UN-3149 [FEAT] Support CSV, TXT, and Excel files in Prompt Studio#1783
UN-3149 [FEAT] Support CSV, TXT, and Excel files in Prompt Studio#1783jagadeeswaran-zipstack wants to merge 2 commits intomainfrom
Conversation
- Store original files in main directory, converted PDFs in converted/ subdir - CSV/TXT/Excel skip PDF conversion entirely and show only Raw View - Backend resolves converted PDF transparently for preview (fetch_contents_ide) - Frontend auto-switches to Raw View for non-PDF file types - Add upload_converted_for_ide() helper method - Include converted/ in directory creation and delete cleanup Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Summary by CodeRabbitRelease Notes
WalkthroughAdds a conversion-to-PDF preview flow: backend stores converted PDFs under a new Changes
Sequence Diagram(s)sequenceDiagram
participant Client
participant BackendAPI
participant Converter
participant FileStorage
participant Database
Client->>BackendAPI: upload_for_ide(file)
BackendAPI->>BackendAPI: check file_converter_plugin & should_convert_to_pdf
alt convert to PDF
BackendAPI->>Converter: convert file -> PDF
Converter-->>BackendAPI: converted PDF
BackendAPI->>FileStorage: upload converted PDF to converted/{name}.pdf
BackendAPI->>FileStorage: store original file in main dir
else store original
BackendAPI->>FileStorage: store original file in main dir
end
BackendAPI->>Database: create document record (original filename)
BackendAPI-->>Client: upload response
Client->>BackendAPI: fetch_contents_ide(file, view=ORIGINAL)
alt converted PDF exists
BackendAPI->>FileStorage: fetch converted/{name}.pdf
FileStorage-->>BackendAPI: converted PDF data
else fetch original
BackendAPI->>FileStorage: fetch original file
FileStorage-->>BackendAPI: original data
end
BackendAPI-->>Client: file content response
Client->>Client: render by mimeType (PDF → PdfViewer, non-PDF → placeholder / Raw View)
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes 🚥 Pre-merge checks | ✅ 2 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (2 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing touches
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. 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: 1
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (3)
frontend/src/components/custom-tools/manage-docs-modal/ManageDocsModal.jsx (1)
592-614:⚠️ Potential issue | 🟡 MinorStale error message: "Only PDF files are allowed" is no longer accurate.
When
ConfirmMultiDocis not available and the file type isn't inDIRECT_UPLOAD_TYPES, the error message still says "Only PDF files are allowed" (Line 607). Now that CSV, TXT, and Excel files are also accepted, this message is misleading.Proposed fix
if (!ConfirmMultiDoc) { setAlertDetails({ type: "error", - content: "Only PDF files are allowed", + content: "Only PDF, CSV, TXT, and Excel files are allowed", }); }backend/utils/file_storage/helpers/prompt_studio_file_helper.py (1)
178-201:⚠️ Potential issue | 🟠 MajorVerify that
text_content_stringis always defined before Line 201.If a file's MIME type is in
allowed_content_typesbut doesn't match any of the explicit branches (PDF, text/plain, text/csv, Excel), execution reaches theelseat Line 198, which only logs a warning. Control then falls to Line 201 wheretext_content_stringis referenced — but it was never assigned, causing anUnboundLocalError.This is a pre-existing issue, but the new branches make it more visible. Consider initializing
text_content_string = ""before the if/elif chain or adding a return/raise in theelsebranch.Proposed fix
def fetch_file_contents( org_id: str, user_id: str, tool_id: str, file_name: str, allowed_content_types: list[str], ) -> dict[str, Any]: """Method to fetch file contents from the remote location. The path is constructed in runtime based on the args """ + text_content_string: str = "" fs_instance = EnvHelper.get_storage(backend/prompt_studio/prompt_studio_core_v2/views.py (1)
576-605:⚠️ Potential issue | 🟡 MinorConversion failure will surface as an unhandled 500 error.
