Skip to content

Execution failed: CAPIError: 400 {"error":{"code":"bad_request","type":"websocket_error","message":"Duplicate item found with id fc_call_ikBrgmi06fcJeO6uIpJJEQbJ. Remove duplicate items from your input and try again."}} #3560

Description

@lgphp

Describe the bug

This started happening suddenly in the evening with the same Copilot CLI version and the same workflow.

Important detail:

  • Plain chat still works.
  • The failure only happens on the turn after a tool/function call.
  • During the day, the same flow worked normally.
  • In the evening, follow-up turns after tool calls started failing with:
    "Duplicate item found with id fc_call_..."

This suggests a server-side regression or rollout affecting tool-call session continuation, rather than a local client version issue.

Affected version

GitHub Copilot CLI 1.0.55. Run 'copilot update' to check for updates.

Steps to reproduce the behavior

No response

Expected behavior

No response

Additional context

No response

Activity

  1. added theissue type on May 28, 2026
  2. salmanhasni commented on May 28, 2026

    @salmanhasni

    +1

  3. added
    area:toolsBuilt-in tools: file editing, shell, search, LSP, git, and tool call behavior
    area:modelsModel selection, availability, switching, rate limits, and model-specific behavior
    on May 28, 2026
  4. Yajur-Grover commented on May 28, 2026

    @Yajur-Grover

    +1 - after starting a new copilot session, the issue did not re-occur.

  5. mir-am commented on May 28, 2026

    @mir-am

    I have the same issue with GPT 5.5 and GPT 5.4.

  6. lsromanls commented on May 28, 2026

    @lsromanls

    +1 — same underlying failure here.

    Observed on Copilot CLI 1.0.54 before auto-update to 1.0.55.
    Model in my logs: gpt-5.4 xhigh.
    Failure is the same websocket_error / duplicate tool-call item pattern:

    Duplicate item found with id fc_call_...

    Important detail from my logs:

    • context usage was only 25.5% (69430 / 272000), so this does not look like a context-window issue
    • the same fc_call_* ID was reused across retries
    • stack points into getCompletionWithTools / runAgenticLoop
    • plain instructions/context size is small; repo instructions file is concise

    This looks consistent with a session/tool-call continuation defect rather than user error.

    Seems to be working fine in v1.0.52 (you have to use --no-auto-update switch)

  7. VONirMan commented on Jun 17, 2026

    @VONirMan

    +1

  8. added a commit that references this issue on Jun 18, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    area:modelsModel selection, availability, switching, rate limits, and model-specific behaviorarea:toolsBuilt-in tools: file editing, shell, search, LSP, git, and tool call behavior

    Type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions