fix: harden input validation, file safety, and error handling#227
Open
ankit1999 wants to merge 2 commits intojamiepine:mainfrom
Open
fix: harden input validation, file safety, and error handling#227ankit1999 wants to merge 2 commits intojamiepine:mainfrom
ankit1999 wants to merge 2 commits intojamiepine:mainfrom
Conversation
- New custom_models.py module for CRUD management of user-defined HF TTS models - New /custom-models API endpoints (list, add, get, delete) - Updated MLX and PyTorch backends to resolve custom model paths (custom:slug format) - Added Custom Models section to ModelManagement UI with add/remove dialogs - Updated GenerationForm and FloatingGenerateBox with grouped model selectors - Added CustomModelCreate/Response types and API client methods - Added instruct field to GenerationRequest type - Graceful actool fallback in build.rs for non-Xcode environments - Added custom_models hidden import for PyInstaller bundling Author: AJ - Kamyab (Ankit Jain)
- ModelManagement.tsx: validate model_name format before removal, guard duplicate clicks - client.ts: URL-encode modelId in removeCustomModel path - useGenerationForm.ts: constrain modelSize with regex pattern - useModelStatus.ts: extract shared hook from GenerationForm/FloatingGenerateBox - custom_models.py: atomic writes, threading lock, corrupt config backup, strict hf_repo_id regex - models.py: add regex pattern to model_size Field - config.py: platform-appropriate default data dir for PyInstaller bundles - main.py: re-raise HTTPException before broad except to preserve 202 responses - custom_models.json: remove pre-seeded entry for clean installs - requirements.txt: add platformdirs dependency - backend/README.md: document --data-dir CLI flag
📝 WalkthroughWalkthroughThis PR adds comprehensive custom HuggingFace TTS model support to the application. Changes include a new backend module for persistent custom model storage, API endpoints for CRUD operations, frontend UI components for model selection and management, and integration throughout the generation workflow to handle both built-in and custom models during voice synthesis. Changes
Sequence Diagram(s)sequenceDiagram
participant User
participant Frontend
participant API_Client
participant Backend_API
participant Custom_Models_Storage
participant Model_Resolver
User->>Frontend: Navigate to Model Management
Frontend->>API_Client: listCustomModels()
API_Client->>Backend_API: GET /custom-models
Backend_API->>Custom_Models_Storage: list_custom_models()
Custom_Models_Storage-->>Backend_API: Custom models list
Backend_API-->>API_Client: CustomModelListResponse
API_Client-->>Frontend: Render custom models UI
User->>Frontend: Add custom model (repo ID, name)
Frontend->>API_Client: addCustomModel(hf_repo_id, display_name)
API_Client->>Backend_API: POST /custom-models
Backend_API->>Custom_Models_Storage: add_custom_model(hf_repo_id, display_name)
Custom_Models_Storage->>Custom_Models_Storage: Validate & generate slug
Custom_Models_Storage->>Custom_Models_Storage: Save to custom_models.json
Custom_Models_Storage-->>Backend_API: CustomModelResponse
Backend_API-->>API_Client: Success
API_Client-->>Frontend: Update UI, show new model
User->>Frontend: Generate audio with custom model
Frontend->>API_Client: submitGenerationForm(model: "custom:slug")
API_Client->>Backend_API: POST /generate
Backend_API->>Model_Resolver: Resolve "custom:slug" to HF repo
Custom_Models_Storage-->>Model_Resolver: HF repo ID
Model_Resolver->>Model_Resolver: Download/load model from HuggingFace
Model_Resolver->>Backend_API: Model ready
Backend_API->>Backend_API: Generate audio
Backend_API-->>API_Client: Audio result
API_Client-->>Frontend: Display result
Frontend-->>User: Play/download audio
Estimated code review effort🎯 4 (Complex) | ⏱️ ~75 minutes Poem
🚥 Pre-merge checks | ✅ 3✅ Passed checks (3 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Tip Try Coding Plans. Let us write the prompt for your AI agent so you can ship faster (with fewer bugs). Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 4
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
app/src/lib/hooks/useGenerationForm.ts (1)
120-132:⚠️ Potential issue | 🟠 MajorHandle download-pending responses before using
resultasGenerationResponse.This block assumes
result.durationexists. When backend returns 202 download-pending payload, this can crash the success path.✅ Safer flow: early-return on undownloaded model
