feat: task-level retry with exponential backoff by JackChen-me · Pull Request #37 · open-multi-agent/open-multi-agent

JackChen-me · 2026-04-03T05:54:51Z

Summary

Closes #30

Add maxRetries, retryDelayMs, retryBackoff to task config
On task failure, retry with exponential backoff (delay * backoff^attempt)
Fresh agent conversation per retry — no corrupted state carry-over
Emit task_retry event via onProgress with attempt number, error, and next delay
Cascade failure to dependents only after all retries exhausted
Works with both runTeam() and runTasks()
Zero behavior change when retry fields are not set (default: no retry)

Usage

const tasks = [
  {
    title: 'Generate report',
    assignee: 'analyst',
    description: '...',
    maxRetries: 2,        // retry up to 2 times
    retryDelayMs: 1000,   // 1s before first retry
    retryBackoff: 2,      // exponential: 1s, 2s
  },
]

const result = await orchestrator.runTasks(team, tasks)

Files changed

File	Change
`src/types.ts`	Add retry fields to `Task`, add `task_retry` to `OrchestratorEvent`
`src/task/task.ts`	`createTask()` accepts and passes through retry config
`src/orchestrator/orchestrator.ts`	Retry loop in `executeQueue()`, `sleep()` helper, plumbing in `runTasks()`/`loadSpecsIntoQueue()`
`tests/task-retry.test.ts`	NEW — 7 tests covering retry success, exhaustion, backoff calculation, no-retry default

Test plan

npm run build — compiles clean
npm run lint — type check passes
npm test — 93/93 tests pass (7 new)

Add `maxRetries`, `retryDelayMs`, and `retryBackoff` to task config. When a task fails and retries remain, the orchestrator waits with exponential backoff and re-runs the task with a fresh agent conversation. A `task_retry` event is emitted via `onProgress` for observability. Cascade failure only occurs after all retries are exhausted. Closes #30

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d7714e6189

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-04-03T05:57:50Z

-          const sharedMem = team.getSharedMemoryInstance()
-          if (sharedMem) {
-            await sharedMem.write(assignee, `task:${task.id}:result`, result.output)
+      const maxAttempts = (task.maxRetries ?? 0) + 1


Validate retry count before deriving attempt loop

The new retry loop computes maxAttempts as (task.maxRetries ?? 0) + 1 without bounding input, so any negative maxRetries (from runTasks input or parsed specs) yields maxAttempts <= 0 and the for loop never executes. Because the task was already set to in_progress, it is never completed or failed, and executeQueue can exit with a silently unexecuted task. Please clamp/validate maxRetries to a non-negative integer (or fail fast) before entering the loop.

Useful? React with 👍 / 👎.

Fixed in #38 (99b028d) — maxRetries is now clamped with Math.max(0, ...) before entering the loop.

chatgpt-codex-connector · 2026-04-03T05:57:50Z

+        maxRetries: spec.maxRetries,
+        retryDelayMs: spec.retryDelayMs,
+        retryBackoff: spec.retryBackoff,


Parse retry fields from coordinator task specs

The runTeam path now threads spec.maxRetries, spec.retryDelayMs, and spec.retryBackoff into created tasks, but parseTaskSpecs() still only copies title/description/assignee/dependsOn, so these retry fields remain undefined for coordinator-generated tasks. In practice, retry settings emitted by the coordinator are dropped, so task-level retries only work via runTasks and not via runTeam as intended.

Useful? React with 👍 / 👎.

By design — retry config is a user-facing option for runTasks(). The coordinator in runTeam() is not expected to generate retry fields, so parseTaskSpecs() intentionally omits them.

- Extract `executeWithRetry()` as a testable exported function - Add `computeRetryDelay()` with 30s max cap (prevents runaway backoff) - Remove retry fields from `ParsedTaskSpec` (dead code for runTeam path) - Deduplicate retry event emission (single code path for both error types) - Injectable delay function for test determinism - Rewrite tests to call the real `executeWithRetry`, not a copy - 15 tests covering: success, retry+success, retry+failure, backoff calculation, delay cap, delay function injection, no-retry default

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 972f9f0c74

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-04-03T06:04:57Z

+      const result = await run()
+      if (result.success) {
+        return result
+      }


Aggregate token usage across retry attempts

When a task succeeds after one or more failed attempts, executeWithRetry returns only the final attempt result here, so token usage (and other execution metadata) from earlier attempts is discarded. Since executeQueue stores this single result in ctx.agentResults and buildTeamRunResult() sums result.tokenUsage, runs with retries underreport actual model consumption, which can skew billing/quota tracking and observability.

Useful? React with 👍 / 👎.

Fixed in d9b20c0 — executeWithRetry now accumulates totalUsage across all attempts and returns it in the final result.

- maxRetries clamped to >= 0 (negative values treated as no retry) - retryBackoff clamped to >= 1 (prevents zero/negative delay oscillation) - retryDelayMs clamped to >= 0 - Add tests for negative maxRetries and negative backoff Addresses Codex review P1 on #37

Previously only the final attempt's tokenUsage was returned, causing under-reporting of actual model consumption when retries occurred. Now all attempts' token counts are summed in the returned result. Addresses Codex review P2 (token usage) on #37

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 63e1f7068a

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-04-03T06:09:59Z

+  onRetry?: (data: { attempt: number; maxAttempts: number; error: string; nextDelayMs: number }) => void,
+  delayFn: (ms: number) => Promise<void> = sleep,
+): Promise<AgentRunResult> {
+  const maxAttempts = Math.max(0, task.maxRetries ?? 0) + 1


Guard maxRetries against non-finite values

executeWithRetry derives maxAttempts directly from task.maxRetries, but Math.max(0, task.maxRetries ?? 0) does not sanitize Infinity/NaN. If a caller passes maxRetries: Infinity (possible via runTasks since this is plain JS input), the for (attempt <= maxAttempts) loop becomes effectively unbounded and the task never exits retrying, stalling executeQueue; NaN similarly produces a broken attempt bound. Please coerce to a finite non-negative integer before entering the loop.

Useful? React with 👍 / 👎.

Fixed in d9b20c0 — all retry fields are now guarded with Number.isFinite() before use.

chatgpt-codex-connector Bot reviewed Apr 3, 2026

View reviewed changes

JackChen-me added 2 commits April 3, 2026 14:05

JackChen-me merged commit 4d7564b into main Apr 3, 2026
3 checks passed

JackChen-me deleted the feat/task-retry branch April 3, 2026 06:08

chatgpt-codex-connector Bot reviewed Apr 3, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: task-level retry with exponential backoff#37

feat: task-level retry with exponential backoff#37
JackChen-me merged 4 commits into
mainfrom
feat/task-retry

JackChen-me commented Apr 3, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Apr 3, 2026

Uh oh!

JackChen-me Apr 3, 2026

Uh oh!

chatgpt-codex-connector Bot Apr 3, 2026

Uh oh!

JackChen-me Apr 3, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Apr 3, 2026

Uh oh!

JackChen-me Apr 3, 2026

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Apr 3, 2026

Uh oh!

JackChen-me Apr 3, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

JackChen-me commented Apr 3, 2026

Summary

Usage

Files changed

Test plan

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Apr 3, 2026

Choose a reason for hiding this comment

Uh oh!

JackChen-me Apr 3, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot Apr 3, 2026

Choose a reason for hiding this comment

Uh oh!

JackChen-me Apr 3, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Apr 3, 2026

Choose a reason for hiding this comment

Uh oh!

JackChen-me Apr 3, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Apr 3, 2026

Choose a reason for hiding this comment

Uh oh!

JackChen-me Apr 3, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant