[Bug]: UI Wrapper Crash: Attempted to access streaming response content, without having called read()

[mitraphix-design]

Bug Description

When using the native Gemini provider, executing any workspace tool triggers a red UI error pop-up. The backend correctly processes the stream, but the UI wrapper throws a streaming read error immediately after tool execution.

Steps to Reproduce

1. Configure config.yaml to use the native 'gemini' provider.
2. Initialize a session in the Hermes Desktop App.
3. Prompt the agent to execute a workspace tool (e.g., git status, file operations).
4. Wait for the tool execution to complete.
5. The red error pop-up appears in the UI immediately after.

Expected Behavior

The tool output should be processed and sent back to the model seamlessly without throwing a UI-level streaming read error pop-up.

Actual Behavior

A red error pop-up appears in the UI stating: "Attempted to access streaming response content, without having called read()".

Affected Component

Other

Messaging Platform (if gateway-related)

No response

Debug Report

N/A - This is a UI-level wrapper/diagnostic artifact triggered by tool execution. Backend httpx logic functions correctly.

Operating System

Windows 11

Python Version

3.11.15

Hermes Version

v0.18.0

Additional Logs / Traceback (optional)

Root Cause Analysis (optional)

No response

Proposed Fix (optional)

The UI monitoring layer or diagnostic tool is attempting to access response.content or response.text on an httpx stream that has already been consumed or closed by the backend's aiter_lines(). The UI wrapper needs to be updated to ignore or correctly handle already-consumed streams.

Are you willing to submit a PR for this?

• I'd like to fix this myself and submit a PR

[mitraphix-design]

hermes_patch.zip Man, that UI error is such a pain. I ran into the same thing with the Gemini provider. It turns out the wrapper was trying to grab the stream after the backend already finished the aiter_lines loop. I found a way to bypass that read error in the UI layer.

Thanks for confirming the root cause! However, for security reasons, I have a strict policy against downloading and executing .zip files or direct patches from issue threads.

Could you please share the specific code changes you made here in a code block, or even better, submit a Pull Request to the repository? That way, the community can review the bypass and the maintainers can merge it safely.

[GottZ]

This was generated by AI during triage.

Summary

One PR addresses #60597 by caching the body drained from the retained streaming httpx response so later Gemini error diagnostics can access it. Its implementation targets the reported cause but is incomplete for both decoded text and empty response bodies.

Related pull requests

• fix(bounded-read): populate response._content after draining streaming error body (#60597) #60786 [closed] partial — (+17/-0) — partial fix: The diff populates response._content after iter_bytes() drains the error stream, directly addressing the downstream response.text crash, but it does so only for non-empty bodies and sets response._text = None, which makes httpx 0.28.1 return None instead of decoding the cached content. Although closed as stale, fix(bounded-read): populate response._content after draining streaming error body (#60597) #60786 remains relevant as the concrete implementation attempt; despite the keep_open review on fix(bounded-read): populate response._content after draining streaming error body (#60597) #60786, the review's own diff-specific findings show that this exact revision should not be reopened or merged unchanged.

Suggested consolidation

Do not reopen or merge #60786 unchanged. Supersede it with a corrected PR that assigns b"".join(chunks) to _content unconditionally, leaves _text unset, and adds real-httpx assertions for .content and .text after both non-empty and empty bounded reads.

Complex graph

flowchart LR
    classDef open fill:#dbeafe,stroke:#1d4ed8,color:#1e3a8a
    classDef merged fill:#dcfce7,stroke:#15803d,color:#14532d
    classDef closed fill:#e5e7eb,stroke:#6b7280,color:#1f2937
    classDef unverified fill:#f3f4f6,stroke:#9ca3af,color:#374151
    classDef best stroke-width:3px,stroke:#b45309
    classDef target stroke-width:3px,stroke:#4338ca
    I60597(["issue #60597 (open)"])
    P60786["PR #60786 (closed)"]
    P60786 -.->|partial| I60597
    class I60597 open
    class I60597 target
    class P60786 closed
    click I60597 "https://github.com/NousResearch/hermes-agent/issues/60597"
    click P60786 "https://github.com/NousResearch/hermes-agent/pull/60786"

Loading

Graph: solid arrow = fixes / best fix, dashed arrow = partial or unverified (see edge label); boxed group = PRs duplicating each other; amber border = best fix; indigo border = target; gray node = closed or no verify verdict yet (state tag in the node label).

Cross-PR triage: Reviewed 1 pull request and 1 issue in this complex. Each diff was read against this issue; Assessment working set: 1 kB of PR diffs, 3 kB of issue/PR text, 2 kB of discussion (4 comments), 1 verify verdict. verdicts reflect diff content, not PR titles. Part of an automated triage batch.