try { const modelStatus = await apiClient.getModelStatus(); const model = modelStatus.models.find((m) => m.model_name === modelName); if (model) { displayName = model.display_name; if (!model.downloaded) { // Not yet downloaded — enable progress tracking UI setDownloadingModelName(modelName); setDownloadingDisplayName(displayName); + if (!model.downloading) { + await apiClient.triggerModelDownload(modelName); + } + toast({ + title: 'Model download in progress', + description: `${displayName} is downloading. Try generating again when ready.`, + }); + return; } } } catch (error) {🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@app/src/lib/hooks/useGenerationForm.ts` around lines 120 - 132, The success handler assumes result is a fully available GenerationResponse and accesses result.duration, which will crash if the backend returned a 202 download-pending payload; update the flow in useGenerationForm where generation.mutateAsync is awaited (the block that constructs toast with result.duration) to first detect the download-pending response (e.g., check a status field or absence of duration/download URL on the returned object) and early-return or show a "download pending" toast instead of accessing result.duration; ensure subsequent logic only runs when result.duration exists (guard on result.duration or result.status === 'ready') so you never dereference duration on a pending response.
🧹 Nitpick comments (5)
backend/config.py (1)
50-57: Consider ensuring the data directory exists inget_data_dir().Unlike
set_data_dir()and the subdirectory getters (e.g.,get_profiles_dir()),get_data_dir()does not callmkdir(). If called beforeset_data_dir()and the directory doesn't exist, callers may encounter errors when writing files.🔧 Proposed fix
def get_data_dir() -> Path: """ Get the data directory path. Returns: Path to the data directory """ + _data_dir.mkdir(parents=True, exist_ok=True) return _data_dir🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@backend/config.py` around lines 50 - 57, get_data_dir() should ensure the underlying directory exists like set_data_dir() and get_profiles_dir() do: before returning the module-level _data_dir Path, call _data_dir.mkdir(parents=True, exist_ok=True) (or equivalent) to create the directory if missing so callers won’t fail when writing files; update the get_data_dir() function to perform this mkdir step and then return _data_dir.backend/backends/mlx_backend.py (1)
40-65: Consider extracting shared custom model resolution logic.The custom model resolution code (lines 40-49) is nearly identical to
pytorch_backend.py(lines 65-73). While this duplication is minimal and acceptable, you could extract it to a shared utility if this pattern expands to more backends.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@backend/backends/mlx_backend.py` around lines 40 - 65, The duplicate custom-model resolution block (checking model_size.startswith("custom:"), extracting custom_id, calling get_hf_repo_id_for_custom_model and raising ValueError) should be extracted into a shared helper so both mlx_backend and pytorch_backend reuse it; create a utility function (e.g., resolve_custom_model_hf_repo or get_hf_repo_id_from_model_size) that accepts model_size, imports/calls get_hf_repo_id_for_custom_model, returns the hf_repo_id or raises the ValueError, and replace the inline code in mlx_backend (the model_size.startswith("custom:") branch) and the corresponding code in pytorch_backend with calls to this new helper; keep existing behavior/logging (print) and preserve mlx_model_map/other logic.backend/custom_models.py (1)
47-62: Uselogger.exceptionto include traceback in error logs.Within exception handlers,
logger.exceptionautomatically includes the traceback, which is more useful for debugging thanlogger.error.♻️ Proposed fix
except json.JSONDecodeError as exc: # Back up the corrupt file so we don't lose data backup = path.with_suffix( f".json.corrupt.{datetime.utcnow().strftime('%Y%m%dT%H%M%S')}" ) try: path.rename(backup) - logger.error( + logger.exception( "Corrupt custom_models.json backed up to %s: %s", backup, exc ) except OSError as rename_err: - logger.error( + logger.exception( "Failed to back up corrupt config %s: %s (original error: %s)", path, rename_err, exc, ) raise🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@backend/custom_models.py` around lines 47 - 62, Replace the logger.error calls inside the JSON decode and nested OSError handlers with logger.exception so the traceback is recorded: in the except json.JSONDecodeError as exc block, change the logger.error("Corrupt custom_models.json backed up to %s: %s", backup, exc) to logger.exception(...) and in the inner except OSError as rename_err block change logger.error("Failed to back up corrupt config %s: %s (original error: %s)", path, rename_err, exc) to logger.exception(...) (preserving the descriptive messages and context) so the full tracebacks for the JSONDecodeError and the rename error are included.backend/main.py (1)
