Add .gitignore and update app.py for platform-specific MLX imports; modify requirements.txt to comment out MLX dependencies

.gitignore

@@ -0,0 +1,237 @@
+# Python
+__pycache__/
+*.py[cod]
+*$py.class
+*.so
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django/Flask stuff
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+.python-version
+
+# pipenv
+Pipfile.lock
+
+# poetry
+poetry.lock
+
+# pdm
+.pdm.toml
+
+# PEP 582
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.env.local
+.env.development.local
+.env.test.local
+.env.production.local
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# MLX/AI Model Files
+*.gguf
+*.bin
+*.safetensors
+*.h5
+*.onnx
+*.pb
+*.tflite
+*.mlmodel
+*.pth
+*.pt
+models/
+checkpoints/
+model_cache/
+.cache/
+huggingface_hub/
+
+# Transformers cache
+~/.cache/huggingface/
+.transformers_cache/
+
+# MLX compiled files
+/tmp/*.so
+/tmp/*.cpp
+*.mlx_cache/
+
+# Gradio temporary files
+gradio_cached_examples/
+flagged/
+
+# Web scraping cache
+.ddgs_cache/
+requests_cache/
+
+# Local development
+local_test/
+test_files/
+debug/
+
+# Docker
+.dockerignore
+
+# Node.js (if any frontend dependencies)
+node_modules/
+npm-debug.log*
+yarn-debug.log*
+yarn-error.log*
+package-lock.json
+yarn.lock
+
+# Build outputs
+dist/
+out/
+
+# Cache directories
+.parcel-cache/
+
+# IDE and editor files
+.vscode/
+.idea/
+*.swp
+*.swo
+*~
+.vim/
+
+# OS generated files
+.DS_Store
+.DS_Store?
+._*
+.Spotlight-V100
+.Trashes
+ehthumbs.db
+Thumbs.db
+desktop.ini
+
+# Logs
+logs/
+*.log
+
+# Runtime data
+pids/
+*.pid
+*.seed
+*.pid.lock
+
+# Temporary files
+tmp/
+temp/
+*.tmp
+*.temp
+
+# API Keys and secrets (extra protection)
+config.json
+secrets.json
+.secrets
+api_keys.txt
+tavily_key.txt
+
+# Large files that shouldn't be committed
+*.zip
+*.tar.gz
+*.rar
+*.7z
+large_files/
+
+# Backup files
+*.bak
+*.backup
+*.orig
+
+# Database files
+*.db
+*.sqlite
+*.sqlite3

app.py

@@ -57,7 +57,13 @@ except Exception:
 
 
 try:
-    from mlx_lm import load as mlx_load, generate as mlx_generate
+    import platform
+    # Only import MLX on macOS as it's not stable on Linux
+    if platform.system() == "Darwin":
+        from mlx_lm import load as mlx_load, generate as mlx_generate
+    else:
+        mlx_load = None
+        mlx_generate = None
 except Exception:
     mlx_load = None
     mlx_generate = None
@@ -473,7 +479,7 @@ def _ensure_local_model(selected_model: str):
     t0 = time.perf_counter()
     if selected_model == "ibraschwan/Kumru-2B-mlx-4Bit":
         if mlx_load is None or mlx_generate is None:
-            raise RuntimeError("mlx-lm is not available.")
+            raise RuntimeError(f"MLX is not available on {platform.system()}. Please use a different model.")
         _backend = "mlx"
         _mlx_model, _mlx_tokenizer = mlx_load(MLX_REPO)
         log_event("model_loaded", model_id=selected_model, backend=_backend, elapsed_ms=int((time.perf_counter() - t0) * 1000))
@@ -568,8 +574,14 @@ def local_stream_generate(selected_model, system_message, user_msg, max_tokens,
 
 
 def pick_inference_model(selected_model: str) -> str:
+    # Models that don't support remote inference
     unsupported = {MLX_REPO, LORA_REPO}
-    return "vngrs-ai/Kumru-2B" if selected_model in unsupported else selected_model
+    if selected_model in unsupported:
+        return "vngrs-ai/Kumru-2B"
+    # Also fallback MLX model selection to base model on non-macOS
+    if selected_model == "ibraschwan/Kumru-2B-mlx-4Bit" and platform.system() != "Darwin":
+        return "vngrs-ai/Kumru-2B"
+    return selected_model
 
 
 def respond(
@@ -738,12 +750,19 @@ def respond(
         )
 
 
-model_dropdown = gr.Dropdown(
-    choices=[
+# Create model choices based on platform
+def get_model_choices():
+    base_choices = [
         "vngrs-ai/Kumru-2B",
-        "ibraschwan/Kumru-2B-mlx-4Bit",
         "ceofast/kumru-2b-lora",
-    ],
+    ]
+    # Only add MLX model on macOS
+    if platform.system() == "Darwin" and mlx_load is not None:
+        base_choices.insert(1, "ibraschwan/Kumru-2B-mlx-4Bit")
+    return base_choices
+
+model_dropdown = gr.Dropdown(
+    choices=get_model_choices(),
     value="vngrs-ai/Kumru-2B",
     label="Model",
 )

requirements.txt

@@ -19,7 +19,8 @@ huggingface_hub>=0.35.3
 requests>=2.32.5
 structlog>=25.4.0
 
-mlx==0.26.3
-mlx-lm==0.26.4
+# MLX only works reliably on macOS
+# mlx==0.26.3
+# mlx-lm==0.26.4
 # llama-cpp-python==0.3.16
 peft==0.17.1