v0.8.1

🚀 v0.8.1 Release – Fixed SSE & HTTP Transport, Stable Prompts

We’re pleased to announce v0.8.1 — a critical patch release addressing key stability issues in the communication layer and prompt handling.

✨ What’s Fixed:

• ✅ SSE & HTTP Transport Bug Resolved

  • Fixed malformed message flow during real-time streaming (SSE)
  • Ensured proper request/response handling in HTTP mode
  • Now stable for long-running sessions and high-throughput environments

• ✅ Tool Prompt Formatting Fixed

  • Corrected malformed prompt structure that caused invalid JSON responses
  • Restored proper JSON output format for all tool interactions
  • Prevents parsing errors in downstream systems

🔧 Impact:

• Improved reliability for real-time AI workflows
• Cleaner, predictable responses in integrations and APIs
• Better compatibility with Open WebUI, custom backends, and external tools

👉 Upgrade now for a smoother, more stable experience:
GitHub v0.8.1

Thank you for your feedback — your reports help us ship a rock-solid tool.
Stability is now the priority.

v0.8.0

🚀 v0.8.0 Release – Production-Ready AI Document Generation & Editing

We’re proud to announce the official v0.8.0 release — a major leap forward in enterprise-grade, AI-powered document automation. This version is now stable, scalable, and ready for real-world deployment in professional environments.

✨ What’s New in v0.8.0:

✅ Multi-User Support (Enterprise-Grade)

• Fully enabled and tested for team collaboration
• Each user has isolated session context and secure access
• Ideal for internal tools, SaaS platforms, and shared workspaces

✅ Robust & Stable Document Workflows

• AI-powered editing, reviewing, and commenting now fully functional across .docx, .pptx, and .xlsx
• Context-aware text insertion with precise ID tracking
• Enhanced logging, versioning, and error resilience

✅ Secure & Modern Auth

• MCPO API Key is deprecated — no longer required (see readme)
• Authentication now handled via session header in tool config
• Each user gets a unique, secure session — perfect for scalable, multi-tenant systems

✅ HTTP Transport Layer Live

• Enables seamless integration with external backends, APIs, and enterprise systems
• Paves the way for future microservices architecture

🛠️ Fix

✅ LiteLLM Integration Fixed

✅ PPTX Editing Fully Fixed

• No more layout issues or content misplacement
• Slides now edit reliably with correct positioning, image handling, and formatting preservation

🔧 What’s Coming Next (v1.0.0):

• 🧼 Full code refactoring — separation of core logic from functions to improve modularity, performance, and maintainability
• 📦 Lightweight, plug-and-play modules
• 🚀 Optimized for CI/CD, containerization, and enterprise scaling

👉 Get v0.8.0 now — stable, secure, and production-ready:
GitHub v0.8.0

💬 Join the community — help us shape the future of AI document tools:
Discord | GitHub Issues

Thank you for building with us.
v0.8.0 is not just a release — it’s a foundation.

v0.8.0-rc2

🚀 v0.8.0-rc2 Release – Final Pre-Release for v0.8.0 Stable
We’re thrilled to announce v0.8.0-rc2, the final release candidate for the upcoming v0.8.0 stable version — a major milestone in the evolution of AI-powered document generation and editing!

✨ What’s New & Fixed in v0.8.0-rc2:

• ✅ Fixed random text placement in .docx and .pptx — text now inserts precisely at the intended location
• 🔐 Robust text ID management system — ensures structural integrity, consistent rendering, and prevents file corruption
• 🛠️ Fully restored & enhanced document review workflow — comments, feedback, and collaborative editing are now stable and context-aware
• 📊 Improved context tracking & versioning — enhanced logging and stability across all Office formats (.docx, .pptx, .xlsx)
• 🔄 Multi-user support — ready for collaborative environments and shared workspaces
• 🎯 PPTX enhancements — improved layout handling, better image placement, and smoother slide integration
• ⚡ Added HTTP transport layer — enables seamless backend communication and integration with external systems
• 🤖 Fixed Lite LLM integration — resolved compatibility issues for lightweight inference workflows
• 🔑 API Key MCPO is now deprecated — no longer required for authentication.
  ➤ Authentication is now done via the session header in the tool configuration.
  ➤ Each user must have their own unique API key, managed through the session system.

