EvoX ドキュメント
  • 製品
  • 料金
  • ドキュメント
  • マーケット
EvoMap に戻る
ログインサインアップ
Overview
Practice cases
Features
Using swarmContinuous swarmRecursive tasksRecovery and follow-upDevice collaboration
Configuration
Developers
Security
Administration
Overview
Practice cases
Features
Using swarmContinuous swarmRecursive tasksRecovery and follow-upDevice collaboration
Configuration
Developers
Security
Administration
EvoX ドキュメント/Product/Continuous swarm

Continuous swarm

While the main AI is working, route additional independent tasks to background work. Continuous swarm routes these new instructions to background work and returns the results to the current conversation.

Use it to add work as you go. To divide one goal from the outset, start with ordinary swarm.

Decide what to add

Suppose the main AI is revising release notes:

New workHow to arrange it
Check links in a separate help directoryIndependent; suitable to add
Verify installation commandsSupply the file and reference material
Rewrite the paragraph currently being editedWait or reassign ownership first
Translate the revision that does not exist yetWait until that revision is ready

Check whether inputs already exist and whether file ownership overlaps. Submit dependent work only after its required inputs are available.

Enable continuous swarm

  1. Open a Code conversation with a connected runtime. In the composer’s + menu, select Continuous swarm. This entry belongs to Code and is hidden until the session has a connection.
  2. A Continuous swarm chip appears in the toolbar. Subsequent text messages use the continuous submission channel, separate from the ordinary Swarm On / Off control.
  3. Submit a self-contained task with source paths, actions, and acceptance criteria. Submission works while the main AI is idle or processing an ordinary turn.
  4. Wait for acknowledgement before adding the next instruction. A conversation permits one unacknowledged request at a time; another send preserves the draft and explains why it must wait.

This channel does not accept attachments. For local material, put files in the current workspace, verify their readable paths, and reference them in the instruction. To send attachments directly, click the continuous-swarm chip to disable the mode and send an ordinary message. An attachment rejection preserves both text and attachments.

Keep each instruction self-contained. Do not send a task’s context, scope, and acceptance criteria as separate instructions or expect a worker to recover the complete main conversation.

Include the context

Background workers do not automatically receive your complete conversation with the main AI. Replace “make the changes we discussed” with concrete inputs, actions, and deliverables.

For example, add these read-only checks independently:

Check whether the installation commands in docs/setup.md match the scripts in package.json. Report mismatches and suggestions without editing files.

Check internal links in docs/help. Report the source file, target, and reason for any inaccessible link. Do not edit files.

These are illustrative paths; substitute real project paths. Each instruction stands on its own.

For edits, specify file ownership, what must remain unchanged, and verification:

Fix only confirmed broken links in docs/help/faq.md. Preserve the meaning of the text and do not change other files. List old and replacement addresses and verify that each target opens.

Track submission and results

Distinguish acknowledgement from execution:

StateMeaning and action
SendingWaiting for acknowledgement of this request; retain it
UnconfirmedNo acknowledgement within 30 seconds, or the connection closed while waiting; this does not establish execution failure
Processing NWork is running; inspect its records and results
Needs attentionRead the rejection reason and repair inputs, permissions, or runtime conditions

For an unconfirmed request, use the retry action beside the chip. It preserves the instruction identity, content, and reasoning effort chosen at submission while sending the same request again. Do not create another identical task. After acceptance, inspect execution state; accepted does not mean completed.

Results return to the conversation. Review scope, evidence, and unfinished items, then decide how the main AI should use them:

Use the link-check findings to update the help documentation. First list the files you intend to edit; do not change the release guide currently being revised.

This connects background findings to subsequent work. A returned result does not mean integration has already happened.

Disable, stop, or retry

Disabling continuous swarm changes how later messages are handled; it does not undo submitted work. Use the appropriate stop action to interrupt execution, then confirm the actual state.

A request retry handles failed sending or uncertain acknowledgement. It does not guarantee resuming execution at the exact interrupted step. Check existing files and results before proceeding; see Recovery and follow-up.

When the goal needs more decomposition

New continuous instructions omit the recursive option. Enabling continuous mode does not turn additional work into recursive tasks. Submit a complete goal requiring hierarchical planning through the conversation path described in Recursive tasks. Existing recursive runs remain observable and recoverable through their own records.

Continuous swarm concerns when new work enters the workflow. Recursive tasks concern how that goal is broken down internally.

前のページUsing swarm次のページRecursive tasks