1546-1552: Don’t silently swallow cache-probe exceptions in model status paths.Current
except Exception: passbehavior hides actionable failures. Log at debug level so support can diagnose cache-state inconsistencies.🧭 Minimal observability patch
+import logging +logger = logging.getLogger(__name__) @@ - except Exception: - pass + except Exception: + logger.debug("Custom model cache probe failed (incomplete check): %s", hf_repo_id, exc_info=True) @@ - except Exception: - pass + except Exception: + logger.debug("Custom model size computation failed: %s", hf_repo_id, exc_info=True) @@ - except Exception: - pass - except Exception: - pass + except Exception: + logger.debug("Custom model fallback size computation failed: %s", hf_repo_id, exc_info=True) + except Exception: + logger.debug("Custom model fallback cache probe failed: %s", hf_repo_id, exc_info=True) @@ - except Exception: - pass + except Exception: + logger.debug("Custom model loaded-state check failed: %s", model_name, exc_info=True) @@ - except Exception: + except Exception: + logger.debug("Custom model status assembly failed: %s", model_name, exc_info=True) is_downloading = model_name in active_download_namesAlso applies to: 1557-1560, 1584-1593, 1598-1601, 1618-1628
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@backend/main.py` around lines 1546 - 1552, Replace silent except blocks that currently do "except Exception: pass" in the model status cache probes (the blocks that compute cache_dir_path, blobs_dir and has_incomplete using hf_constants.HF_HUB_CACHE and hf_repo_id) with exception handlers that call the module logger at debug level including a descriptive message and the exception details (e.g., "failed probing HF cache for repo {hf_repo_id}, cache_dir={cache_dir_path}") so failures aren't swallowed; make the same change for the other similar blocks flagged (the ones around lines computing blobs_dir/has_incomplete in the same function) to ensure consistent debug-level observability while preserving existing control flow.app/src/components/Generation/GenerationForm.tsx (1)
146-174: Extract shared model-option rendering/mapping to avoid drift.The built-in/custom select rendering and
model_name→sizeValuemapping is duplicated inFloatingGenerateBox.tsx; fallback text has already started diverging. Consider a shared helper/component for this block.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@app/src/components/Generation/GenerationForm.tsx` around lines 146 - 174, The built-in/custom model option rendering and the model_name→sizeValue mapping in GenerationForm (see builtInModels, customModels, SelectGroup/SelectItem usage and the sizeValue = model.model_name.replace('qwen-tts-', '')) are duplicated in FloatingGenerateBox; extract a shared helper or small presentational component (e.g., renderModelOptions or ModelSelectGroup component and a mapModelNameToSize utility) that returns the same SelectGroup/SelectItem structure and performs the model_name→sizeValue transformation, then replace the duplicated loops in both GenerationForm.tsx and FloatingGenerateBox.tsx to call the new shared function/component so fallback text and mapping remain consistent.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@app/src/components/Generation/GenerationForm.tsx`:
- Line 37: The hook useModelStatus currently classifies builtInModels using
model_name.startsWith('qwen-tts'), which incorrectly flags repos like
"qwen-tts/my-voice" as built-in; update the predicate in useModelStatus (the
filter that builds builtInModels/customModels) to only treat true built-ins such
as the core qwen-tts identifier (e.g., model_name === 'qwen-tts' or match a
pattern that disallows a following '/'), for example replace
startsWith('qwen-tts') with a stricter check (exact equality or a regex like
/^qwen-tts($|[:@])/) so repo-qualified names with a slash go to customModels.
