When a native function call (e.g., web search) times out or fails without producing a function_call_output, convert_output_to_messages() emits an assistant message with tool_calls and no following tool role message. All providers (OpenAI, Azure, Google) reject this as invalid, returning "Provider returned error" for every subsequent message in that chat — permanently breaking the conversation.
After the final flush_pending() call, this PR checks if the last message is an assistant with tool_calls and appends synthetic tool role messages so the conversation remains valid.
Test plan
New test: normal flow (function_call + function_call_output) — no change in behavior
New test: orphaned function_call at end — synthetic tool response added
New test: multiple orphaned tool calls — all get synthetic responses
New test: empty output list — returns empty
New test: output with only messages (no tool calls) — no change
All tests in backend/open_webui/test/test_convert_output.py.
🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.
## 📋 Pull Request Information
**Original PR:** https://github.com/open-webui/open-webui/pull/21940
**Author:** [@maxkuminov](https://github.com/maxkuminov)
**Created:** 2/27/2026
**Status:** ❌ Closed
**Base:** `main` ← **Head:** `fix/orphaned-tool-calls`
---
### 📝 Commits (1)
- [`959ff2b`](https://github.com/open-webui/open-webui/commit/959ff2b479edb92c3ecb5f63dcf9aeb255b5f105) fix: add synthetic tool responses for orphaned tool calls
### 📊 Changes
**2 files changed** (+123 additions, -0 deletions)
<details>
<summary>View changed files</summary>
➕ `backend/open_webui/test/test_convert_output.py` (+106 -0)
📝 `backend/open_webui/utils/misc.py` (+17 -0)
</details>
### 📄 Description
## Summary
- When a native function call (e.g., web search) times out or fails without producing a `function_call_output`, `convert_output_to_messages()` emits an assistant message with `tool_calls` and no following `tool` role message. All providers (OpenAI, Azure, Google) reject this as invalid, returning "Provider returned error" for **every subsequent message** in that chat — permanently breaking the conversation.
- After the final `flush_pending()` call, this PR checks if the last message is an assistant with `tool_calls` and appends synthetic `tool` role messages so the conversation remains valid.
## Test plan
- [x] New test: normal flow (function_call + function_call_output) — no change in behavior
- [x] New test: orphaned function_call at end — synthetic tool response added
- [x] New test: multiple orphaned tool calls — all get synthetic responses
- [x] New test: empty output list — returns empty
- [x] New test: output with only messages (no tool calls) — no change
All tests in `backend/open_webui/test/test_convert_output.py`.
Fixes #14577, #12135, #18121, #9435
🤖 Generated with [Claude Code](https://claude.com/claude-code)
---
<sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
📋 Pull Request Information
Original PR: https://github.com/open-webui/open-webui/pull/21940
Author: @maxkuminov
Created: 2/27/2026
Status: ❌ Closed
Base:
main← Head:fix/orphaned-tool-calls📝 Commits (1)
959ff2bfix: add synthetic tool responses for orphaned tool calls📊 Changes
2 files changed (+123 additions, -0 deletions)
View changed files
➕
backend/open_webui/test/test_convert_output.py(+106 -0)📝
backend/open_webui/utils/misc.py(+17 -0)📄 Description
Summary
function_call_output,convert_output_to_messages()emits an assistant message withtool_callsand no followingtoolrole message. All providers (OpenAI, Azure, Google) reject this as invalid, returning "Provider returned error" for every subsequent message in that chat — permanently breaking the conversation.flush_pending()call, this PR checks if the last message is an assistant withtool_callsand appends synthetictoolrole messages so the conversation remains valid.Test plan
All tests in
backend/open_webui/test/test_convert_output.py.Fixes #14577, #12135, #18121, #9435
🤖 Generated with Claude Code
🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.