General
{
const val = e.target.value;
setForm((f) => ({ ...f, taskPrefix: val || undefined }));
if (val && !/^[A-Z]{1,10}$/.test(val)) {
setPrefixError("Prefix must be 1–10 uppercase letters");
} else {
setPrefixError(null);
}
}}
/>
{prefixError && {prefixError}}
{!prefixError && Prefix for new task IDs (e.g. KB, PROJ)}
When enabled, AI-generated task specifications require manual approval before moving to Todo
When enabled (default), Fusion spawns short-lived
executor-FN-XXXX agents to run each task. When disabled, only permanent agents execute tasks and the scheduler auto-assigns work using the agent reporting chain. Tasks with no eligible permanent agent stay queued.
Controls how future task specs handle release-note artifacts at completion. Use changeset mode for repositories that follow
.changeset workflows, or changelog mode when contributors should update an existing changelog file.
Show the floating chat button in the dashboard. Chat is still accessible from the Chat tab in the mobile navigation.
Chat history
Delete chat sessions and rooms that have been idle for this many days. Default: Off.
Delete inbox/outbox messages older than this many days. Default: Off. 7 days is the suggested setting.
Chat Rooms
setForm((f) => ({ ...f, chatRoomRecentVerbatimMessages: Number(e.target.value) || undefined }))
}
/>
Number of most-recent chat-room messages kept verbatim in the responder transcript. Older messages are compacted into a summary block. Default: 25.
setForm((f) => ({ ...f, chatRoomCompactionFetchLimit: Number(e.target.value) || undefined }))
}
/>
Upper bound on messages fetched from the room store for compaction consideration. Default: 200.
setForm((f) => ({ ...f, chatRoomSummaryMaxChars: Number(e.target.value) || undefined }))
}
/>
Hard cap on the synthesized "Earlier room context" summary block. Default: 3000.
Capacity Risk Banner
Warn on the board when todo work exceeds the threshold and no idle agents are available.
setForm((f) => ({
...f,
capacityRiskTodoThreshold:
e.target.value === ""
? 0
: Math.max(0, Number.parseInt(e.target.value, 10) || 0),
}))
}
/>
Banner fires when todo count is strictly greater than this value (default 20). Applies when the banner is enabled.
GitHub Tracking
Controls whether newly created tasks have GitHub issue tracking enabled by default. Individual tasks can still override this from the task detail modal.
Tracking issues use this task's title. If a task has no title yet, Fusion can summarize its description using the title summarization model in Project Models.
{!form.autoSummarizeTitles && !form.useAiMergeCommitSummary && !form.githubTrackingEnabledByDefault
? " Enable summarization in Project Models to configure that model."
: ""}
When enabled, Fusion checks open and closed issues in the target repo for likely duplicates (using File Scope paths and key symptoms) before creating a new tracking issue. Uncheck to always create a new issue.
);
case "global-general":
return (
<>
{renderScopeBanner()}
General
When disabled, tool rows are still logged but detailed tool payloads are omitted.
Very large tool payloads may still be clipped even when this stays enabled.
Save AI thinking logs
Leave both thinking toggles off to keep the original default behavior. This only controls persistedthinking rows and does not affect assistant text or tool rows.
When enabled, the dashboard probes for a globally-installed{" "}
fn / fusion CLI by spawning{" "}
<bin> --version. Disable this if your local
dev process is the source of truth and you don't want any
outdated globally-installed binary executed during the probe.
Updates
When enabled, Fusion checks npm for new versions of{" "}
@runfusion/fusion and shows update notices in the CLI and dashboard.
Cadence is governed by the frequency below.
Controls how often the dashboard re-fetches the npm registry.
Use the version + refresh control in the header to trigger an
immediate check at any time.
When enabled (default), the dashboard automatically reloads when it
detects a new build version — either from server rebuilds or service
worker updates. Disable this to stay on the current version until you
manually refresh.
);
case "global-models": {
const selectedValue = form.defaultProvider && form.defaultModelId
? `${form.defaultProvider}/${form.defaultModelId}`
: "";
const globalModelLanes = MODEL_LANES.filter(
(lane) => lane.laneId !== "default",
);
return (
<>
{renderScopeBanner()}
{/* --- Default Model --- */}
Default Model
{modelsLoading ? (Loading available models…
) : availableModels.length === 0 ? (
No models available. Configure authentication first.
) : (
<>
Controls how much reasoning effort the AI model uses. Higher levels produce better results but cost more.
);
})()}
{availableModels.length > 0 && (
<>
Model Lanes
Global baseline models for each AI role. Project settings can override these per-project.
{globalModelLanes.map((lane) => { const provider = form[lane.globalProviderKey as keyof Settings] as string | undefined; const model = form[lane.globalModelKey as keyof Settings] as string | undefined; const value = provider && model ? `${provider}/${model}` : ""; return (Startup Model Sync
When enabled, startup fetches the latest available models from the OpenRouter API so
model pickers always include the newest catalog.
When enabled, startup refreshes models through the local
opencode models opencode --refresh
flow and publishes them under the opencode-go provider in model pickers.
OpenRouter advanced
setForm((f) => ({
...f,
openrouterAppAttribution: {
...(f.openrouterAppAttribution || {}),
referer: e.target.value,
},
}))}
/>
Leave empty to omit this header. Default: https://runfusion.ai.
setForm((f) => ({
...f,
openrouterAppAttribution: {
...(f.openrouterAppAttribution || {}),
title: e.target.value,
},
}))}
/>
Leave empty to omit this header. Default: Fusion.
{
const parsed = fromCommaSeparatedInput(e.target.value);
setForm((f) => ({
...f,
openrouterModelFilters: {
...(f.openrouterModelFilters || {}),
supported_parameters: parsed.length > 0 ? parsed : undefined,
},
}));
}}
/>
Comma-separated values sent to OpenRouter model sync.
{
const parsed = fromCommaSeparatedInput(e.target.value);
setForm((f) => ({
...f,
openrouterModelFilters: {
...(f.openrouterModelFilters || {}),
output_modalities: parsed.length > 0 ? parsed : undefined,
},
}));
}}
/>
Comma-separated values sent to OpenRouter model sync.
{
const parsed = fromCommaSeparatedInput(e.target.value);
setForm((f) => ({
...f,
openrouterProviderPreferences: {
...(f.openrouterProviderPreferences || {}),
order: parsed.length > 0 ? parsed : undefined,
},
}));
}}
/>
{
const parsed = fromCommaSeparatedInput(e.target.value);
setForm((f) => ({
...f,
openrouterProviderPreferences: {
...(f.openrouterProviderPreferences || {}),
ignore: parsed.length > 0 ? parsed : undefined,
},
}));
}}
/>
{
const parsed = fromCommaSeparatedInput(e.target.value);
setForm((f) => ({
...f,
openrouterProviderPreferences: {
...(f.openrouterProviderPreferences || {}),
only: parsed.length > 0 ? parsed : undefined,
},
}));
}}
/>
Secrets
Token Cap
{
const val = e.target.value;
setForm((f) => ({ ...f, tokenCap: val ? parseInt(val, 10) : null } as SettingsFormState));
}}
/>
{form.tokenCap != null && (
)}
Automatically compact context when approaching this token count. Leave empty for no cap (compact only on overflow errors). Set a number to proactively compact when reaching this token count.
Model Lanes
Override global model settings at the project level. Each lane controls a specific AI usage context. Unset lanes inherit from the corresponding global lane. The Project Default Model is the fallback for this project when a more specific lane is unset.
{modelsLoading ? (Loading available models…
) : availableModels.length === 0 ? (
No models available. Configure authentication first.
) : (
<>
{projectModelLanes.map((lane) => {
const status = getLaneStatus(lane);
const value = getLaneValue(lane);
const isOverridden = status === "overridden";
const laneLabel = getProjectLaneLabel(lane);
return (
{isOverridden ? "Override (Project)" : "Inherited (Global)"}
Fallback Models
{modelsLoading ? (Loading available models…
) : availableModels.length === 0 ? (
No models available.
) : (
<>
Model Presets
{presets.length === 0 ? (
) : null}
{presetDraft ? (
No presets configured yet.
) : (
{presets.map((preset) => {
const selection = applyPresetToSelection(preset);
const summary = `${selection.executorValue || "default"} / ${selection.validatorValue || "default"}`;
return (
);
})}
)}
{!presetDraft ? (
{preset.name}
{summary}
{
const name = e.target.value;
setPresetDraft((current) => current ? { ...current, name } : current);
}}
/>
{availableModels.length === 0 ? (
No models available. Configure authentication first.
) : (
<>
{(["S", "M", "L"] as const).map((sizeKey) => (
))}
) : null}
{/* --- AI Title and Git Commit Message Summarization --- */}
AI Title and Git Commit Message Summarization
Configures the model used for two short-summary jobs: auto-generating task titles from long descriptions, and generating merge commit summaries from step commits and diff stats.
When enabled, tasks created without a title but with descriptions over 200 characters
will automatically get an AI-generated title (max 60 characters). The same model is
also used to generate fallback merge commit message bodies when the branch's commit
log is empty (e.g. squash merges with no unique commits), and GitHub tracking issue
titles when a tracked task has no title yet.
When enabled, merge commit messages include an AI-generated subject plus body summary (narrative + bullets + diff-stat) instead of just listing step commit subjects. Uses the title summarization model.
{(form.autoSummarizeTitles || form.useAiMergeCommitSummary || form.githubTrackingEnabledByDefault || false) && (
<>
{modelsLoading ? (
Loading available models...
) : availableModels.length === 0 ? (
No models available. Configure authentication first.
) : (
{
if (!val) {
setForm((f) => ({
...f,
titleSummarizerProvider: undefined,
titleSummarizerModelId: undefined,
}));
return;
}
const slashIdx = val.indexOf("/");
setForm((f) => ({
...f,
titleSummarizerProvider: val.slice(0, slashIdx),
titleSummarizerModelId: val.slice(slashIdx + 1),
}));
}}
placeholder="Use fallback model"
favoriteProviders={favoriteProviders}
onToggleFavorite={handleToggleFavorite}
favoriteModels={favoriteModels}
onToggleModelFavorite={handleToggleModelFavorite}
/>
)}
Also used to summarize task descriptions into GitHub tracking issue titles when a task has no title yet.
{form.titleSummarizerProvider && form.titleSummarizerModelId
? "Using explicitly configured model"
: resolvedTitleSummarizerModel.provider && resolvedTitleSummarizerModel.modelId
? resolvedTitleSummarizerModel.provider === resolvedPlanningModel.provider
&& resolvedTitleSummarizerModel.modelId === resolvedPlanningModel.modelId
? "(using planning model)"
: resolvedTitleSummarizerModel.provider === resolvedDefaultModel.provider
&& resolvedTitleSummarizerModel.modelId === resolvedDefaultModel.modelId
? form.defaultProviderOverride && form.defaultModelIdOverride
? "(using project default model)"
: "(using global default model)"
: "(using global summarization model)"
: "(using automatic model selection)"}
Appearance
Suppress the “needs your input” banner that appears when AI sessions are awaiting input or have failed.
);
case "scheduling":
return (
<>
{renderScopeBanner()}
Scheduling
{
const val = e.target.value;
globalConcurrencyDirtyRef.current = true;
setGlobalMaxConcurrent(val === "" ? undefined : Number(val));
}}
/>
Maximum concurrent agents across all projects
{
const val = e.target.value;
setForm((f) => ({ ...f, maxConcurrent: val === "" ? undefined : Number(val) } as SettingsFormState));
}}
/>
{
const val = e.target.value;
setForm((f) => ({ ...f, maxTriageConcurrent: val === "" ? undefined : Number(val) } as SettingsFormState));
}}
/>
Maximum concurrent planning agents
{
const val = e.target.value;
setForm((f) => ({ ...f, pollIntervalMs: val === "" ? undefined : Number(val) } as SettingsFormState));
}}
/>
Strict — coordination-focused; higher per-tick tokens. Lite — pre-2026-05-11 behavior. Off — minimal procedure.
{
const val = e.target.value;
const num = Number(val);
setForm((f) => ({ ...f, taskStuckTimeoutMs: val && num > 0 ? num * 60000 : undefined }));
}}
/>
Timeout in minutes for detecting stuck tasks. When a task's agent session shows no activity for longer than this duration, the task is terminated and retried. Leave empty to disable. Suggested: 10.
{
const val = e.target.value;
const num = Number(val);
setForm((f) => ({
...f,
staleHighFanoutBlockerAgeThresholdMs: val && num > 0 ? num * 3600000 : undefined,
}));
}}
/>
Escalate high fan-out blockers only after they remain in in-progress or in-review for this many hours (age source: columnMovedAt, fallback updatedAt). Default: 2 hours.
When the stuck detector kills and re-queues a task, keep completed step statuses so the agent can resume from where it left off. Disable to reset every step to pending on each stuck retry. Default: enabled.
