Thanks to visit codestin.com
Credit goes to github.com

Skip to content

fix(installer): resolve default model falling back to Qwen on Windows#949

Merged
kovtcharov merged 2 commits into
amd:mainfrom
theonlychant:fix/default-model-qwen-fallback
May 3, 2026
Merged

fix(installer): resolve default model falling back to Qwen on Windows#949
kovtcharov merged 2 commits into
amd:mainfrom
theonlychant:fix/default-model-qwen-fallback

Conversation

@theonlychant

Copy link
Copy Markdown
Contributor

Summary

Fixes GAIA ignoring the new Gemma default model and falling back to Qwen
on Windows 11, causing the wrong model to load in the frontend.

Why

After commit 5d37771 made Gemma-4-E4B the default model, Windows users
reported that GAIA still attempts to load Qwen instead. This left the
new default model effectively unreachable on Windows, making the frontend
unusable for anyone who hadn't manually configured a model.

Linked issue

Closes #948

Changes

  • Fixed model selection logic to correctly resolve the new Gemma default
    on Windows instead of falling back to Qwen

Test plan

  • pytest tests/unit/ - passing locally
  • python util/lint.py --all - no failures
  • Manual: launch gaia chat --ui on Windows and verify Gemma loads
    instead of Qwen

Checklist

  • I have linked a GitHub issue above (Closes #948).
  • I have described why this change is being made, not just what changed.
  • I have run linting and tests locally.
  • I have updated documentation if user-visible behavior changed.

@theonlychant

Copy link
Copy Markdown
Contributor Author
~/gaia$  PYTHONPATH=src python3 -c "from gaia.agents.chat.agent import ChatAgent; print('ChatAgent import OK', ChatAgent.__name__)"
[2026-05-03 05:48:31] | INFO | faiss.loader.<module> | loader.py:132 | Loading faiss with AVX2 support.
[2026-05-03 05:48:31] | INFO | faiss.loader.<module> | loader.py:134 | Successfully loaded faiss with AVX2 support.
ChatAgent import OK ChatAgent

@theonlychant theonlychant reopened this May 3, 2026
@theonlychant

Copy link
Copy Markdown
Contributor Author

@kovtcharov sorry I closed this by accident

@kovtcharov

Copy link
Copy Markdown
Collaborator

@kovtcharov sorry I closed this by accident

No prob, if you can fix the lint error we can complete it asap.

@theonlychant

Copy link
Copy Markdown
Contributor Author

I just fixed that I tested it locally and here's the output

============================================================
[OK] All documentation version references are consistent!
============================================================


================================================================
                    LINT SUMMARY REPORT                        
================================================================

[STATS] Project Statistics:
   - Python Files: 387
   - Lines of Code: 164,113
   - Directories: src/gaia, tests

[RESULTS] Quality Check Results:

+--------------------------------+------------+-----------+
| Check                          | Status     | Issues    |
+--------------------------------+------------+-----------+
| Code Formatting (Black)        | [OK] PASS  | -         |
| Import Sorting (isort)         | [OK] PASS  | -         |
| Critical Errors (Pylint)       | [OK] PASS  | -         |
| Style Compliance (Flake8)      | [OK] PASS  | -         |
| Type Checking (MyPy)           | [!] WARN   | 1002 warns |
| Import Validation              | [OK] PASS  | -         |
| Security Check (Bandit)        | [!] WARN   | 52 warns  |
| Agent Conventions              | [!] WARN   | 11 warns  |
| Doc Version Consistency        | [OK] PASS  | -         |
+--------------------------------+------------+-----------+

[SUMMARY] Statistics:
   - Total Checks Run: 9
   - Passed: 6 (66.7%)
   - Failed: 0 (0.0%)
   - Warnings: 3 (33.3%)

============================================================
[SUCCESS] ALL QUALITY CHECKS PASSED!
[WARNING] 3 warning(s) found (non-blocking)
============================================================

[OK] Your code meets quality standards!
[OK] Ready for PR submission

@kovtcharov kovtcharov added this pull request to the merge queue May 3, 2026
Merged via the queue into amd:main with commit 63e4995 May 3, 2026
23 of 24 checks passed
theonlychant added a commit to theonlychant/gaia that referenced this pull request May 5, 2026
…amd#949)

## Summary
Fixes GAIA ignoring the new Gemma default model and falling back to Qwen
on Windows 11, causing the wrong model to load in the frontend.

## Why
After commit 5d37771 made Gemma-4-E4B the default model, Windows users
reported that GAIA still attempts to load Qwen instead. This left the
new default model effectively unreachable on Windows, making the
frontend
unusable for anyone who hadn't manually configured a model.

## Linked issue
Closes amd#948

## Changes
- Fixed model selection logic to correctly resolve the new Gemma default
  on Windows instead of falling back to Qwen

## Test plan
- [x] `pytest tests/unit/` - passing locally
- [x] `python util/lint.py --all` - no failures
- [ ] Manual: launch `gaia chat --ui` on Windows and verify Gemma loads
      instead of Qwen

## Checklist
- [x] I have linked a GitHub issue above (`Closes amd#948`).
- [x] I have described **why** this change is being made, not just what
changed.
- [x] I have run linting and tests locally.
- [ ] I have updated documentation if user-visible behavior changed.