In `@app/src/components/ServerSettings/ModelManagement.tsx`:
- Around line 617-623: The Remove button currently disables only when
model.loaded or isUnregistering is true; update the disable condition for the
Button (the component with onClick={onRemove}) to also check the showDownloading
flag so removal is prevented while a download is active. Locate the Button that
uses onRemove and the disabled prop, and add showDownloading (or its local/state
variable) into the combined disabled expression alongside model.loaded and
isUnregistering.
In `@backend/custom_models.py`:
- Line 10: Remove the unused fcntl import from the top of custom_models.py
(delete the `import fcntl` statement), verify there are no remaining references
to fcntl in the file (the module already uses threading.Lock for
synchronization), and run tests/linting to confirm no regressions.
In `@backend/voicebox-server.spec`:
- Around line 16-19: The spec unconditionally calls collect_all('mlx') and
collect_all('mlx_audio') which fails when MLX isn't installed; wrap those
collect_all calls in a runtime presence check (e.g., try/except ImportError or
importlib.util.find_spec) before invoking collect_all so that if 'mlx' or
'mlx_audio' is missing you skip updating tmp_ret/datas/binaries/hiddenimports.
Specifically, guard the collect_all('mlx') and collect_all('mlx_audio') blocks
(the places that assign tmp_ret and then do datas += tmp_ret[0]; binaries +=
tmp_ret[1]; hiddenimports += tmp_ret[2]) so they only run when the package is
importable.
---
Outside diff comments:
In `@app/src/lib/hooks/useGenerationForm.ts`:
- Around line 120-132: The success handler assumes result is a fully available
GenerationResponse and accesses result.duration, which will crash if the backend
returned a 202 download-pending payload; update the flow in useGenerationForm
where generation.mutateAsync is awaited (the block that constructs toast with
result.duration) to first detect the download-pending response (e.g., check a
status field or absence of duration/download URL on the returned object) and
early-return or show a "download pending" toast instead of accessing
result.duration; ensure subsequent logic only runs when result.duration exists
(guard on result.duration or result.status === 'ready') so you never dereference
duration on a pending response.
---
Nitpick comments:
In `@app/src/components/Generation/GenerationForm.tsx`:
- Around line 146-174: The built-in/custom model option rendering and the
model_name→sizeValue mapping in GenerationForm (see builtInModels, customModels,
SelectGroup/SelectItem usage and the sizeValue =
model.model_name.replace('qwen-tts-', '')) are duplicated in
FloatingGenerateBox; extract a shared helper or small presentational component
(e.g., renderModelOptions or ModelSelectGroup component and a mapModelNameToSize
utility) that returns the same SelectGroup/SelectItem structure and performs the
model_name→sizeValue transformation, then replace the duplicated loops in both
GenerationForm.tsx and FloatingGenerateBox.tsx to call the new shared
function/component so fallback text and mapping remain consistent.
In `@backend/backends/mlx_backend.py`:
- Around line 40-65: The duplicate custom-model resolution block (checking
model_size.startswith("custom:"), extracting custom_id, calling
get_hf_repo_id_for_custom_model and raising ValueError) should be extracted into
a shared helper so both mlx_backend and pytorch_backend reuse it; create a
utility function (e.g., resolve_custom_model_hf_repo or
get_hf_repo_id_from_model_size) that accepts model_size, imports/calls
get_hf_repo_id_for_custom_model, returns the hf_repo_id or raises the
ValueError, and replace the inline code in mlx_backend (the
model_size.startswith("custom:") branch) and the corresponding code in
pytorch_backend with calls to this new helper; keep existing behavior/logging
(print) and preserve mlx_model_map/other logic.
In `@backend/config.py`:
- Around line 50-57: get_data_dir() should ensure the underlying directory
exists like set_data_dir() and get_profiles_dir() do: before returning the
module-level _data_dir Path, call _data_dir.mkdir(parents=True, exist_ok=True)
(or equivalent) to create the directory if missing so callers won’t fail when
writing files; update the get_data_dir() function to perform this mkdir step and
then return _data_dir.
In `@backend/custom_models.py`:
- Around line 47-62: Replace the logger.error calls inside the JSON decode and
nested OSError handlers with logger.exception so the traceback is recorded: in
the except json.JSONDecodeError as exc block, change the logger.error("Corrupt
custom_models.json backed up to %s: %s", backup, exc) to logger.exception(...)
and in the inner except OSError as rename_err block change logger.error("Failed
to back up corrupt config %s: %s (original error: %s)", path, rename_err, exc)
to logger.exception(...) (preserving the descriptive messages and context) so
the full tracebacks for the JSONDecodeError and the rename error are included.
In `@backend/main.py`:
- Around line 1546-1552: Replace silent except blocks that currently do "except
Exception: pass" in the model status cache probes (the blocks that compute
cache_dir_path, blobs_dir and has_incomplete using hf_constants.HF_HUB_CACHE and
hf_repo_id) with exception handlers that call the module logger at debug level
including a descriptive message and the exception details (e.g., "failed probing
HF cache for repo {hf_repo_id}, cache_dir={cache_dir_path}") so failures aren't
swallowed; make the same change for the other similar blocks flagged (the ones
around lines computing blobs_dir/has_incomplete in the same function) to ensure
consistent debug-level observability while preserving existing control flow.
ℹ️ Review info
Configuration used: defaults
Review profile: CHILL
Plan: Pro
⛔ Files ignored due to path filters (2)
bun.lockis excluded by!**/*.locktauri/src-tauri/Cargo.lockis excluded by!**/*.lock
📒 Files selected for processing (19)
app/src/components/Generation/FloatingGenerateBox.tsxapp/src/components/Generation/GenerationForm.tsxapp/src/components/ServerSettings/ModelManagement.tsxapp/src/lib/api/client.tsapp/src/lib/api/types.tsapp/src/lib/hooks/useGenerationForm.tsapp/src/lib/hooks/useModelStatus.tsbackend/README.mdbackend/backends/mlx_backend.pybackend/backends/pytorch_backend.pybackend/build_binary.pybackend/config.pybackend/custom_models.pybackend/main.pybackend/models.pybackend/requirements.txtbackend/voicebox-server.specdata/custom_models.jsontauri/src-tauri/build.rs
| const { form, handleSubmit, isPending } = useGenerationForm(); | ||
|
|
||
| // Use shared hook for model status fetching and grouping | ||
| const { builtInModels, customModels } = useModelStatus(); |
There was a problem hiding this comment.
Prevent built-in/custom option overlap from hook classification.
builtInModels from useModelStatus is based on model_name.startsWith('qwen-tts'). A custom repo like qwen-tts/my-voice can appear in both groups and be misinterpreted in built-in mapping.
🔧 Proposed fix (in app/src/lib/hooks/useModelStatus.ts)
-const builtInModels =
- modelStatus?.models.filter((m) => m.model_name.startsWith('qwen-tts')) || [];
+const builtInModels =
+ modelStatus?.models.filter((m) => !m.is_custom && m.model_name.startsWith('qwen-tts-')) || [];🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@app/src/components/Generation/GenerationForm.tsx` at line 37, The hook
useModelStatus currently classifies builtInModels using
model_name.startsWith('qwen-tts'), which incorrectly flags repos like
"qwen-tts/my-voice" as built-in; update the predicate in useModelStatus (the
filter that builds builtInModels/customModels) to only treat true built-ins such
as the core qwen-tts identifier (e.g., model_name === 'qwen-tts' or match a
pattern that disallows a following '/'), for example replace
startsWith('qwen-tts') with a stricter check (exact equality or a regex like
/^qwen-tts($|[:@])/) so repo-qualified names with a slash go to customModels.