When enabled, tasks with stale plans (PROMPT.md older than the threshold) are automatically sent back to planning for replanning
{
const val = e.target.value;
const num = Number(val);
setForm((f) => ({ ...f, specStalenessMaxAgeMs: val !== "" ? num * 3600000 : undefined }));
}}
disabled={!form.specStalenessEnabled}
/>
Maximum age in hours before a plan is considered stale. Default: 6 hours.
Completed tasks older than the threshold are moved out of the active task database.
{
const val = e.target.value;
const num = Number(val);
setForm((f) => ({
...f,
autoArchiveDoneAfterMs: val === "" ? undefined : num * MS_PER_DAY,
}));
}}
disabled={form.autoArchiveDoneTasksEnabled === false}
/>
Number of days a task can stay in Done before it is archived. Default: 2 days (48 hours).
Compact mode keeps archive size low while preserving recent agent activity for context.
{
const val = e.target.value;
const num = Number(val);
setForm((f) => ({ ...f, maxStuckKills: val && num > 0 ? num : undefined }));
}}
/>
Maximum stuck-detector retries before a task is marked failed. Default: 6.
When enabled, tasks that modify the same files are queued serially to avoid merge conflicts
Optional file or directory paths to ignore when overlap serialization is enabled.
Paths are project-relative (for example
docs/ or generated/*).
{(form.overlapIgnorePaths && form.overlapIgnorePaths.length > 0 ? form.overlapIgnorePaths : [""]).map((path, index) => (
))}
handleOverlapIgnorePathChange(index, e.target.value)}
/>
Step Execution
Run each task step in its own fresh agent session for better isolation and error recovery. Failed steps can be retried individually.
{
const val = e.target.value;
setForm((f) => ({ ...f, maxParallelSteps: val === "" ? undefined : Number(val) }));
}}
disabled={!form.runStepsInNewSessions}
/>
Maximum number of steps to run in parallel when file scopes don't overlap (1-4)
);
case "scheduled-evals": {
const evalSettings = form.evalSettings ?? {};
const isScheduledEvalEnabled = evalSettings.enabled ?? false;
return (
<>
{renderScopeBanner()}
Scheduled Evals
setForm((current) => ({
...current,
evalSettings: {
...(current.evalSettings ?? {}),
intervalMs: event.target.value === "" ? undefined : Number(event.target.value),
},
}))
}
/>
setForm((current) => ({
...current,
evalSettings: {
...(current.evalSettings ?? {}),
evaluatorProvider: event.target.value.trim() === "" ? undefined : event.target.value,
},
}))
}
placeholder="openai"
/>
setForm((current) => ({
...current,
evalSettings: {
...(current.evalSettings ?? {}),
evaluatorModelId: event.target.value.trim() === "" ? undefined : event.target.value,
},
}))
}
placeholder="gpt-5"
/>
Leave provider and model blank to inherit the project validator lane model settings.
setForm((current) => ({
...current,
evalSettings: {
...(current.evalSettings ?? {}),
retentionDays: event.target.value === "" ? undefined : Number(event.target.value),
},
}))
}
/>
);
}
case "node-routing":
return (
<>
{renderScopeBanner()}
Node Routing
Configure how tasks are routed to execution nodes.
These settings apply at the project level.
{(() => {
const selectedNode = nodes.find((node) => node.id === form.defaultNodeId);
if (!selectedNode) return null;
return (
Selected node:
);
})()}
Used when a task has no node override. Node status is shown for safer routing selection.
Worktrees
{
const val = e.target.value;
setForm((f) => ({ ...f, maxWorktrees: val === "" ? undefined : Number(val) } as SettingsFormState));
}}
/>
Limits total git worktrees including in-review tasks
setForm((f) => ({ ...f, worktreeInitCommand: e.target.value }))
}
/>
Shell command to run in each new worktree after creation
Off by default (opt-in). When enabled, completed task worktrees are returned to an idle pool instead of being deleted, preserving build caches for faster startup
Discouraged. This restores the legacy behavior where a live
fusion/<task-id> branch collision silently forks work onto sibling branches like -2 and can hide prior commits from the default recovery flow.
{form.recycleWorktrees
? "Naming style is not applicable when recycling worktrees — pooled worktrees retain their existing names"
: "How to name fresh worktree directories. Only applies when recycling is off."}
setForm((f) => ({ ...f, worktreesDir: e.target.value }))
}
/>
{form.worktrunk?.enabled === true
? "Disabled because Worktrunk integration is enabled — worktrunk manages the worktree directory layout. Disable worktrunk integration to use a custom directory."
: <>
Optional. Supports ~ and {"{repo}"}. Defaults to <projectRoot>/.worktrees when unset. Only affects newly-created worktrees.
}
When enabled, the merger fetches from the configured remote and rebases the task branch onto the latest default-branch tip before merging — catching concurrent pushes from other collaborators or fusion workers. Any conflicts the rebase surfaces flow into the existing smart/AI resolve pipeline.
{form.worktreeRebaseBeforeMerge !== false && (
Which remote to fetch for the pre-merge rebase. "Use git default" falls back to the remote configured for the default branch (typically
)}
origin).
In addition to the remote rebase above, also rebase the task branch onto the local default-branch HEAD (rootDir). This catches sibling tasks that merged locally but haven't been pushed yet — without it, two concurrent tasks where one deletes code can have the other silently re-introduce it via the fallback strategy. Enabled by default; only disable if it causes issues with your workflow.
Worktrunk integration
Disabled by default (opt-in). When enabled, Fusion shells out to
worktrunk for worktree create, sync, prune, and remove operations and follows worktrunk's directory layout.
{!worktrunkInstallVerified && form.worktrunk?.enabled !== true && (
Install the worktrunk binary below to enable this integration.
)}
{worktrunkInstall.status === "installed" && (
worktrunk {worktrunkInstall.version ?? ""} installed at {worktrunkInstall.installPath ?? "~/.fusion/bin/worktrunk"}
)}
{(worktrunkInstall.status === "missing" || worktrunkInstall.status === "installing") && (
<>
Enable worktrunk and request approval to install the pinned release.
)}
{worktrunkInstall.status === "pending-approval" && (
<>
Awaiting approval — open Approvals to continue.
)}
{(worktrunkInstall.status === "denied" || worktrunkInstall.status === "failed") && (
<>
{worktrunkInstall.error ?? "Worktrunk install failed."}
)}
setForm((f) => ({
...f,
worktrunk: {
enabled: f.worktrunk?.enabled === true,
binaryPath: e.target.value,
onFailure: f.worktrunk?.onFailure ?? "fail",
},
}))
}
/>
Optional. Leave blank to auto-resolve; Fusion will offer to install on first use.
fail stops on worktrunk errors for explicit operator recovery; fallback-native keeps progress moving by switching to Fusion's built-in worktree backend.