If
process_file(Line 581) raises an exception (e.g., corrupted file, unsupported format variant), there's no try/except around the conversion block, so the entire upload request fails with a generic 500. Consider wrapping the conversion in a try/except and either falling back to storing the original without a converted preview, or returning a descriptive error message.Proposed approach
if file_converter_service.should_convert_to_pdf(file_type): + try: converted_data, converted_name = file_converter_service.process_file( uploaded_file, file_name ) PromptStudioFileHelper.upload_converted_for_ide( org_id=UserSessionUtils.get_organization_id(request), user_id=custom_tool.created_by.user_id, tool_id=str(custom_tool.tool_id), file_name=converted_name, file_data=converted_data, ) + except Exception: + logger.warning( + f"Failed to convert {file_name} to PDF for preview, " + "storing original only" + ) # Reset uploaded_file for storing original in main dir uploaded_file.seek(0) file_data = uploaded_file
🤖 Fix all issues with AI agents
In `@backend/prompt_studio/prompt_studio_core_v2/views.py`:
- Line 577: The conversion branch uses the browser-supplied
uploaded_file.content_type (referenced as file_type at the if with
file_converter_plugin) instead of the server-detected MIME from FileValidator,
causing inconsistent behavior; fix by persisting the server-detected MIME (e.g.,
add a property or return value like detected_mime on FileValidator or its
validate() result) or re-run magic.from_buffer() in the view before the
conversion decision, then use that server-detected MIME (detected_mime) in the
conditional (the same place that checks file_converter_plugin and file_type) so
conversion vs direct upload is based on the validated content type.
🧹 Nitpick comments (3)
frontend/src/components/custom-tools/manage-docs-modal/ManageDocsModal.jsx (1)
593-600:DIRECT_UPLOAD_TYPESis re-created on everybeforeUploadcall inside theonloadcallback.This is a minor allocation concern. Consider hoisting it to module scope as a constant since it never changes.
Proposed refactor
Move it outside the component, near the top of the file:
const DIRECT_UPLOAD_TYPES = new Set([ "application/pdf", "text/plain", "text/csv", "application/vnd.ms-excel", "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet", "application/vnd.ms-excel.sheet.macroenabled.12", ]);backend/utils/file_storage/helpers/prompt_studio_file_helper.py (1)
86-116:upload_converted_for_idelargely duplicatesupload_for_ide.The two methods share the same storage-init and path-resolution logic. Consider extracting a private helper that accepts a subdirectory parameter (e.g.,
Nonefor main dir,"converted"for the converted subdir) to reduce duplication.frontend/src/components/custom-tools/document-manager/DocumentManager.jsx (1)
357-370: Consider disabling the "PDF View" tab for non-PDF files instead of showing a placeholder.Users can still manually switch to the PDF View tab for non-PDF files and see a "preview not available" placeholder. Disabling or hiding the tab would be a cleaner UX, though the current approach is functional.
| file_type = uploaded_file.content_type | ||
| # Convert non-PDF files | ||
|
|
||
| if file_converter_plugin and file_type != "application/pdf": |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
# First, let's see the full context around line 577 in views.py
find . -name "views.py" -path "*prompt_studio_core_v2*" -type fRepository: Zipstack/unstract
Length of output: 115
🏁 Script executed:
# Get the full context around line 577
wc -l backend/prompt_studio/prompt_studio_core_v2/views.pyRepository: Zipstack/unstract
Length of output: 117
🏁 Script executed:
# Read the context around line 577
sed -n '550,610p' backend/prompt_studio/prompt_studio_core_v2/views.pyRepository: Zipstack/unstract
Length of output: 3014
🏁 Script executed:
# Search for should_convert_to_pdf implementation
rg -n "should_convert_to_pdf" --type=py -B3 -A5Repository: Zipstack/unstract
Length of output: 1108
🏁 Script executed:
# Look for file_converter plugin definition
rg -n "file_converter" --type=py -A2 | head -100Repository: Zipstack/unstract
Length of output: 2773
🏁 Script executed:
# Check for MIME type validation or magic byte checking
rg -n "content_type|magic|sniff" --type=py | grep -v test | grep -v migration | head -50Repository: Zipstack/unstract
Length of output: 5750
🏁 Script executed:
# Look for allowed_content_types usage
rg -n "allowed_content_types" --type=py -B2 -A2Repository: Zipstack/unstract
Length of output: 3946
🏁 Script executed:
# Find the file_converter plugin implementation
find . -path "*file_converter*" -name "*.py" -type f | grep -v test | grep -v migration | head -20Repository: Zipstack/unstract
Length of output: 43
🏁 Script executed:
# Check FileKey.FILE_UPLOAD_ALLOWED_MIME to understand supported formats
rg -n "FILE_UPLOAD_ALLOWED_MIME|ExtentedFileInformationKey" --type=py -A10 | head -80Repository: Zipstack/unstract
