Repository navigation
feat(watch): /fleet hybrid dispatch — 2.9x faster parallel execution #776
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,5 @@ | ||
| --- | ||
| "@bradygaster/squad-cli": minor | ||
| --- | ||
|
|
||
| Add /fleet hybrid dispatch mode for squad watch --execute (#775) — enables parallel issue analysis via Copilot CLI /fleet, 2.9x faster than sequential dispatch for read-heavy work |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,11 @@ | ||
| --- | ||
| "@bradygaster/squad-sdk": minor | ||
| --- | ||
|
|
||
| Add `.squad/.scratch/` directory for organized temp file management (#790) | ||
|
|
||
| - `scratchDir()` — resolve and optionally create the scratch directory | ||
| - `scratchFile()` — create named temp files inside `.scratch/` | ||
| - Path traversal protection: prefix/ext sanitized to prevent `../` attacks | ||
| - Monotonic counter ensures unique filenames even within the same millisecond | ||
| - Agents and CLI should use these utilities instead of writing temp files to repo root |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
|
|
@@ -3,6 +3,8 @@ | |
| */ | ||
|
|
||
| import { execFile, type ChildProcess } from 'node:child_process'; | ||
| import { existsSync } from 'node:fs'; | ||
| import path from 'node:path'; | ||
| import type { WatchCapability, WatchContext, PreflightResult, CapabilityResult } from '../types.js'; | ||
| import type { MachineCapabilities } from '@bradygaster/squad-sdk/ralph/capabilities'; | ||
|
|
||
|
|
@@ -15,18 +17,32 @@ export interface ExecutableWorkItem { | |
| assignees: Array<{ login: string }>; | ||
| } | ||
|
|
||
| /** Check whether an issue carries a squad or squad:* label. */ | ||
| function hasSquadLabel(issue: ExecutableWorkItem): boolean { | ||
| return issue.labels.some(l => l.name === 'squad' || l.name.startsWith('squad:')); | ||
| } | ||
|
|
||
| /** Keywords that indicate read-heavy / analysis work. */ | ||
| const READ_KEYWORDS = [ | ||
| 'research', 'review', 'analyze', 'investigate', 'audit', | ||
| 'check', 'scan', 'assess', 'evaluate', 'fact-check', | ||
| 'document', 'report', | ||
| ]; | ||
|
|
||
| /** Keywords that indicate write-heavy / implementation work. */ | ||
| const WRITE_KEYWORDS = [ | ||
| 'fix', 'implement', 'create', 'build', 'refactor', | ||
| 'add', 'update', 'migrate', 'deploy', 'feature', | ||
| ]; | ||
|
|
||
| /** Classify an issue as read-heavy or write-heavy by title keywords. */ | ||
| export function classifyIssue(title: string): 'read' | 'write' { | ||
| const lower = title.toLowerCase(); | ||
| const isRead = READ_KEYWORDS.some(k => lower.includes(k)); | ||
| const isWrite = WRITE_KEYWORDS.some(k => lower.includes(k)); | ||
| if (isRead && !isWrite) return 'read'; | ||
| return 'write'; // default to write (safer — gets full agent session) | ||
| } | ||
|
Comment on lines
+25
to
+45
There was a problem hiding this comment. New dispatch-mode behavior (
Collaborator
Author
There was a problem hiding this comment. Fixed in commit e07e8f9. Added unit tests in \ est/cli/watch-execute.test.ts\ covering:
All 16 watch-execute tests pass. |
||
|
|
||
| /** Build agent command for a prompt. */ | ||
| function buildAgentCommand( | ||
|
|
@@ -46,38 +62,92 @@ function buildAgentCommand( | |
| return { cmd: 'gh', args }; | ||
| } | ||
|
|
||
| /** Labels that indicate an issue should not be auto-executed. */ | ||
| const BLOCKING_LABELS = ['status:blocked', 'status:wontfix', 'status:on-hold', 'blocked']; | ||
|
|
||
| /** Check whether an issue has a blocking status label. */ | ||
| function hasBlockingLabel(issue: ExecutableWorkItem): boolean { | ||
| return issue.labels.some(l => BLOCKING_LABELS.includes(l.name)); | ||
| } | ||
|
|
||
| /** Check whether an issue is already assigned to a human. */ | ||
| function isAssigned(issue: ExecutableWorkItem): boolean { | ||
| return issue.assignees.length > 0; | ||
| } | ||
|
|
||
| /** Find issues eligible for autonomous execution. | ||
| * | ||
| * Pre-filters to keep only clearly actionable items: | ||
| * - must have a squad/squad:* label | ||
| * - must not be assigned to a human (agent decides once it reads ralph-instructions.md) | ||
| * - must not carry a blocking status label | ||
| * | ||
| * Matches the PS1 ralph-watch pre-filter design. | ||
| */ | ||
| export function findExecutableIssues( | ||
| _roster: Array<{ name: string; label: string; expertise: string[] }>, | ||
| _capabilities: MachineCapabilities | null, | ||
| issues: ExecutableWorkItem[], | ||
| ): ExecutableWorkItem[] { | ||
| return issues.filter( | ||
| issue => hasSquadLabel(issue) && !isAssigned(issue) && !hasBlockingLabel(issue), | ||
| ); | ||
| } | ||
|
|
||
| /** Format issue list for the agent prompt. */ | ||
| function formatIssueList(issues: ExecutableWorkItem[]): string { | ||
| return issues.map(i => { | ||
| const labels = i.labels.map(l => l.name).join(', '); | ||
| const assigned = i.assignees?.length | ||
| ? `assigned: ${i.assignees.map(a => a.login).join(',')}` | ||
| : 'unassigned'; | ||
| return `- #${i.number}: ${i.title} [${labels}] ${assigned}`; | ||
| }).join('\n'); | ||
| } | ||
|
|
||
| /** Build the rich agent prompt matching PS1 ralph-watch design. */ | ||
| export function buildAgentPrompt( | ||
| issues: ExecutableWorkItem[], | ||
| teamRoot: string, | ||
| ): string { | ||
| const issueList = formatIssueList(issues); | ||
| const hasInstructions = existsSync(path.join(teamRoot, '.squad', 'ralph-instructions.md')); | ||
|
|
||
| if (hasInstructions) { | ||
| return [ | ||
| 'Ralph, Go! Read .squad/ralph-instructions.md for your full instructions. Follow ALL sections there. MAXIMIZE PARALLELISM — spawn agents for ALL actionable issues simultaneously.', | ||
| '', | ||
| 'Here are the current open squad issues:', | ||
| issueList, | ||
| '', | ||
| 'Task: Read the issues, follow your instructions in .squad/ralph-instructions.md, and work on what\'s actionable.', | ||
| 'WHY: Keep the squad pipeline moving — no idle work.', | ||
| 'Success: Issues get branches, PRs, and progress.', | ||
| 'Escalation: If blocked, comment on the issue and move to next.', | ||
| ].join('\n'); | ||
| } | ||
|
|
||
| // Fallback when ralph-instructions.md does not exist | ||
| return [ | ||
| 'You are Ralph, the autonomous work monitor. Review the open squad issues below and work on every actionable one. Skip issues that are blocked, waiting on external input, or already assigned.', | ||
| '', | ||
| 'Here are the current open squad issues:', | ||
| issueList, | ||
| '', | ||
| 'Task: Triage the list, pick up unblocked/unassigned issues, create branches and PRs.', | ||
| 'WHY: Keep the squad pipeline moving — no idle work.', | ||
| 'Success: Issues get branches, PRs, and progress.', | ||
| 'Escalation: If blocked, comment on the issue and move to next.', | ||
| ].join('\n'); | ||
| } | ||
|
|
||
| /** Spawn a single agent session for all eligible issues. */ | ||
| async function executeAll( | ||
| issues: ExecutableWorkItem[], | ||
| context: WatchContext, | ||
| timeoutMs: number, | ||
| ): Promise<{ success: boolean; error?: string }> { | ||
| const prompt = buildAgentPrompt(issues, context.teamRoot); | ||
| const { cmd, args } = buildAgentCommand(prompt, context); | ||
|
|
||
| return new Promise<{ success: boolean; error?: string }>((resolve) => { | ||
|
|
@@ -115,7 +185,6 @@ export class ExecuteCapability implements WatchCapability { | |
|
|
||
| async execute(context: WatchContext): Promise<CapabilityResult> { | ||
| try { | ||
| const timeout = ((context.config['timeout'] as number) ?? 30) * 60_000; | ||
|
|
||
| // Fetch open issues with squad label | ||
|
|
@@ -127,31 +196,25 @@ export class ExecuteCapability implements WatchCapability { | |
| assignees: wi.assignedTo ? [{ login: wi.assignedTo }] : [], | ||
| })); | ||
|
|
||
| // Minimal filter: must have squad or squad:* label (agent decides the rest) | ||
| const eligible = findExecutableIssues(context.roster, null, issues); | ||
|
|
||
| if (eligible.length === 0) { | ||
| return { success: true, summary: 'no squad-labeled issues found' }; | ||
| } | ||
|
|
||
| // Single agent invocation with all issues — agent reads ralph-instructions.md | ||
| const result = await executeAll(eligible, context, timeout); | ||
|
|
||
| return { | ||
| success: result.success, | ||
| summary: result.success | ||
| ? `agent dispatched with ${eligible.length} issues` | ||
| : `agent failed: ${result.error}`, | ||
| data: { dispatched: eligible.length, success: result.success }, | ||
| }; | ||
| } catch (e) { | ||
| return { success: false, summary: `execute error: ${(e as Error).message}` }; | ||
| } | ||
| } | ||
| } | ||
Oops, something went wrong.
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
--dispatch-modeis cast to the union type without runtime validation. If a user passes an invalid value, it will silently flow into config and later behave like task mode (because it won’t matchfleet|hybrid), which is hard to diagnose. Consider validating against{task,fleet,hybrid}and either erroring or warning + defaulting totask(and, iffleet|hybrid, auto-enablefleet-dispatchto match the documented usage).