Commands
setForm((f) => ({ ...f, testCommand: e.target.value || undefined }))
}
/>
Command used to run tests — injected into generated task specs
setForm((f) => ({ ...f, buildCommand: e.target.value || undefined }))
}
/>
Command used to build the project — injected into generated task specs
);
case "merge":
return (
<>
{renderScopeBanner()}
Merge
More details
When enabled, tasks that pass review are automatically merged into the main branchMore details
AI mode merges the task branch into an isolated clean-room checkout at the target branch's tip, has an AI reviewer audit the squash (with corrective retries — advisory concerns land with a logged warning, an unfixable correctness concern hard-fails), then fast-forwards the target branch and syncs your local checkout (AI reconciles a conflicting restore). Each task merges to its own target branch, or the default integration branch. The legacy merge settings below do not apply while AI merge is on.
setForm((f) => ({ ...f, merger: { ...(f.merger ?? {}), maxReviewPasses: e.target.value === "" ? undefined : Number(e.target.value) } }))
}
/>
AI corrective rounds before landing the best result (advisory concern) or hard-failing (unfixable correctness concern). Default 3. The reviewer uses your project's reviewer/validator model.
More details
Dangerous compatibility escape hatch. Leave off unless you explicitly want the legacy stash → fast-forward → restore behavior when your checked-out integration branch has unrelated local edits. When off, AI merge blocks before advancing the branch so dirty project-root edits cannot contaminate a completed merge.More details
Forces all AI lanes to use the deterministic mock provider. No network calls, zero token cost.More details
When enabled, workflow revision feedback that explicitly names files outside the original task's declared File Scope is split into a dependent follow-up task instead of being appended to the current task's PROMPT.md.
{
const rawValue = e.target.value;
if (rawValue === "") {
setForm((f) => ({ ...f, verificationFixRetries: undefined } as SettingsFormState));
return;
}
const parsedValue = Number.parseInt(rawValue, 10);
if (!Number.isFinite(parsedValue)) {
setForm((f) => ({ ...f, verificationFixRetries: undefined } as SettingsFormState));
return;
}
const clampedValue = Math.max(0, Math.min(3, parsedValue));
setForm((f) => ({ ...f, verificationFixRetries: clampedValue } as SettingsFormState));
}}
/>
More details
Controls auto-fix retry attempts after deterministic test/build verification failures — applies to both executor-time and in-merge verification (0-3).More details
Controls what happens after a task reaches In Review. Direct mode merges into the current branch locally. Pull request mode keeps the task in In Review while Fusion waits for GitHub reviews and required checks before merging the PR.
{(() => {
const currentValue = form.integrationBranch ?? "";
const valueIsKnown = currentValue.length > 0 && integrationBranchOptions.includes(currentValue);
const isCustomMode = integrationBranchCustomMode || (currentValue.length > 0 && !valueIsKnown);
if (isCustomMode) {
return (
{form.mergeStrategy !== "pull-request" && (form.merger?.mode ?? "ai") !== "ai" && (
<>
{
const trimmed = e.target.value.trim();
setForm((f) => ({
...f,
integrationBranch: trimmed.length === 0 ? undefined : trimmed,
}));
}}
data-testid="integration-branch-custom-input"
/>
);
}
const CUSTOM = "__fusion-custom__";
const AUTO = "";
return (
);
})()}
More details
The canonical branch Fusion merges tasks into and uses as the reference for all ahead/behind / overlap / pre-rebase computations. Leave on auto-detect to resolve via the standard cascade (integrationBranch → legacy baseBranch →
origin/HEAD symbolic ref → fallback main). Pick a
local branch from the dropdown — common integration names like main,
master, trunk, and develop are listed
first — or choose Custom… to type a branch that doesn't exist
locally yet. Applies to both direct merges and pull-request mode; individual
tasks can still override via task metadata.
More details
Auto keeps today's squash behavior for branches with zero or one substantive commit, but switches multi-substantive branches to a history-preserving rebase-and-merge path. Individual tasks can override this in PROMPT.md with**Direct Merge Commit Strategy:** auto|always-squash|always-rebase.
Auto-merge runs in the task worktree by default. Switch to the legacy project-root path only if you need the pre-FN-5279 fallback; worktrunk-managed projects still defer to worktrunk.
{(form.mergeIntegrationWorktree ?? "reuse-task-worktree") !== "reuse-task-worktree" && (
Legacy integration-branch mode.{" "}
Auto-merge will run rebase, conflict resolution, and squash commits inside the
project root (the user's checked-out integration-branch worktree) instead of
the task worktree. Fusion assumes that directory is already on the integration
branch and clean; if it isn't, merges may fail or touch the user's working
tree. Reuse-task-worktree is the recommended default (FN-5279). Switch back unless
you have a specific reason to opt in (FN-5348).
)}
More details
After Fusion advances the integration branch ref, the merger can auto-sync other worktrees still checked out on that branch (typically your project-root checkout).Stash + fast-forward snapshots real local edits as a patch
against the previous tip, snaps the worktree to the new tip, then reapplies the
patch — untracked files that collide with newly-tracked paths are left in a temp
dir for manual recovery. Fast-forward only snaps cleanly when the
worktree has no edits and skips otherwise. Off is the legacy
behavior: git status in your project root will show the new commits
inverted as "staged changes" until you pull manually. Only applies to direct
merges.
More details
When enabled, Fusion holds the PR in In Review until at least one approving GitHub review has been submitted. Useful on free private repos where GitHub's required-reviewer enforcement isn't available — without this, a fresh PR with no required checks is treated as immediately mergeable.GitHub Authentication
setForm((f) => ({ ...f, githubAuthToken: e.target.value || undefined }))
}
/>
)}
More details
When disabled, merge commit messages omit the task ID from the scope (e.g.feat: ... instead of feat(KB-001): ...)