Length of output: 4511
🏁 Script executed:
# Look for should_convert_to_pdf implementation in file_converter
find . -path "*file_converter*" -name "*.py" -type f -exec grep -l "should_convert_to_pdf" {} \;Repository: Zipstack/unstract
Length of output: 43
🏁 Script executed:
# Check the upload_for_ide method more completely to see if there's validation
sed -n '568,620p' backend/prompt_studio/prompt_studio_core_v2/views.pyRepository: Zipstack/unstract
Length of output: 2535
🏁 Script executed:
# Search for where file_converter plugin is loaded and its definition
rg -n "get_plugin.*file_converter" --type=py -B2 -A2Repository: Zipstack/unstract
Length of output: 1079
🏁 Script executed:
# Check if there's any file extension or magic byte validation in upload methods
rg -n "upload_for_ide|upload_converted_for_ide" --type=py -B3 -A8 | head -60Repository: Zipstack/unstract
Length of output: 4941
🏁 Script executed:
# Look for FileViewTypes to understand the upload logic better
rg -n "FileViewTypes" --type=py -B2 -A2Repository: Zipstack/unstract
Length of output: 2391
🏁 Script executed:
# Check FileUploadIdeSerializer to see if there's MIME validation
rg -n "FileUploadIdeSerializer" --type=py -A15Repository: Zipstack/unstract
Length of output: 5493
🏁 Script executed:
# Look at the file_converter plugin to understand should_convert_to_pdf behavior
find . -path "*plugins*" -name "*.py" -type f | xargs grep -l "should_convert_to_pdf" 2>/dev/nullRepository: Zipstack/unstract
Length of output: 45
🏁 Script executed:
# Check if there's any magic byte detection or file extension validation
rg -n "magic\.|mimetypes|sniff" backend/prompt_studio/prompt_studio_core_v2/ --type=pyRepository: Zipstack/unstract
Length of output: 191
🏁 Script executed:
# Get the FILE_UPLOAD_ALLOWED_EXT and details about FileKey constants
rg -n "FILE_UPLOAD_ALLOWED" --type=py -B2 -A2 | head -100Repository: Zipstack/unstract
Length of output: 2646
🏁 Script executed:
# Check if ExtentedFileInformationKey is from a plugin
find . -path "*file_converter*" -name "constants.py" -type fRepository: Zipstack/unstract
Length of output: 43
🏁 Script executed:
# Search for ExtentedFileInformationKey definition
rg -n "ExtentedFileInformationKey\|class.*FileInformationKey" --type=py -A15 | head -150Repository: Zipstack/unstract
Length of output: 43
🏁 Script executed:
# Try to find ExtentedFileInformationKey in the entire codebase
rg -n "ExtentedFileInformationKey" --type=py -B5 -A15 | head -120Repository: Zipstack/unstract
Length of output: 1953
🏁 Script executed:
# Look for plugins directory and file_converter
find . -type d -name "plugins" | head -5Repository: Zipstack/unstract
Length of output: 138
🏁 Script executed:
# Search for get_extented_file_information_key implementation
rg -n "get_extented_file_information_key" --type=py -A8Repository: Zipstack/unstract
Length of output: 973
🏁 Script executed:
# Look for the file_converter plugin constants
find ./backend/plugins -path "*file_converter*" -name "constants.py" -type fRepository: Zipstack/unstract
Length of output: 43
🏁 Script executed:
# Look at the structure of the plugins directory
ls -la ./backend/plugins/ | head -30Repository: Zipstack/unstract
Length of output: 661
🏁 Script executed:
# Search for file_converter in the plugins directory
find ./backend/plugins -path "*file_converter*" -type f -name "*.py" | head -20Repository: Zipstack/unstract
Length of output: 43
🏁 Script executed:
# Look for file_converter in the workers plugins
find ./workers/plugins -path "*file_converter*" -type f -name "*.py" | head -20Repository: Zipstack/unstract
Length of output: 43
🏁 Script executed:
# Search the entire codebase for file_converter plugin
rg -n "file_converter" --type=py | grep -i "constant\|mime\|extension" | head -30Repository: Zipstack/unstract
Length of output: 183
🏁 Script executed:
# Check if file_converter is an external plugin by searching plugin references
rg -n "should_convert_to_pdf\|get_extented_file_information_key" --type=py -B3 -A10Repository: Zipstack/unstract
Length of output: 43
🏁 Script executed:
# Look for FileValidator implementation
rg -n "class FileValidator" --type=py -A20Repository: Zipstack/unstract
Length of output: 1521
🏁 Script executed:
# Check if there's any file extension checking alongside MIME type
rg -n "FILE_UPLOAD_ALLOWED_EXT" --type=py -B2 -A2Repository: Zipstack/unstract
Length of output: 1787
🏁 Script executed:
# Look for any additional validation in the upload_for_ide or before line 577
sed -n '560,577p' backend/prompt_studio/prompt_studio_core_v2/views.pyRepository: Zipstack/unstract
Length of output: 838
🏁 Script executed:
# Look at the FileValidator implementation more closely
cat -n backend/utils/FileValidator.py | head -100Repository: Zipstack/unstract
Length of output: 4505
🏁 Script executed:
