentTasksByIssue, which would collateral-cancel parallel
@-mention agents on the same issue.
3. GetLastTaskSession now considers both completed and failed tasks
(mirroring GetLastChatTaskSession), ordering by the most recent
timestamp. With UpdateAgentTaskSession pinning session_id/work_dir
mid-flight, an auto-retry or manual rerun of a daemon-crash failure
now actually resumes the prior conversation context instead of
starting fresh — matching the stated B-branch behaviour.
go build / go vet pass; the existing service and agent test suites pass.
runtime_sweeper / handler integration tests require a local DB with the
055 migration (and the pre-existing 050 first_executed_at column).
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
---------
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
OpenClaw's `--json` result blob carries the actual LLM identifier in
`meta.agentMeta.model` (e.g. `deepseek-chat`, `claude-sonnet-4`),
alongside `provider` and the usage breakdown. The backend was reading
the surrounding `agentMeta.usage` and `agentMeta.sessionId` but skipping
the `model` field entirely, then attributing every run's tokens to
`opts.Model` — which for openclaw is the *agent name* passed via
`--agent`, not a real model identifier — falling all the way through to
"unknown" when no agent.model was configured.
Surface the runtime-reported model:
- `openclawEventResult` gains a `model` string.
- `buildOpenclawEventResult` reads `agentMeta.model` (trimmed; empty
string when absent for forward-compat with older runtimes / partial
outputs).
- `processOutput` propagates it through the result-blob branch.
- `Execute`'s usage map prefers `scanResult.model`, falling back to
`opts.Model` then `"unknown"` — preserving the prior behavior path
for any runtime that doesn't surface its own model yet.
Two unit tests cover both the populated and missing cases.
Refs: #1395
Surface the actual exec path + argv for every agent backend at INFO
so operators can see the exact command without flipping to debug.
Also add the missing log line in pi.go for consistency with the
other nine backends.
* feat(agent): add Kimi CLI as agent runtime
Adds support for Moonshot AI's Kimi Code CLI (https://github.com/MoonshotAI/kimi-cli)
as a new agent runtime, alongside Claude, Codex, OpenCode, OpenClaw, Hermes,
Gemini, Pi, Cursor and Copilot.
Kimi Code CLI implements the standard Agent Client Protocol (ACP) via the
`kimi acp` subcommand, so the new `kimiBackend` reuses the existing
hermesClient JSON-RPC transport in the agent package — only the binary,
client identity, log prefix, and tool-name extraction differ.
Wiring:
- server/pkg/agent: new kimiBackend + kimi_test.go; registered in New(),
LaunchHeader map, and the supported-types coverage test.
- server/internal/daemon/config.go: probes `kimi` (overridable via
MULTICA_KIMI_PATH / MULTICA_KIMI_MODEL).
- server/internal/daemon/execenv: writes AGENTS.md as the runtime context
file (Kimi reads AGENTS.md natively via /init), and writes skills under
`.kimi/skills/` so they are auto-discovered by the project-level skill
loader.
- packages/views/runtimes: ProviderLogo gains a Kimi mark.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* feat(agent/kimi): support per-agent model selection via ACP set_model
Wire Kimi into the model dropdown introduced in #1399:
- ListModels gets a 'kimi' case that drives the same ACP
initialize + session/new handshake as Hermes; both share a new
discoverACPModels helper and parseACPSessionNewModels parser
so future ACP backends only need a small provider entry.
- kimiBackend now issues session/set_model after session/new when
opts.Model is non-empty, mirroring the Hermes flow. Failures
fail the task instead of silently falling back to Kimi's
default model — silent fallback would hide that the dropdown
pick wasn't honoured.
Verified: go build ./..., go test ./pkg/agent/... ./internal/daemon/... ./internal/handler/..., pnpm typecheck and pnpm test (138 passed).
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* refactor(agent): address code review feedback on Kimi runtime
- Share ACP provider-error sniffer between hermes and kimi. Previously
only hermes promoted stderr-observed 4xx/5xx into a failed task;
kimi would report "completed + empty output" when the Moonshot
upstream rejected a request (expired token, rate limit, …). Rename
hermesProviderErrorSniffer → acpProviderErrorSniffer and parameterise
the provider name; wire it into kimiBackend.Execute the same way.
- Rename extractHermesSessionID → extractACPSessionID (shared by all
ACP backends) so the name matches parseACPSessionNewModels.
- Drop the redundant second argument to kimiToolNameFromTitle; the
Message struct has only one relevant field (Tool), so passing it
twice was a dead fallback. Document that the function normalises
residual capitalised kimi titles not caught by hermesToolNameFromTitle.
- Remove kimi-only cmd.WaitDelay override; the hermes baseline is
fine for both and divergence adds noise.
- Add TestKimiBackendSetModelFailureFailsTask: fake `kimi acp` binary
that returns a JSON-RPC error for session/set_model, asserts that
the task result surfaces status=failed with the model name + upstream
message and preserves the session id.
- Fix stale agent listings in agent.go / daemon/config.go doc comments
(missing cursor, gemini, copilot).
All: `go build ./...`, `go vet ./...`, `go test ./pkg/agent/...
./internal/daemon/... ./internal/handler/...` green.
* fix(agent/kimi): pass --yolo so Shell tools don't hang on approval
Kimi's default config has `default_yolo = false`. Every Shell/file-mutating
tool call causes kimi acp to send a `session/request_permission` request
and block (up to 300s) waiting for a response. The daemon's hermesClient
only handles `session/update` notifications — permission requests go
unanswered, the tool call times out, and the UI loop eventually dies
("UI loop timed out"). Observed with the first real kimi task: agent sat
as Live for ~7 minutes before the daemon killed it.
The fix mirrors hermes' HERMES_YOLO_MODE=1 override: pass `--yolo` to
`kimi` so it auto-approves everything. `--yolo` is a top-level flag on
the `kimi` CLI (not a flag on `kimi acp`), so it must come before the
`acp` subcommand in argv. Added to kimiBlockedArgs so user custom_args
can't strip it.
While here, fix a related bug that made kimi tool names show up empty
in the daemon log ("tool #1: "): hermesToolNameFromTitle's fallback
returned `kind` when neither title-with-colon nor kind matched a known
tool. Kimi's ACP `tool_call` emits bare titles like "Shell" or "Read
file" with no `kind` at all, so we'd drop the title on the floor before
kimiToolNameFromTitle ever got a chance to map it. Now: preserve the
title when kind is unclassified; hermes titles always carry a colon so
this branch never fires for hermes.
Tests:
- TestKimiBackendPassesYoloFlag — fake binary that records its argv,
asserts --yolo comes before acp.
- TestHermesToolNameFromTitle rows for bare kimi-style titles.
- Existing suite green: go build, go vet, full pkg/agent + daemon +
handler test packages.
* fix(agent/acp): auto-approve session/request_permission from agent
The previous attempt (`kimi --yolo acp`) was a no-op. Inspected the
kimi-cli source: the `acp` Typer subcommand takes no parameters, so
flags on the root `kimi` command are dropped before `acp_main()` runs
— it's impossible to opt into YOLO mode through CLI flags for ACP.
The real fix is on our side: respond to session/request_permission.
ACP is bidirectional. When kimi runs a Shell or file-write tool, it
sends `session/request_permission` (agent → client, JSON-RPC request
with id + method) and waits up to 300s for a response. Our existing
hermesClient.handleLine only dispatched: (id + result/error) →
handleResponse, and (no id + method) → handleNotification. A request
with BOTH id and method fell through and got silently dropped — kimi
timed out, UI loop died, task sat stuck for 7 minutes.
Add handleAgentRequest: for session/request_permission, echo the id
and respond with outcome=selected, optionId=approve_for_session. The
daemon is headless; there's no user to prompt. `approve_for_session`
lets the agent remember the action so subsequent identical calls
(every Shell, every file write) skip the round-trip entirely. For any
other agent → client method, reply with standard -32601 method-not-
found so the agent doesn't block.
Also:
- Add writeMu so request() (main goroutine) and handleAgentRequest
(reader goroutine) don't interleave JSON frames on stdin.
- Revert the `--yolo acp` flag — it's a no-op, and carrying it in
kimiBlockedArgs gives the wrong impression that it does something.
Comment in kimi.go now points at handleAgentRequest as the real fix.
Tests:
- TestHermesClientAutoApprovesPermissionRequest: inject a
session/request_permission, assert the reply echoes the id and
carries {outcome: selected, optionId: approve_for_session}.
- TestHermesClientReplesMethodNotFoundForUnknownAgentRequest: confirm
unknown agent → client methods get JSON-RPC -32601 instead of silence.
- TestKimiBackendInvokesACPSubcommand replaces the yolo-flag assertion
with a negative assertion: no dead --yolo / --auto-approve / -y on
argv, since they'd pretend to do something they can't.
All: go build ./..., go vet ./..., go test ./pkg/agent/... green.
* fix(agent/acp): surface kimi tool input/output via content blocks
Kimi-cli emits tool_call and tool_call_update ACP frames with the
input/output inside a `content` array of ContentToolCallContent
blocks (shape: {type:"content", content:{type:"text", text:"..."}}),
not in the hermes-style `rawInput` map / `rawOutput` string. Our
parser only looked at rawInput/rawOutput, so the daemon recorded
empty Input and Output for every kimi tool — the execution-history
UI showed blank terminal panels even for commands that ran fine.
Add extractACPToolCallText() and a fallback in handleToolCallStart /
handleToolCallUpdate: when rawInput is nil / rawOutput is empty, pull
the text out of the content blocks. rawInput / rawOutput still take
precedence so hermes' behaviour is untouched. Terminal /
FileEditToolCallContent blocks are skipped (we have nothing to render
them as — kimi only emits TerminalToolCallContent when the client
advertises terminal capability, which we don't).
Tests:
- TestHermesClientHandleToolCallStartKimiContent — content array →
Input.text populated.
- TestHermesClientHandleToolCallCompleteKimiContent — multi-block
content → Output concatenated with newline separator.
- TestHermesClientHandleToolCallRawOutputTakesPrecedence — hermes
rawOutput still wins when both are present.
- TestExtractACPToolCallText — unit coverage for the helper
(single/multiple text blocks, terminal-block skip, empty input).
* fix(agent/acp): buffer streaming tool args so Input isn't empty in UI
kimi-cli streams tool args token-by-token via tool_call_update frames
— the initial tool_call carries an empty content block and each
subsequent in_progress update carries the cumulative JSON so far
(`{`, `{"comma`, `{"command": "echo`, …). The final completed update
then carries the tool's stdout, not the args. Observed per kimi-cli
acp/session.py::_send_tool_call{,_part,_result} and confirmed by
driving a real Shell call end-to-end: 10 in_progress frames, last
with `{"command": "echo hello world"}`, then completed with `hello
world\n`.
Our previous handleToolCallStart emitted MessageToolUse on the first
tool_call frame, capturing the empty content — so every kimi tool
appeared in the execution-history UI with a blank input. Output was
correct (fix 4335c198) but command was missing.
Changes:
- hermesClient now tracks pending tool calls per toolCallId. Hermes
path is unchanged — rawInput is present at tool_call time, so
emit-immediately-then-flag-emitted still fires on the initial frame.
- kimi path defers MessageToolUse until status=completed / failed.
tool_call_update in_progress frames update the buffered argsText
(cumulative, so overwrite); on completion we parse the accumulated
JSON into Message.Input. Malformed JSON falls back to `{"text": …}`
so non-JSON tool args still render.
- Orphan completion frames (no matching tool_call seen — e.g. daemon
restarted mid-task) synthesise ToolUse from the update's own
title/kind/rawInput so the UI still gets a header.
- extractACPToolCallText now also renders FileEditToolCallContent
blocks as a compact header ("--- path / +++ path / (edited: N → M
bytes)"). kimi emits these for Write / StrReplaceFile / Patch when
the tool's display block is a DiffDisplayBlock.
Tests:
- TestHermesClientKimiStreamingToolCall: empty tool_call + 5 streaming
in_progress + completed. Asserts no emission until complete, then
[ToolUse(Input.command="echo hi"), ToolResult(Output="hi\n")].
- TestHermesClientKimiMalformedArgsFallback: non-JSON argsText → falls
back to Input.text.
- TestHermesClientHandleToolCallCompleteOrphan: completed frame
without a start → ToolUse synthesised from update's rawInput.
- TestExtractACPToolCallText: diff + new-file-diff cases.
All agent / daemon / handler test packages green.
---------
Co-authored-by: Eve <8b0578a3-cf72-4394-9e38-b328eca92463@users.noreply.multica.ai>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Eve <eve@multica.ai>
Co-authored-by: Lambda <f252c2c5-7d1d-4f3c-b394-a61abfe673fc@users.noreply.multica.ai>
Adds a first-class `model` field on agents so users can pick the LLM model from the create / settings UI instead of editing `custom_env` / `custom_args`. Each provider's dropdown is populated from the live CLI when possible (`opencode models`, `pi --list-models`, `openclaw agents list --json`, `cursor-agent --list-models`, hermes ACP `session/new` → `SessionModelState`), with a static catalog for providers that don't enumerate.
Daemon resolves the runtime model as `agent.model → MULTICA_<PROVIDER>_MODEL → ""` — empty passes through so each backend's CLI picks its own default, avoiding static-guess drift.
Per-provider honouring:
- Claude / Codex / OpenCode / Cursor / Gemini / Pi / Copilot — CLI `--model` / thread payload.
- OpenClaw — `opts.Model` is mapped to `--agent <name>` (the CLI rejects `--model`).
- Hermes — `session/set_model` ACP RPC; stderr is sniffed for provider-level errors so HTTP 4xx from the configured LLM surfaces instead of "empty output"; explicit-model failures mark the task `failed`.
Supporting changes: migration 050 adds `agent.model`; daemon ↔ server heartbeat piggyback carries a model-discovery request; new REST endpoints under `/api/runtimes/{id}/models`; `multica agent create --model` / `update --model`; shared `ModelDropdown` in `packages/views/agents` (searchable, creatable, provider-grouped, default-badge, runtime-supported gate).
* fix(agent/codex): surface stderr tail in initialize / turn startup errors
When codex app-server exits before the JSON-RPC handshake completes —
e.g. because the user put a flag in custom_args that the subcommand
rejects — the Result.Error users see is `codex initialize failed:
codex process exited`, while codex's actual complaint (typically
something like `error: unexpected argument '-m' found`) only lives in
daemon logs.
Wrap the stderr writer with a bounded stderrTail that still forwards
to the slog logWriter but also retains the last 2 KiB of bytes
written. Include that tail on the three startup failure paths
(initialize, startOrResumeThread, turn/start). Runtime cancellation
paths are left untouched — they're our own abort and the stderr
context isn't a clear signal there.
Refs #1308. Complement to #1310 / #1312 — lets "bad custom_args fail
loudly" actually be workable by giving the failure a real message.
* fix(agent/codex): join cmd.Wait() before sampling stderr tail
Addressing review of #1314: reading stderrBuf.Tail() right after
c.request returns "codex process exited" was racy. Nothing in that
path synchronizes with os/exec's internal stderr copy goroutine —
cmd.Wait() is the only documented join point. The original defer ran
cmd.Wait() later, but by then we had already built Result.Error from
a potentially-empty Tail().
Replace the ad-hoc deferred stdin.Close()/cmd.Wait() with a
sync.Once-wrapped drainAndWait closure. Call it explicitly on the
three startup failure paths before sampling the tail; keep it as the
cleanup defer so the success path behaves identically.
Also add TestCodexExecuteSurfacesStderrWhenChildExitsEarly: spawns a
real subprocess that prints to stderr and exits before responding to
initialize, runs it through Execute, and asserts Result.Error
contains the stderr hint. This covers the full timing path the
reviewer flagged, which the helper-level tests in this PR did not.
Fixes#1332.
Two regressions introduced in #910 (2026-04-14, "OpenClaw backend P0+P1
improvements") that together block all openclaw users:
1. `openclaw agent` does not accept `--model` or `--system-prompt`, so
any agent configured with a Model field crashed in ~700ms with
`exit status 1`. Remove both forwards, and add them to
openclawBlockedArgs so custom_args can't reintroduce the crash.
Model is bound at registration time via `openclaw agents
add/update --model`.
2. AgentInstructions were written to `{workDir}/AGENTS.md` by
execenv.InjectRuntimeConfig, but openclaw loads bootstrap files
from its own workspace dir — the file was never read, so every
agent's Instructions field was silently discarded. Populate
opts.SystemPrompt for the openclaw provider in runTask and
prepend it to the `--message` payload in the backend so the
model actually receives the instructions.
Other providers surface instructions through their native runtime
config file (CLAUDE.md / AGENTS.md / GEMINI.md) and are intentionally
left unchanged to avoid double injection.
Extract buildOpenclawArgs so arg construction is directly testable;
add unit tests covering the removed flags, the SystemPrompt prepend,
and custom_args filtering.
* fix(chat): preserve chat session resume pointer across failures
The chat 'forgets earlier messages' bug came from PriorSessionID being
silently lost in several edge cases:
- UpdateChatSessionSession unconditionally overwrote chat_session.session_id,
so any task that completed without a session_id (early agent crash,
missing result) wiped the resume pointer to NULL.
- CompleteAgentTask + UpdateChatSessionSession ran in separate calls. A
follow-up chat message claimed in between resumed against a stale (or
NULL) session and started over.
- FailAgentTask never wrote session_id back, so a task that established
a real session before failing lost its resume pointer.
- ClaimTaskByRuntime only trusted chat_session.session_id and never
fell back to the existing GetLastChatTaskSession query, so a single
bad turn could permanently drop the conversation memory.
This change:
- Use COALESCE in UpdateChatSessionSession so empty inputs preserve the
existing pointer; surface DB errors instead of swallowing them.
- Run CompleteAgentTask/FailAgentTask + UpdateChatSessionSession inside
the same transaction (TaskService now takes a TxStarter).
- Extend FailAgentTask + the daemon FailTask path (client, handler,
service) to forward session_id/work_dir, so failed/blocked tasks that
built a real session still record it.
- Fall back to GetLastChatTaskSession in ClaimTaskByRuntime when the
chat_session pointer is missing, and include failed tasks in that
lookup so a single failure can't lose the conversation.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* fix(daemon): forward session_id/work_dir on blocked + timeout paths
runTask previously dropped result.SessionID and env.WorkDir on the
non-completed return paths:
- timeout returned a naked error, so handleTask called FailTask with
empty session info and the chat resume pointer was either left stale
or eventually overwritten with NULL.
- blocked / failed (default branch) returned a TaskResult without
SessionID / WorkDir, so even though FailTask now COALESCEs into
chat_session, there was no value to write through.
- the empty-output completion path was the same: it raised an error
even when a real session_id had been built.
All three paths now return a TaskResult that carries the SessionID /
WorkDir the backend produced. Combined with the COALESCE-based update
in UpdateChatSessionSession and the FailTask plumbing introduced in
PR #1360, the next chat turn can always resume from the latest agent
session — even when the previous turn timed out, was rate-limited, or
returned an empty completion — instead of starting over with no memory
of the conversation.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* fix(copilot): capture session id from session.start as fallback
The Copilot backend only read sessionId from the synthetic 'result'
event, ignoring the one already present on session.start. When the CLI
was killed before result arrived (timeout, cancel, crash, or a
session.error mid-turn), the daemon reported SessionID="" and the
chat-session resume pointer could not advance — causing the chat to
silently drop conversation memory on the next turn.
Capture session.start.sessionId into state up front, and only let
'result' overwrite it when it actually carries one. result still wins
when present (it is the authoritative end-of-turn record).
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* fix(copilot): parse premiumRequests as float to preserve session id
Copilot CLI v1.0.32 serializes premiumRequests as a float (e.g. 7.5),
not an integer. Our copilotResultUsage struct typed it as int, which
made the entire 'result' line fail json.Unmarshal — silently dropping
sessionId on every turn.
This was the real cause of chat memory loss: the daemon reported
SessionID="" to the server, chat_session.session_id stayed NULL, and
the next chat turn never received --resume <id>, so each turn started
a fresh Copilot session with no prior context.
Add a regression test using the real JSON line from CLI v1.0.32 that
asserts sessionId is preserved when premiumRequests is fractional.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
---------
Co-authored-by: Devv <devv@Devvs-Mac-mini.local>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Eve <eve@multica.ai>
Co-authored-by: yushen <ldnvnbl@gmail.com>
* feat(agent): add LaunchHeader per agent type
Each backend in server/pkg/agent/ hardcodes a stable command skeleton
(e.g. `codex app-server --listen stdio://`, `hermes acp`) before
appending opts.CustomArgs. Surfacing that skeleton lets the UI tell
users which command their custom_args are being appended to, so a
Codex user doesn't mistakenly add `-m gpt-5.4-mini` expecting it to
reach the CLI when the subcommand is actually `app-server`.
Expose only the minimum that aids judgment — binary + subcommand, or a
short mode label when there is no subcommand — and deliberately omit
transport values, internal flags, and env to keep the surface small
and renaming-safe.
Refs #1308.
* feat(handler/runtime): surface launch_header on runtime response
runtimeToResponse now derives launch_header from agent.LaunchHeader,
piggybacking on the runtime's existing provider field so the
frontend's RuntimeDevice gains the skeleton without a new endpoint or
DB query. Client gets the header for free whenever it lists agents'
runtimes — which the custom-args tab already does.
Refs #1308.
* feat(ui/agents): show launch mode preview in custom args tab
Thread the resolved RuntimeDevice from AgentDetail into CustomArgsTab
and render its launch_header as a one-line preview above the args
list, so users see `codex app-server <your args>` (or equivalent per
provider) and can tell whether a CLI-style flag like `--model` will
actually reach the invoked subcommand. Source of truth stays in the
Go backend; the TS type just carries the string.