More details
When enabled, commits made by Fusion keep your git identity as the primary author and append aCo-authored-by trailer crediting
Fusion (recognized by GitHub for shared attribution).
setForm((f) => ({
...f,
commitAuthorName: e.target.value || undefined,
}))
}
/>
Name used in the
Co-authored-by trailer
setForm((f) => ({
...f,
commitAuthorEmail: e.target.value || undefined,
}))
}
/>
Email used in the
)}
Co-authored-by trailer
More details
When enabled, lock files (package-lock.json, pnpm-lock.yaml, etc.), generated files (dist/*, *.gen.ts), and trivial whitespace conflicts are resolved automatically without AI intervention. Complex code conflicts still require AI review.More details
When enabled, lock files (package-lock.json, pnpm-lock.yaml, etc.) are resolved using 'ours' strategy, generated files (dist/*, *.gen.ts) using 'theirs' strategy, and trivial whitespace conflicts are auto-resolved without spawning an AI agent. Complex code conflicts still require AI review.More details
Both Smart options start with a best-effortgit fetch + fast-forward of local main from origin (so a freshly-pushed sibling commit doesn't get clobbered), then run an AI agent, then auto-resolve handles lock/generated/trivial files. They differ only in the final fallback:
{" "}
Smart, prefer main uses -X ours so main wins — protects just-merged sibling work and is the new default.
{" "}
Smart, prefer task uses -X theirs so the task branch wins — fast, but can resurrect code an earlier sibling task deleted (the FN-2887 class of regression).
{" "}
AI only retries the AI agent rather than auto-picking a side.
{" "}
Abort stops after the first AI attempt and waits for a human.
{" "}
Legacy "smart" and "prefer-main" values from older settings are migrated automatically.
When using smart-prefer-main, automatically prefer the branch side for files that main has recently modified to avoid silently discarding branch work.
Controls the post-merge audit gate. Warn (default) logs findings but auto-completes the merge. Block is the stricter opt-in mode that refuses to auto-complete merges with duplicate-subject or touched-file overlap risks. Off skips the audit entirely. Switching to Off is recommended only if you trust your branches don't silently drop edits.
)}
More details
When enabled, the merged result is automatically pushed to the configured git remote. This includes pulling the latest from the remote first (rebase) and resolving any conflicts with AI if needed.
setForm((f) => ({ ...f, pushRemote: e.target.value || undefined }))
}
/>
)}
);
case "agent-permissions":
return (
<>
{renderScopeBanner()}
More details
Git remote to push to (e.g. "origin"). Can include branch name (e.g. "origin main"). Default: "origin".Agent Permissions
Per-agent settings override project defaults. Each category controls a separate approval gate.
Agent Provisioning Approvals
Configure project-level approval behavior for durable provisioning tools (fn_agent_create/fn_agent_delete).
Memory
Memory lives in
.fusion/memory/. Agents search with qmd first, fall back to local files when qmd is missing, and open exact line windows only when needed.
Agents get memory_search, memory_get, and memory_append tools. Search defaults to qmd with a local file fallback.
{backendLoading ? (
Checking memory write access...
) : backendError ? (
Failed to load backend status: {backendError}
) : null}
{backendStatusResolved && backendStatus.qmdAvailable === false && (
qmd is not installed. Search will use local files.
Install indexed retrieval:
)}
{backendStatus.qmdInstallCommand || "bun install -g @tobilu/qmd"}
Automatically compact memory when it exceeds the threshold on a schedule
{(form.memoryAutoSummarizeEnabled || false) && (
<>
setForm((f) => ({
...f,
memoryAutoSummarizeThresholdChars: parseInt(e.target.value, 10) || 50000,
}))
}
min={1000}
/>
Memory will be compacted when it exceeds this character count
setForm((f) => ({ ...f, memoryAutoSummarizeSchedule: e.target.value }))
}
placeholder="0 3 * * *"
/>
Cron expression for auto-summarize schedule (default: daily at 3 AM)
)}
Turns daily notes into DREAMS.md and promotes reusable lessons into MEMORY.md.
{isMemoryEnabled && form.memoryDreamsEnabled === true && (
<>
setForm((f) => ({ ...f, memoryDreamsSchedule: e.target.value }))
}
/>
Cron expression for dream processing.
Manually trigger dream processing now.
)}
setMemoryTestQuery(e.target.value)}
placeholder="Search memory with qmd"
/>
Runs the same qmd-backed memory_search path agents use.
{memoryTestResult.results.length} result{memoryTestResult.results.length === 1 ? "" : "s"}
{" "}for "{memoryTestResult.query}"
qmd {memoryTestResult.qmdAvailable ? "available" : "missing"} · {memoryTestResult.usedFallback ? "local fallback used" : "qmd path used"}
{memoryTestResult.results.length > 0 ? (
)}
-
{memoryTestResult.results.map((result, index) => (
-
{result.path}:{result.lineStart}
{result.snippet}
))}
Memory is currently disabled. You can view the file, but editing is read-only until memory is re-enabled.
)}
{isMemoryEnabled && backendStatusResolved && !isBackendWritable && (
Memory is configured with a read-only backend. You can view the file, but saving is disabled.
)}
{memoryLoading ? (
Loading memory…
) : (
{memoryDirty
? "Save or discard the current edits before switching files."
: "Choose any project memory file to view or edit. Dreams is selected by default."}
{selectedMemoryFile && (
{memoryLayerNames[selectedMemoryFile.layer]}
{selectedMemoryFile.path}
{selectedMemoryFile.size.toLocaleString()} bytes · updated {new Date(selectedMemoryFile.updatedAt).toLocaleString()}
)}
{selectedMemoryFile?.layer === "long-term" && "Curated durable decisions, conventions, constraints, and pitfalls promoted from dreams."}
{selectedMemoryFile?.layer === "daily" && "Raw daily observations, open loops, and running context for dream processing."}
{selectedMemoryFile?.layer === "dreams" && "Synthesized patterns and open loops promoted from daily memory."}
{!selectedMemoryFile && "Edits the selected memory file."}
{
setMemoryContent(content);
setMemoryDirty(true);
}}
readOnly={!isEditingAllowed}
filePath={selectedMemoryPath}
/>
{memoryDirty
? "Save or discard edits before compacting this file."
: `Compacts ${selectedMemoryPath} and writes the result back to the same file.`}
)}
{memoryDirty && isEditingAllowed && (
Cannot save: {isMemoryEnabled ? "Backend is read-only" : "Memory is disabled"}
)}
);
}
case "research-global": {
const resolvedProvider =
form.researchGlobalWebSearchProvider ??
form.researchGlobalDefaults?.searchProvider ??
"builtin";
const externalProvider =
resolvedProvider === "searxng" ||
resolvedProvider === "brave" ||
resolvedProvider === "google" ||
resolvedProvider === "tavily";
const selectedCredentialProvider =
resolvedProvider === "brave" || resolvedProvider === "tavily" ? resolvedProvider : null;
const hasMissingResearchCredential = selectedCredentialProvider
? authProviders.some((provider) => provider.id === selectedCredentialProvider && !provider.authenticated)
: false;
const setSearchProvider = (provider: Settings["researchGlobalWebSearchProvider"]) => {
setForm((current) => ({
...current,
researchGlobalWebSearchProvider: provider,
researchGlobalDefaults: {
...(current.researchGlobalDefaults ?? {}),
searchProvider: provider,
},
}));
};
return (
<>
{renderScopeBanner()}
Research Defaults
Searches and fetches use the agent's native WebSearch/WebFetch tools. No API key required.