# Check what happens in should_convert_to_pdf - does it do its own MIME checking?
rg -n "process_file|should_convert_to_pdf" --type=py -A5 | head -60Repository: Zipstack/unstract
Length of output: 4561
🏁 Script executed:
# Look for any file extension-based checks alongside MIME type checks
rg -n "\.split.*\." backend/prompt_studio/prompt_studio_core_v2/views.pyRepository: Zipstack/unstract
Length of output: 43
🏁 Script executed:
# Check what FILE_UPLOAD_ALLOWED_EXT contains (look for ExtentedFileInformationKey constant definition)
rg -n "FILE_UPLOAD_ALLOWED_EXT\|FILE_UPLOAD_ALLOWED_MIME" --type=py -B3 -A3Repository: Zipstack/unstract
Length of output: 43
🏁 Script executed:
# Verify the file extension is being checked
sed -n '192,204p' backend/prompt_studio/prompt_studio_core_v2/serializers.pyRepository: Zipstack/unstract
Length of output: 509
🏁 Script executed:
# Check if file_name contains the actual extension being used
sed -n '571,580p' backend/prompt_studio/prompt_studio_core_v2/views.pyRepository: Zipstack/unstract
Length of output: 567
🏁 Script executed:
# Look at what happens with the file extension in the rest of the code
rg -n "file_name\|\.split\|splitext" backend/prompt_studio/prompt_studio_core_v2/views.py | head -20Repository: Zipstack/unstract
Length of output: 43
🏁 Script executed:
# Verify what constants FileKey resolves to by checking imports and fallback
cat -n backend/prompt_studio/prompt_studio_core_v2/serializers.py | sed -n '28,34p'Repository: Zipstack/unstract
Length of output: 321
🏁 Script executed:
# Check the FileValidator more thoroughly - especially the magic.from_buffer call
sed -n '72,85p' backend/utils/FileValidator.pyRepository: Zipstack/unstract
Length of output: 670
🏁 Script executed:
# Key insight: Check if file_type is being used for conversion decision after validation
# Let's see if should_convert_to_pdf receives the browser-supplied MIME or actual file MIME
sed -n '575,590p' backend/prompt_studio/prompt_studio_core_v2/views.pyRepository: Zipstack/unstract
Length of output: 968
🏁 Script executed:
# Search for ExtentedFileInformationKey to understand what file types are actually supported
rg -n "FILE_UPLOAD_ALLOWED" --type=py -B5 -A5 | grep -A5 -B5 "EXT\|MIME" | head -80Repository: Zipstack/unstract
Length of output: 4765
🏁 Script executed:
# Check what attributes are available on uploaded_file after validation
rg -n "class InMemoryUploadedFile\|class UploadedFile" --type=py -A10 | head -40Repository: Zipstack/unstract
Length of output: 43
🏁 Script executed:
# Confirm: does FileValidator's magic.from_buffer result get stored anywhere?
rg -n "magic\.from_buffer\|\.content_type" backend/utils/FileValidator.pyRepository: Zipstack/unstract
Length of output: 43
🏁 Script executed:
# Check if there's a way to get the actual detected MIME type at line 577
rg -n "file_type\|content_type" backend/prompt_studio/prompt_studio_core_v2/views.py | head -20Repository: Zipstack/unstract
Length of output: 43
🏁 Script executed:
# Get the last few lines of FileValidator to see the full validation flow
sed -n '72,110p' backend/utils/FileValidator.pyRepository: Zipstack/unstract
Length of output: 1672
🏁 Script executed:
# Check if magic.from_buffer result is stored anywhere
grep -n "magic\|_check_file" backend/utils/FileValidator.pyRepository: Zipstack/unstract
Length of output: 481
🏁 Script executed:
# Understand the complete validation: check the __call__ method
sed -n '100,120p' backend/utils/FileValidator.pyRepository: Zipstack/unstract
Length of output: 656
file_type at line 577 relies on browser-supplied MIME, not actual file content detection.