---------

Signed-off-by: theonlychant <[email protected]>
@itomek itomek mentioned this pull request May 7, 2026
6 tasks
theonlychant pushed a commit to theonlychant/gaia that referenced this pull request May 7, 2026
## Why this matters

Ships the v0.17.6 patch: a new Email Triage Agent with Gmail (every
email body stays on local Lemonade), the OAuth PKCE foundation that
backs it, settings UI card layout, and a sweep of installer fixes that
close the remaining first-launch failures uncovered after v0.17.5.
Custom Python agents that follow the template's
`super().__init__(**kwargs)` pattern no longer crash on the first
message in the Agent UI.

Full notes: `docs/releases/v0.17.6.mdx`.

## What's New

- **Email Triage Agent with Gmail**
([amd#965](amd#965)) — `EmailTriageAgent` with
25 tools across read / organize / reply / delete / calendar mixins.
Every email body processed locally on Lemonade; seven destructive tools
confirmation-gated; three layers of prompt-injection defense; SQLite
undo log via `DatabaseMixin`.
- **OAuth PKCE foundation for Google connections**
([amd#926](amd#926)) — Self-contained
`gaia.connections` module: refresh tokens in the OS keychain (Keychain /
DPAPI / SecretService), per-agent grants in
`~/.gaia/connections/grants.json`, async token cache with refresh
rotation. Baseline for the v0.17.7 connectors framework.
- **Settings UI card layout**
([amd#969](amd#969)) — Outlined cards with
accent left-stripe replace margin-separated blocks across all settings
sections; light + dark themes both updated.

## Bug Fixes

- **Custom Python agents crashed on first message**
([amd#974](amd#974), closes
[amd#973](amd#973)) — `python_factory` now
introspects the target class's `__init__` chain and only forwards kwargs
the chain accepts, so the bare `super().__init__(**kwargs)` template
pattern no longer crashes with `unexpected keyword argument
'rag_documents'`.
- **Windows installer failed at `ensure-uv`**
([amd#968](amd#968), closes
[amd#966](amd#966)) — `uv` binary now
bundled for `win-x64`; packaged Windows rescue installer included.
- **macOS installer failed at `ensure-uv` on clean Apple Silicon**
([amd#967](amd#967), closes
[amd#941](amd#941)) — Pinned `uv` v0.5.14
(`aarch64-apple-darwin`) shipped in the DMG; new `dmg-structural-smoke`
CI job blocks future drift.
- **AppImage users hit `gaia: command not found`**
([amd#942](amd#942), closes
[amd#782](amd#782)) — Startup writes a
`~/.local/bin/gaia` shim so `gaia` is on PATH after first launch (skips
creation if already present).
- **Windows fell back to Qwen instead of Gemma 4 default**
([amd#949](amd#949), closes
[amd#948](amd#948)) — Model-resolution logic
fixed so Gemma loads correctly on Windows.

## Tooling & Docs

- **`gaia-release` skill**
([amd#939](amd#939)) — Phased release flow
with hard gates before every irreversible step; encodes the manual
pre-tag verification that caught two release-blocking bugs in v0.17.4.
- **Internal-task issue template**
([amd#906](amd#906)) — Third issue template
for agent-assignable internal work.
- **Outlook via Power Automate plan**
([amd#954](amd#954)) — Enterprise-bypass spec
for v0.17.7 Outlook integration.
- **PR description guidance sharpened**
([amd#947](amd#947)) — `CLAUDE.md` "tight and
value-focused" rule with anti-patterns.
- **Stale `macOS uv fetch removed` orphan comment removed**
([amd#975](amd#975)).

## Thanks

External contributors in this release:

- [@theonlychant](https://github.com/theonlychant) — installer fixes
([amd#968](amd#968),
[amd#942](amd#942),
[amd#949](amd#949))
- [@BlueriteSoul](https://github.com/BlueriteSoul) — reported AppImage
`gaia: command not found`
([amd#782](amd#782))
- [@nuts23](https://github.com/nuts23) — reported Windows Gemma/Qwen
fallback ([amd#948](amd#948))

## Release checklist

- [x] `util/validate_release_notes.py docs/releases/v0.17.6.mdx --tag
v0.17.6` passes
- [x] `src/gaia/version.py` → `0.17.6`
- [x] `src/gaia/apps/webui/package.json` → `0.17.6`
- [x] Navbar label in `docs/docs.json` → `v0.17.6 · Lemonade 10.2.0`
- [x] All 14 commits in the range (v0.17.5..HEAD) are represented in the
notes
- [ ] Review from @kovtcharov-amd addressed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[Bug]: continues to try to use or download Qwen 3.5 even when Gemma is default

2 participants