Advanced — external search providers
setForm((current) => ({
...current,
researchGlobalSearxngUrl: event.target.value || undefined,
}))
}
placeholder="https://searx.example.com"
/>
setForm((current) => ({
...current,
researchGlobalGoogleSearchCx: event.target.value || undefined,
}))
}
placeholder="custom-search-engine-id"
/>
Configure Brave, Tavily, and Google API keys in Authentication.
setForm((current) => ({
...current,
researchGlobalMaxConcurrentRuns: event.target.value === "" ? undefined : Number(event.target.value),
}))
}
/>
setForm((current) => ({
...current,
researchGlobalMaxSourcesPerRun: event.target.value === "" ? undefined : Number(event.target.value),
researchGlobalDefaults: {
...(current.researchGlobalDefaults ?? {}),
maxSourcesPerRun: event.target.value === "" ? undefined : Number(event.target.value),
},
}))
}
/>
setForm((current) => ({
...current,
researchGlobalDefaultTimeout: event.target.value === "" ? undefined : Number(event.target.value),
}))
}
/>
setForm((current) => ({
...current,
researchGlobalFetchTimeoutMs: event.target.value === "" ? undefined : Number(event.target.value),
}))
}
/>
setForm((current) => ({
...current,
researchGlobalMaxSynthesisRounds: event.target.value === "" ? undefined : Number(event.target.value),
}))
}
/>
Missing credentials for the selected research provider.
)}
);
}
case "research-project": {
const limits = form.researchSettings?.limits;
const sources = form.researchSettings?.enabledSources;
return (
<>
{renderScopeBanner()}
Project Research Settings
Web search is always enabled. Configure the search provider under Research Defaults.
{[
["pageFetch", "Page Fetch"],
["github", "GitHub"],
["localDocs", "Local Docs"],
["llmSynthesis", "LLM Synthesis"],
].map(([key, label]) => (
))}
setForm((current) => ({
...current,
researchSettings: {
...(current.researchSettings ?? {}),
limits: {
...(current.researchSettings?.limits ?? {}),
maxConcurrentRuns: event.target.value === "" ? undefined : Number(event.target.value),
},
},
}))
}
/>
setForm((current) => ({
...current,
researchSettings: {
...(current.researchSettings ?? {}),
limits: {
...(current.researchSettings?.limits ?? {}),
maxSourcesPerRun: event.target.value === "" ? undefined : Number(event.target.value),
},
},
}))
}
/>
setForm((current) => ({
...current,
researchSettings: {
...(current.researchSettings ?? {}),
limits: {
...(current.researchSettings?.limits ?? {}),
maxDurationMs: event.target.value === "" ? undefined : Number(event.target.value),
},
},
}))
}
/>
setForm((current) => ({
...current,
researchSettings: {
...(current.researchSettings ?? {}),
limits: {
...(current.researchSettings?.limits ?? {}),
requestTimeoutMs: event.target.value === "" ? undefined : Number(event.target.value),
},
},
}))
}
/>
{researchLimitError && {researchLimitError}}
Experimental Features
Experimental features are early capabilities that are not yet fully stable.
Enable them to test new functionality, but be aware they may change or be removed.
{featureFlags.map(([key, enabled]) => (
))}
Database Backups
When enabled, the database is backed up automatically on a schedule
setForm((f) => ({ ...f, autoBackupSchedule: e.target.value }))
}
disabled={!form.autoBackupEnabled}
/>
Cron expression for backup timing. Default: 0 2 * * * (daily at 2 AM).
Examples: 0 * * * * (hourly), 0 0 * * 0 (weekly), */15 * * * * (every 15 min)
{form.autoBackupSchedule && !/^[\s\d*,/-]+$/.test(form.autoBackupSchedule) && (
Invalid cron expression format
)}
{
const val = e.target.value;
setForm((f) => ({ ...f, autoBackupRetention: val === "" ? undefined : Number(val) }));
}}
disabled={!form.autoBackupEnabled}
/>
Number of backup files to keep (oldest are deleted first). Range: 1-100.
{form.autoBackupRetention !== undefined && (form.autoBackupRetention < 1 || form.autoBackupRetention > 100) && (
Must be between 1 and 100
)}
setForm((f) => ({ ...f, autoBackupDir: e.target.value }))
}
disabled={!form.autoBackupEnabled}
/>
Directory for backup files, relative to project root
{form.autoBackupDir && form.autoBackupDir.includes("..") && (
Path cannot contain parent directory traversal (..)
)}
Database Maintenance
Prune append-only operational logs (activity log, agent logs, run audit, heartbeats) older than this
many days during periodic maintenance. Keeps the database from growing without bound — large databases
are slower to checkpoint and more prone to corruption. Default: 30 days.
Memory Backups
When enabled, project and agent memory files are backed up automatically on a schedule.
setForm((f) => ({ ...f, memoryBackupSchedule: e.target.value }))}
disabled={!form.memoryBackupEnabled}
/>
Cron expression for memory backup timing. Default: 0 3 * * * (daily at 3 AM).
{form.memoryBackupSchedule && !/^[\s\d*,/-]+$/.test(form.memoryBackupSchedule) && (
Invalid cron expression format
)}
{
const val = e.target.value;
setForm((f) => ({ ...f, memoryBackupRetention: val === "" ? undefined : Number(val) }));
}}
disabled={!form.memoryBackupEnabled}
/>
Number of memory backups to keep (oldest are deleted first). Range: 1-100.
{form.memoryBackupRetention !== undefined && (form.memoryBackupRetention < 1 || form.memoryBackupRetention > 100) && (
Must be between 1 and 100
)}
setForm((f) => ({ ...f, memoryBackupDir: e.target.value }))}
disabled={!form.memoryBackupEnabled}
/>
Directory for memory backups, relative to project root.
{form.memoryBackupDir && form.memoryBackupDir.includes("..") && (
Path cannot contain parent directory traversal (..)