Comment on lines
+617
to
+623
| <Button | ||
| size="sm" | ||
| onClick={onRemove} | ||
| variant="ghost" | ||
| title="Remove custom model from list" | ||
| disabled={model.loaded || isUnregistering} | ||
| > |
There was a problem hiding this comment.
Disable “remove custom model” while download is active.
Unregistering during showDownloading can create inconsistent UX/state around in-flight downloads.
🛠️ Small UI guard improvement
<Button
size="sm"
onClick={onRemove}
variant="ghost"
title="Remove custom model from list"
- disabled={model.loaded || isUnregistering}
+ disabled={model.loaded || isUnregistering || showDownloading}
>🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@app/src/components/ServerSettings/ModelManagement.tsx` around lines 617 -
623, The Remove button currently disables only when model.loaded or
isUnregistering is true; update the disable condition for the Button (the
component with onClick={onRemove}) to also check the showDownloading flag so
removal is prevented while a download is active. Locate the Button that uses
onRemove and the disabled prop, and add showDownloading (or its local/state
variable) into the combined disabled expression alongside model.loaded and
isUnregistering.
| @author AJ - Kamyab (Ankit Jain) | ||
| """ | ||
|
|
||
| import fcntl |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
rg 'fcntl' backend/custom_models.pyRepository: jamiepine/voicebox
Length of output: 74
🏁 Script executed:
rg 'fcntl[\.(\s]' backend/custom_models.pyRepository: jamiepine/voicebox
Length of output: 44
🏁 Script executed:
head -20 backend/custom_models.pyRepository: jamiepine/voicebox
Length of output: 467
Remove unused fcntl import on line 10.
The fcntl module is imported but never used in this file. Since fcntl is Unix-only, this import will cause an ImportError on Windows. The module already uses threading.Lock for synchronization, making fcntl unnecessary.
Proposed fix
-import fcntl
import json
import logging📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
Suggested change
| import fcntl | |
| import json | |
| import logging |
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@backend/custom_models.py` at line 10, Remove the unused fcntl import from the
top of custom_models.py (delete the `import fcntl` statement), verify there are
no remaining references to fcntl in the file (the module already uses
threading.Lock for synchronization), and run tests/linting to confirm no
regressions.
Comment on lines
+16
to
+19
| tmp_ret = collect_all('mlx') | ||
| datas += tmp_ret[0]; binaries += tmp_ret[1]; hiddenimports += tmp_ret[2] | ||
| tmp_ret = collect_all('mlx_audio') | ||
| datas += tmp_ret[0]; binaries += tmp_ret[1]; hiddenimports += tmp_ret[2] |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
git ls-files | head -20Repository: jamiepine/voicebox
Length of output: 446
🏁 Script executed:
find . -name "*.spec" -o -name "*requirements*.txt" -o -name "pyproject.toml" -o -name "setup.cfg" | grep -v ".git" | head -20Repository: jamiepine/voicebox
Length of output: 169
🏁 Script executed:
ls -la backend/ 2>/dev/null || echo "backend dir not found"Repository: jamiepine/voicebox
Length of output: 1843
🏁 Script executed:
# Check if the spec file exists and examine it
if [ -f "backend/voicebox-server.spec" ]; then
echo "=== backend/voicebox-server.spec ==="
head -30 backend/voicebox-server.spec
fiRepository: jamiepine/voicebox
Length of output: 1802
🏁 Script executed:
echo "=== backend/requirements.txt ===" && cat backend/requirements.txt && echo && \
echo "=== backend/requirements-mlx.txt ===" && cat backend/requirements-mlx.txtRepository: jamiepine/voicebox
Length of output: 672
🏁 Script executed:
echo "=== Build workflows ===" && \
for f in .github/workflows/*.yml; do
echo "--- $f ---"
cat "$f" | head -50
doneRepository: jamiepine/voicebox
Length of output: 2778
🏁 Script executed:
echo "=== Check Makefile for build commands ===" && \
cat Makefile 2>/dev/null | head -80Repository: jamiepine/voicebox
Length of output: 3156
🏁 Script executed:
# Verify PyInstaller collect_all behavior and check if error handling exists in build_binary.py
echo "=== backend/build_binary.py ===" && \
head -50 backend/build_binary.py && \
echo "..." && \
tail -30 backend/build_binary.pyRepository: jamiepine/voicebox
Length of output: 2784
🏁 Script executed:
# Check if there's any error handling around collect_all in the repo
rg -n "collect_all|try.*collect" backend/Repository: jamiepine/voicebox
Length of output: 268
Guard MLX collect_all calls for environments where MLX is absent.
The spec file unconditionally calls collect_all('mlx') and collect_all('mlx_audio') (lines 16–19), but MLX is declared as Apple Silicon-only in requirements-mlx.txt and conditionally installed only on aarch64-apple-darwin. Windows and Intel Mac CI builds skip MLX dependencies, causing PyInstaller to fail when the spec file is executed on these platforms. Add package-presence guards before collecting.
Suggested fix
+from importlib.util import find_spec
+
tmp_ret = collect_all('mlx')
datas += tmp_ret[0]; binaries += tmp_ret[1]; hiddenimports += tmp_ret[2]
tmp_ret = collect_all('mlx_audio')
datas += tmp_ret[0]; binaries += tmp_ret[1]; hiddenimports += tmp_ret[2]
+if find_spec("mlx") is not None:
+ _d, _b, _h = collect_all("mlx")
+ datas += _d
+ binaries += _b
+ hiddenimports += _h
+
+if find_spec("mlx_audio") is not None:
+ _d, _b, _h = collect_all("mlx_audio")
+ datas += _d
+ binaries += _b
+ hiddenimports += _h📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
Suggested change
| tmp_ret = collect_all('mlx') | |
| datas += tmp_ret[0]; binaries += tmp_ret[1]; hiddenimports += tmp_ret[2] | |
| tmp_ret = collect_all('mlx_audio') | |
| datas += tmp_ret[0]; binaries += tmp_ret[1]; hiddenimports += tmp_ret[2] | |
| from importlib.util import find_spec | |
| if find_spec("mlx") is not None: | |
| _d, _b, _h = collect_all("mlx") | |
| datas += _d | |
| binaries += _b | |
| hiddenimports += _h | |
| if find_spec("mlx_audio") is not None: | |
| _d, _b, _h = collect_all("mlx_audio") | |
| datas += _d | |
| binaries += _b | |
| hiddenimports += _h |
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@backend/voicebox-server.spec` around lines 16 - 19, The spec unconditionally
calls collect_all('mlx') and collect_all('mlx_audio') which fails when MLX isn't
installed; wrap those collect_all calls in a runtime presence check (e.g.,
try/except ImportError or importlib.util.find_spec) before invoking collect_all
so that if 'mlx' or 'mlx_audio' is missing you skip updating
tmp_ret/datas/binaries/hiddenimports. Specifically, guard the collect_all('mlx')
and collect_all('mlx_audio') blocks (the places that assign tmp_ret and then do
datas += tmp_ret[0]; binaries += tmp_ret[1]; hiddenimports += tmp_ret[2]) so
they only run when the package is importable.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Addresses code review findings across frontend and backend — input validation, file integrity, error handling, and code deduplication.
Changes
Frontend
model_nameformat before removal; disable Remove button during pending mutationmodelIdinremoveCustomModelpathmodelSizewith regex (built-in sizes + custom slugs)GenerationForm/FloatingGenerateBoxBackend
owner/reporegex forhf_repo_idmodel_sizeField--onefilebundles (viaplatformdirs)HTTPExceptionbefore broadexcept Exceptionso 202 download-in-progress responses aren't swallowed as 500splatformdirs>=4.0.0--data-dirCLI flagSkipped
main.pyprobe_model_cache_status refactoring (pure refactoring, deferred to reduce regression risk)Verification
tsc --noEmit— zero errorspy_compile— all modified files passSummary by CodeRabbit
Release Notes
New Features
--data-diroption.Documentation