api错误

[panyuxi514888]

[jatmn]

there is a few pending PR that address this i think .

#783 #895 #910

[kevincodex1]

hello this is already fixed in current v0.7.0 version. we just release it earlier

[nickmesen]

hello this is already fixed in current v0.7.0 version. we just release it earlier

@kevincodex1 Thanks again for the update.

I tested this again with the current v0.7.0, and it looks like the issue may still be present in at least one of the scenarios covered by this PR.

In particular, I’m still seeing an error related to reasoning_content after running with v0.7.0.

It’s possible the v0.7.0 fix covers a different path than the one I hit — I’d love to understand if these 4 cases are already handled, or if they need additional coverage.

• assistant messages with tool calls but no visible thinking block
• assistant messages where content is a plain string
• synthetic interrupted tool execution messages
• conversation recovery preserving thinking blocks before provider-specific conversion

For that reason, I think it may be better to keep PR #895 open for now, or at least compare the current v0.7.0 implementation against the cases handled here.

I’m attaching the screenshots/log evidence below so you can review the exact error I’m seeing.

[kevincodex1]

copy on it @nickmesen

[jatmn]

This looks partially covered by #1248, which landed the resume/recovery-side fix for reasoning-echo providers.

The remaining runtime/shim edge cases described above appear to be handled by #1155, which is approved but still open. Since the v0.7.0 path was reported as still reproducing here, I’m leaving this issue open until #1155 lands or a successor PR covers the same cases.