)}
Loading backup info…
) : backupInfo ? (
{backupInfo.count}
backups
{backupInfo.totalSize > 1024 * 1024
? `${(backupInfo.totalSize / (1024 * 1024)).toFixed(1)} MB`
: `${(backupInfo.totalSize / 1024).toFixed(1)} KB`}
total size
View {backupInfo.backups.length} backup(s)
-
{backupInfo.backups.slice(0, 10).map((backup) => (
-
{backup.filename}{backup.size > 1024 * 1024 ? `${(backup.size / (1024 * 1024)).toFixed(1)} MB` : `${(backup.size / 1024).toFixed(1)} KB`}
))}
{backupInfo.backups.length > 10 && (
- ...and {backupInfo.backups.length - 10} more )}
Notifications
Sticky-only suppresses recovered failures; terminal-only waits for paused/in-review failed tasks; all restores legacy alerts.
{
const parsed = Number(e.target.value);
setForm((f) => ({
...f,
failureNotificationDelayMs: Number.isFinite(parsed) && parsed >= 0 ? parsed : 0,
}));
}}
/>
How long a failure must persist before a push notification is sent. 0 = notify immediately.
ntfy
{form.ntfyEnabled && (
{
const val = e.target.value;
setForm((f) => ({ ...f, ntfyTopic: val || undefined }));
}}
/>
Your ntfy.sh topic name (1–64 alphanumeric/hyphen/underscore characters).{" "}
Learn more about ntfy.sh
{form.ntfyTopic && !/^[a-zA-Z0-9_-]{1,64}$/.test(form.ntfyTopic) && (
Topic must be 1–64 alphanumeric, hyphen, or underscore characters
)}
Advanced
{
const value = e.target.value;
setForm((f) => ({ ...f, ntfyBaseUrl: value || undefined }));
}}
/>
Leave blank to keep the default server: https://ntfy.sh. Custom servers must use http:// or https://.
{
const value = e.target.value;
setForm((f) => ({ ...f, ntfyAccessToken: value || undefined }));
}}
/>
Leave blank to publish without authentication. When set, Fusion sends an Authorization Bearer header with ntfy requests.
{NOTIFICATION_EVENT_OPTIONS.map(({ event, label, description }) => {
const checked = form.ntfyEvents?.includes(event) ?? true;
return (
{description}
);
})}
{
const val = e.target.value;
setForm((f) => ({ ...f, ntfyDashboardHost: val || undefined }));
}}
/>
Base URL for deep links in notifications. When set, clicking a notification
opens the dashboard directly to the task.
{form.ntfyDashboardHost && !/^https?:\/\/.+/.test(form.ntfyDashboardHost) && (
Must be a valid URL starting with http:// or https://
)}
{testNotificationResult["ntfy"] && (
General: {testNotificationResult["ntfy"].message}
)}
{testNotificationResult["ntfy-message"] && (
Message inbox: {testNotificationResult["ntfy-message"].message}
)}
{testNotificationResult["ntfy-room"] && (
Room reply: {testNotificationResult["ntfy-room"].message}
)}
)}
Webhook
{form.webhookEnabled && (
{
const val = e.target.value;
setForm((f) => ({ ...f, webhookUrl: val || undefined }));
}}
/>
{NOTIFICATION_EVENT_OPTIONS.map(({ event, label, description }) => {
const currentEvents = form.webhookEvents ?? [...DEFAULT_NTFY_EVENTS];
const checked = currentEvents.includes(event);
return (
{description}
);
})}
{testNotificationResult["webhook"].message}
)}
Node Sync
Automatically synchronize settings between this node and connected remote nodes
{form.settingsSyncEnabled && (
<>
Include API keys and OAuth tokens in sync operations
Remote Access
{tunnelState}
{remoteStatus?.provider && · {remoteStatus.provider}}
{remoteStatus?.url &&
{tunnelState === "stopped" && externalTunnel && (
{remoteStatus.url}}
{remoteStatus?.lastError && {remoteStatus.lastError}}
External {externalTunnel.provider} tunnel detected
{externalTunnel.url && {externalTunnel.url}}
{tunnelShareLink?.qrSvg && (
Scan to open:
}`})
)}
{tailnetUrl && (
);
})()}
Tailnet URL:
)}
{accessCode && (
{tailnetUrl}
Remote access code:
)}
{tunnelShareLink?.qrSvg && (
{accessCode}
Scan to connect:
)}
cloudflared is installed
)}
{activeProvider === "cloudflare" && remoteStatus?.cloudflaredAvailable === false && (
cloudflared is not installed
{cloudflaredInstallError && {cloudflaredInstallError}}
Manual install:
{cloudflaredManualInstallCommand()}
{cloudflaredMacFallbackCommand()
? If Homebrew is unavailable: {cloudflaredMacFallbackCommand()}
: null}
{activeProvider === "tailscale" ? (
<>
Tailscale Funnel will expose this dashboard on your tailnet's public {`https://..ts.net/`} URL — no hostname or port configuration needed.
) : (
<>
{(remoteForm.remoteCloudflareQuickTunnel ?? true)
? "Using Quick Tunnel — automatically creates a random trycloudflare.com URL, no account needed."
: "Named Tunnel mode enabled — configure tunnel name, token, and ingress URL below."}
)}
{
const detailsOpen = event.currentTarget.open;
setForm((f) => {
const currentQuickTunnel = Boolean((f as Record).remoteCloudflareQuickTunnel ?? true);
const nextQuickTunnel = !detailsOpen;
if (currentQuickTunnel === nextQuickTunnel) {
return f;
}
return { ...f, remoteCloudflareQuickTunnel: nextQuickTunnel } as SettingsFormState;
});
}}
>
)}
Advanced (Named Tunnel)
{!(remoteForm.remoteCloudflareQuickTunnel ?? true) ? (
setForm((f) => ({ ...f, remoteCloudflareTunnelName: e.target.value } as SettingsFormState))} />
setForm((f) => ({ ...f, remoteCloudflareTunnelToken: e.target.value } as SettingsFormState))} />
setForm((f) => ({ ...f, remoteCloudflareIngressUrl: e.target.value } as SettingsFormState))} />
) : null}
{tunnelState === "running" || tunnelState === "starting" ? (
) : (
<>
{externalTunnel ? (
) : (
)}
{activeProvider === "cloudflare" && remoteStatus?.cloudflaredAvailable === false ? (
cloudflared must be installed to start the tunnel
) : null}
)}
Advanced Settings
setForm((f) => ({ ...f, remoteShortLivedTtlMs: Number(e.target.value || 900000) } as SettingsFormState))} />
{remoteShortLivedToken && Last short-lived token expires at {new Date(remoteShortLivedToken.expiresAt).toLocaleString()} ({remoteShortLivedToken.ttlMs}ms)}
Automatically restore tunnel on startup if it was running when last stopped.
{remoteUrlPreview.url}
Token type: {remoteUrlPreview.tokenType}
{remoteUrlPreview.expiresAt ? ` · Expires at ${new Date(remoteUrlPreview.expiresAt).toLocaleString()}` : " · No expiry"}
)}
{remoteQrSvg && (
Scan this QR code on your phone
QR SVG markup
{remoteQrSvg}
Prompts
Plugins
{activePluginsSubsection === "fusion-plugins" && (
<>
{activePluginsSubsection === "pi-extensions" && (
)}
);
case "authentication": {
// CLI-backed providers (currently just claude-cli) render their own
// compact card with Enable/Disable + Test actions — bypassing the
// OAuth/API-key rendering below. Filter them out of the standard
// sort and render alongside.
const cliAuthProviders = authProviders.filter((p) => p.type === "cli");
const nonCliProviders = authProviders.filter((p) => p.type !== "cli");
// Sort providers: authenticated first, then unauthenticated. Within each bucket, sort alphabetically by name.
const sortedProviders = [...nonCliProviders].sort((a, b) => {
if (a.authenticated !== b.authenticated) {
return a.authenticated ? -1 : 1;
}
return a.name.localeCompare(b.name);
});
const authenticatedProviders = sortedProviders.filter(p => p.authenticated);
const unauthenticatedProviders = sortedProviders.filter(p => !p.authenticated);
// CLI-backed providers live in whichever bucket matches their current
// auth state (Authenticated when signed in, Available otherwise).
const claudeCliProvider = cliAuthProviders.find((p) => p.id === "claude-cli");
const cursorCliProvider = cliAuthProviders.find((p) => p.id === "cursor-cli");
const llamaCppProvider = cliAuthProviders.find((p) => p.id === "llama-cpp");
const claudeCliCard = claudeCliProvider ? (
Authentication
{authLoading ? (Loading authentication status…
) : authProviders.length === 0 ? (
No providers available
) : (
Sign in to at least one provider to get started with AI models.
)}
{showAuthenticatedGroup && (
Authenticated
{claudeCliProvider?.authenticated && claudeCliCard}
{cursorCliProvider?.authenticated && cursorCliCard}
{llamaCppProvider?.authenticated && llamaCppCard}
{authenticatedProviders.map((provider) => (
{/* Stable icon wrapper contract for auth card tests: auth-provider-icon- */}
{provider.name}
✓ Active
{provider.authenticated && provider.keyHint && (
Key: {provider.keyHint}
)}
{provider.type === "api_key" ? (
setApiKeyInputs((prev) => ({ ...prev, [provider.id]: e.target.value }))}
disabled={authActionInProgress === provider.id}
/>
{provider.authenticated && !apiKeyInputs[provider.id] ? (
) : (
)}
{authActionInProgress === provider.id && (
Saving…
)}
{apiKeyErrors[provider.id] && (
{apiKeyErrors[provider.id]}
)}
{(provider.id === "opencode" || provider.id === "opencode-go") && opencodeApiKeyRefreshStatus[provider.id] && (
{opencodeApiKeyRefreshStatus[provider.id].message}
)}
{authActionInProgress === provider.id ? (
) : provider.loginInProgress ? (
) : (
)}
)}
Available
{claudeCliProvider && !claudeCliProvider.authenticated && claudeCliCard}
{cursorCliProvider && !cursorCliProvider.authenticated && cursorCliCard}
{llamaCppProvider && !llamaCppProvider.authenticated && llamaCppCard}
{unauthenticatedProviders.map((provider) => (
{/* Stable icon wrapper contract for auth card tests: auth-provider-icon- */}
{provider.name}
✗ Not connected
{provider.type === "api_key" ? (
setApiKeyInputs((prev) => ({ ...prev, [provider.id]: e.target.value }))}
disabled={authActionInProgress === provider.id}
/>
{authActionInProgress === provider.id && (
Saving…
)}
{apiKeyErrors[provider.id] && (
{apiKeyErrors[provider.id]}
)}
{(provider.id === "opencode" || provider.id === "opencode-go") && opencodeApiKeyRefreshStatus[provider.id] && (
{opencodeApiKeyRefreshStatus[provider.id].message}
)}
{authActionInProgress === provider.id ? (
) : provider.loginInProgress ? (
) : (
)}
{provider.id === "github-copilot" && deviceCodes[provider.id] && (provider.loginInProgress || authActionInProgress === provider.id) && (
setManualCodeInputs((prev) => ({ ...prev, [provider.id]: value }))}
onSubmit={() => void handleSubmitManualCode(provider.id)}
prompt={manualCodeConfigs[provider.id].prompt}
placeholder={manualCodeConfigs[provider.id].placeholder}
helpText={manualCodeConfigs[provider.id].helpText}
disabled={manualCodeSubmitInProgress === provider.id}
submitLabel={manualCodeSubmitInProgress === provider.id ? "Submitting…" : "Submit code"}
data-testid={`auth-manual-code-${provider.id}`}
/>
)}
)}
Enter this code on GitHub
)}
{loginInstructions[provider.id] && (provider.loginInProgress || authActionInProgress === provider.id) && (
{deviceCodes[provider.id].userCode}
Re-run the setup wizard to review or update your AI provider and model configuration.
)}
Hermes Runtime
OpenClaw Runtime
Paperclip Runtime
Settings
Loading…
) : (
{showMobileSectionPicker && (
)}
)}
{renderSectionFields()}
event.stopPropagation()}>
Select ignored overlap path
Choose a file to ignore directly, or navigate into a folder and select the current directory.
Current directory:
{overlapPathPickerCurrentPath === "." ? "(project root)" : overlapPathPickerCurrentPath}
event.stopPropagation()}>
Select worktrees directory
Navigate to the folder where Fusion should create task worktrees, then select the current directory.
Current directory:
{worktreesDirPickerCurrentPath === "." ? "(project root)" : worktreesDirPickerCurrentPath}
e.target === e.currentTarget && setImportDialogOpen(false)} role="dialog" aria-modal="true">
)}
Import Settings
Review the settings to be imported:
{importPreview.global && Object.keys(importPreview.global).length > 0 && (
Global Settings:
)}
{importPreview.project && Object.keys(importPreview.project).length > 0 && (
-
{Object.entries(importPreview.global)
.filter(([, v]) => v !== undefined)
.map(([key]) => (
- {key} ))}
Project Settings:
)}
-
{Object.entries(importPreview.project)
.filter(([, v]) => v !== undefined)
.map(([key]) => (
- {key} ))}
If unchecked, existing settings will be replaced with imported values.