Refs #1308.
When --resume targets a dead session, claude prints
"No conversation found with session ID: ..." to stderr, emits a stream-json
system init with a fresh session_id, then exits with code 1. The backend
was treating that fresh id as the authoritative session, so
daemon.go's retry-with-fresh-session fallback (SessionID == "" guard)
never triggered. Every subsequent task for the same (issue, agent) pair
stayed permanently broken until the server-side session_id was cleared by
hand.
Fix: when --resume was requested but the emitted session_id differs AND
the run failed, drop the fresh id from Result so the daemon's existing
fallback can do its job. Factored into a pure helper and unit-tested.
Fixes#1284
Co-authored-by: fuxiao <fuxiao@zyql.com>
* fix(agent): add per-agent mcp_config field to restore MCP access
Closes#1111
The --strict-mcp-config flag was added defensively in #592 to prevent
Claude agents from inheriting MCP state from the outer Claude Code session.
It was meant to be paired with --mcp-config <path> to inject a controlled
set of MCPs, but that path was never implemented, which silently stripped
all user-scope MCPs from spawned agents.
This PR completes the original design by:
- Adding a nullable mcp_config jsonb column to the agents table
- Wiring mcp_config through AgentResponse, Create/Update requests
- Piping it into ExecOptions.McpConfig in the daemon
- Serializing to a temp file and passing --mcp-config <path> in buildClaudeArgs
- Blocklisting --mcp-config in claudeBlockedArgs to prevent override
via custom_args
Does not touch Codex provider (tracked separately in #674).
Does not implement Multica MCP auto-injection (out of scope).
* fix: disambiguate JSON null vs absent for mcp_config
Follow-up to #1192. Document the v2 protocol contract that the
dispatch-level threadId guard relies on, and lock down the two leakage
paths the guard closes:
- turn/completed from a subagent thread must not call onTurnDone
- item/completed (agentMessage, final_answer) from a subagent thread
must neither leak text into the output builder nor terminate the turn
Without these tests a future refactor that drops or relocates the guard
would not be caught by CI, since existing notification tests omit the
top-level threadId field and pass through unfiltered.
* fix(daemon): filter thread/status/changed by threadId to prevent subagent interference
When Codex CLI has memories enabled, the app-server spawns a memory
consolidation subagent as a separate thread within the same stdio
connection. When that subagent thread finishes and transitions to idle,
the daemon's codex backend mistakenly interprets the idle signal as the
main turn completing, causing it to close stdin and cancel the context
before the real turn produces any output.
Add a threadId check to the thread/status/changed handler so only
status changes from the tracked thread trigger turn completion. Signals
from subagent threads (threadId != c.threadID) are now ignored.
Fixes#1181
* fix(codex): dispatch-level threadId filter for subagent notifications
Codex multiplexes subagent threads (e.g. memory consolidation) on
the same stdio pipe. Previously only thread/status/changed had a
threadId guard, but item/completed (agentMessage + final_answer),
turn/completed, and turn/started from subagent threads could still
trigger onTurnDone or contaminate output.
Move the threadId check to the top of handleRawNotification so all
notification handlers are protected. Remove the now-redundant
per-handler check on thread/status/changed.
Fixesmultica-ai/multica#1181
---------
Co-authored-by: fuxiao <fuxiao@zyql.com>
Every other backend (Claude, Gemini, OpenCode, OpenClaw, Hermes) honors
ExecOptions.ResumeSessionID — only Codex didn't. That's why users on
the Codex runtime saw each new comment on an issue start a fresh Codex
conversation: the daemon persists Result.SessionID per (agent, issue)
and passes it back as PriorSessionID, but codex.go always called
thread/start and never populated SessionID, so the value round-tripped
as empty.
Wire the missing half:
- Extract startOrResumeThread on codexClient. When ResumeSessionID is
set, call thread/resume (per the Codex app-server protocol), passing
only cwd / model / developerInstructions overrides so the thread
keeps its persisted model and reasoning effort. If resume fails
(unknown thread, schema drift, transport error) fall back to
thread/start so the task still runs on a fresh thread.
- Surface the live threadID as Result.SessionID on the final emit so
the daemon stores it and feeds it back into ResumeSessionID on the
next claim.
Tests drive the new helper through the fake stdin harness, covering:
fresh start, successful resume, fallback on resume error, fallback
when resume returns no thread ID, and surfacing of thread/start
failures.