The FileValidator class performs server-side MIME sniffing using magic.from_buffer() to validate uploads, which correctly identifies actual file types. However, after validation passes, this detected MIME type is discarded. The conversion logic at line 577 then uses uploaded_file.content_type (browser-supplied), not the server-detected MIME.
This creates an inconsistency: a CSV file with content_type="application/octet-stream" (common browser behavior) will pass validation (actual MIME detected as text/csv), but the conversion decision will use the unreliable browser value. If the file converter plugin supports multiple formats beyond PDF, the conversion-vs-direct-upload logic could behave unexpectedly. Consider either storing the detected MIME type during validation or re-detecting it before the conversion decision.
🤖 Prompt for AI Agents
In `@backend/prompt_studio/prompt_studio_core_v2/views.py` at line 577, The
conversion branch uses the browser-supplied uploaded_file.content_type
(referenced as file_type at the if with file_converter_plugin) instead of the
server-detected MIME from FileValidator, causing inconsistent behavior; fix by
persisting the server-detected MIME (e.g., add a property or return value like
detected_mime on FileValidator or its validate() result) or re-run
magic.from_buffer() in the view before the conversion decision, then use that
server-detected MIME (detected_mime) in the conditional (the same place that
checks file_converter_plugin and file_type) so conversion vs direct upload is
based on the validated content type.
jagadeeswaran-zipstack
changed the title
jagadeeswaran-zipstack
changed the title
Test ResultsSummary
Runner Tests - Full Report
SDK1 Tests - Full Report
|
jagadeeswaran-zipstack
requested review from
Deepak-Kesavan,
athul-rs,
harini-venkataraman and
vishnuszipstack
|
There was a problem hiding this comment.
Actionable comments posted: 2
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
frontend/src/components/custom-tools/document-manager/DocumentManager.jsx (1)
155-161:⚠️ Potential issue | 🟠 MajorBug:
fileDatais not reset when switching documents, causing stale mimeType.When
selectedDocchanges,fileUrlandextractTxtare reset butfileData(which holdsmimeType) is not. This means between the reset and the new response arriving (or if the new request fails), the previous document'smimeTypepersists. SinceisContentAvailable(line 482-485) andrenderDoc(line 358) both depend onfileData.mimeType, stale state can cause incorrect rendering — e.g., showing the non-PDF placeholder for what should be a PDF, or marking content as available when the fetch actually failed.Proposed fix
useEffect(() => { setFileUrl(""); setExtractTxt(""); + setFileData({}); Object.keys(viewTypes).forEach((item) => { handleFetchContent(viewTypes[item]); }); }, [selectedDoc]);Also reset in
handleFetchContent:const handleFetchContent = (viewType) => { if (viewType === viewTypes.original) { setFileUrl(""); setFileErrMsg(""); + setFileData({}); }
🤖 Fix all issues with AI agents
In `@frontend/src/components/custom-tools/document-manager/DocumentManager.jsx`:
- Around line 482-485: The isContentAvailable prop can remain as written but you
must prevent stale fileData from making it true after a failed fetch: update
handleGetDocsError to clear the previous fileData (e.g., reset fileData to
null/empty object) and/or set an explicit fetch-failed flag (e.g.,
fileFetchError) and include that in the isContentAvailable check; specifically,
modify handleGetDocsError to reset the state that holds the prior mimeType
(refer to fileData and handleGetDocsError) so that isContentAvailable (which
depends on fileUrl and fileData.mimeType) cannot be true after a fetch failure.
- Around line 259-261: The FileReader.onerror handler currently throws an
exception inside the async callback (reader.onerror) which becomes unhandled;
instead, update the component's error state inside that callback (e.g., call
your state setter such as setError or setFileError with a descriptive message
and the event/error details) and avoid throwing — locate reader.onerror in
DocumentManager.jsx and replace the throw new Error("Fail to load the file")
with a call that sets the component error state and optionally logs the event so
the UI can render the error.
🧹 Nitpick comments (1)
frontend/src/components/custom-tools/document-manager/DocumentManager.jsx (1)
356-370: UnuseddocNameparameter inrenderDoc.After the refactor to mimeType-based rendering,
docNameis no longer used insiderenderDoc. Consider removing it to avoid confusion.Proposed cleanup
- const renderDoc = (docName, fileUrl, highlightData) => { + const renderDoc = (fileUrl, highlightData) => {And update the call site at line 489:
- {renderDoc(selectedDoc?.document_name, blobFileUrl, highlightData)} + {renderDoc(blobFileUrl, highlightData)}
| reader.onerror = () => { | ||
| throw new Error("Fail to load the file"); | ||
| }; |
There was a problem hiding this comment.
Unhandled throw inside reader.onerror callback.
Throwing inside an async callback (FileReader.onerror) won't be caught by any surrounding try/catch or .catch() — it results in an unhandled error. This should set the error state instead.
Proposed fix
reader.onerror = () => {
- throw new Error("Fail to load the file");
+ setFileErrMsg("Failed to load the file");
};📝 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.
| reader.onerror = () => { | |
| throw new Error("Fail to load the file"); | |
| }; | |
| reader.onerror = () => { | |
| setFileErrMsg("Failed to load the file"); | |
| }; |
🤖 Prompt for AI Agents
In `@frontend/src/components/custom-tools/document-manager/DocumentManager.jsx`
around lines 259 - 261, The FileReader.onerror handler currently throws an
exception inside the async callback (reader.onerror) which becomes unhandled;
instead, update the component's error state inside that callback (e.g., call
your state setter such as setError or setFileError with a descriptive message
and the event/error details) and avoid throwing — locate reader.onerror in
DocumentManager.jsx and replace the throw new Error("Fail to load the file")
with a call that sets the component error state and optionally logs the event so
the UI can render the error.
| isContentAvailable={ | ||
| fileUrl?.length > 0 || | ||
| (fileData.mimeType && fileData.mimeType !== "application/pdf") | ||
| } |
There was a problem hiding this comment.
isContentAvailable can be true even when document fetch fails.
This condition considers content available when fileData.mimeType is set to a non-PDF value. However, since fileData isn't cleared on fetch errors (see handleGetDocsError), a previously loaded non-PDF document's mimeType can leak into the next document's state. This ties back to the stale fileData issue above — once that's fixed, this logic should work correctly.
🤖 Prompt for AI Agents
In `@frontend/src/components/custom-tools/document-manager/DocumentManager.jsx`
around lines 482 - 485, The isContentAvailable prop can remain as written but
you must prevent stale fileData from making it true after a failed fetch: update
handleGetDocsError to clear the previous fileData (e.g., reset fileData to
null/empty object) and/or set an explicit fetch-failed flag (e.g.,
fileFetchError) and include that in the isContentAvailable check; specifically,
modify handleGetDocsError to reset the state that holds the prior mimeType
(refer to fileData and handleGetDocsError) so that isContentAvailable (which
depends on fileUrl and fileData.mimeType) cannot be true after a fetch failure.
2 participants
What
Add support for CSV, TXT, and Excel files in Prompt Studio for enterprise users. These file types are stored as-is (no PDF conversion) and displayed in Raw View only.
Why
How
converted/subdirectory for previewfetch_contents_idechecksconverted/{name}.pdffirst for enterprise; falls back to original fileFiles Changed:
backend/prompt_studio/prompt_studio_core_v2/views.py- Upload stores original + converted; fetch resolves converted PDFbackend/utils/file_storage/helpers/prompt_studio_file_helper.py- Addconverted/dir,upload_converted_for_ide(), handle CSV/Excel MIMEfrontend/.../DocumentManager.jsx- Use mimeType for rendering, auto-switch to Raw Viewfrontend/.../ManageDocsModal.jsx- Allow CSV/TXT/Excel in upload validationCan this PR break any existing features. If yes, please list possible items. If no, please explain why.
No - Changes are enterprise-gated via
get_plugin("file_converter"). OSS behavior is unchanged (PDF-only). Existing DOCX/image/PPT uploads now store both original AND converted PDF, maintaining backward compatibility for preview while enabling original file indexing.Database Migrations
None
Env Config
None
Related Issues or PRs
Notes on Testing
file.docxin main dir +converted/file.pdfexists → PDF View shows PDF → Index uses original DOCXdata.csvin main dir, no converted file → Auto-switches to Raw View → Index uses CSVconverted/directoryChecklist
I have read and understood the Contribution Guidelines.