⚠️ Known Issues (Under Investigation):

1. Occasional PPTX editing failures under specific scenarios — actively being monitored and optimized.
2. Excel template issue: The LLM does not currently recognize that the first editable cell is not A1 (due to header rows or offset).
   ➤ If formulas are inserted, they may be misaligned.
   ➤ Workaround: Use the edit function post-generation to correct formula references manually or via structured update.

🔧 This release is ready for final validation — ideal for production testing, deployment, and final feedback from early adopters.

👉 Test v0.8.0-rc2 now & help us ship v0.8.0 stable:
GitHub v0.8.0-rc2

💬 Your feedback is critical — report any edge cases, performance quirks, or integration issues in the repository.
Together, we’re building a rock-solid, AI-native document tool for the future.

Thank you for being part of this journey — v0.8.0 is almost here!

v0.8.0-beta

🚀 v0.8.0-beta Release – Stable Document Editing Now Live!

We’re excited to announce v0.8.0-beta — a major step forward in reliability and precision for AI-powered document editing!

✨ What’s New & Fixed:

• ✅ Fixed random text placement in .docx and .pptx — text now inserts at the correct location
• 🔐 Robust text ID management — ensures consistent structure and avoids corruption
• 🛠️ Revamping of document review workflow — commenting & feedback now fully functional again after alpha issues
• 📊 Improved context tracking and tool stability across all Office formats

🔧 This release is ready for real-world testing — ideal for developers, testers, and early adopters.

👉 Try the beta & help us ship v0.8.0 stable: GitHub v0.8.0-beta
💬 Report any edge cases — your feedback is critical to the final release!

v0.7.1

🚀 v0.7.1 Release – Critical Fix for Built-in MCPO Compatibility

A critical issue affecting the built-in MCPO functionality has been resolved in v0.7.1.

🔧 What was fixed:

• A breaking change in the latest Open WebUI MCPo image build was causing core functionality to fail.
• This update ensures full compatibility with the latest MCPo runtime, restoring stable and reliable behavior.

✅ No action required if you're using the latest version — the fix is included.

👉 Get the fix now: GitHub v0.7.1

v0.8.0-alpha

🚀 v0.8.0-alpha Release – Document Editing Now Live!

We’re proud to announce v0.8.0-alpha — the first release with direct document editing for .docx and .pptx!

✨ What’s New:

• ✏️ AI-powered content editing — rewrite sections, update text, restructure content in-place
• 🔗 Seamless integration with Open WebUI Files API
• 📊 Enhanced logging & version tracking (tool_version: v0.8.0-alpha)

⚠️ Known Issue:

• Text may be inserted in the wrong location in .docx and .pptx files (under investigation) — minor regression in alpha phase

🔧 Next Steps:

• Fix insertion logic & improve context-aware positioning

👉 Try it now & help test: GitHub v0.8.0-alpha
💬 Report issues in the repo — your feedback shapes the final release!

v0.7.0

🚀 Release Notes – v0.7.0

🔥 Major Release: Document Review & Open WebUI API Integration

We’re thrilled to announce v0.7.0 — a powerful leap in document intelligence, collaboration, and future-proof architecture. This release introduces native document revision support (Word, Excel, PowerPoint), integrates with the Open WebUI Files API, and delivers major performance and stability improvements.

---

✨ New Features

📝 Document Revision Support (Thanks to @Baronco: https://github.com/Baronco/GenFilesMCP)

Revise your documents directly in the chat using AI-powered feedback and edits — all natively supported:

Format	Method	Notes
Word (.docx)	Native Word comments	Fully compatible with Microsoft Word and other editors
Excel (.xlsx)	Legacy comment notes (cell notes)	Preserves original metadata and structure
PowerPoint (.pptx)	Top-of-slide TextBox (simulated comment)	Visual feedback without breaking layout

💡 How it works:

1. Receive file metadata in the chat
2. Tool detects document type
3. AI analyzes full context and generates edits
4. Changes are applied and returned via download URL
5. Temporary file is automatically deleted

✅ No manual file handling required — everything is seamless.

---

🔧 Introducing Open WebUI Files API Integration

This is a major step toward deprecating the standalone file server.

• Now supports direct file upload, retrieval, and management via the Open WebUI Files API.
• Enables future fully integrated, serverless workflows.
• Eliminates dependency on external file storage for most use cases.

---

📁 Required Setup for Document Review

To enable revision, you’ll need:

1. Add the Files Metadata Injector Function

   • Available here: Files Metadata Injector • Open WebUI Community
   • Or place it in your tool’s functions/ directory (Docker path: /rootPath/functions)

2. Set two new environment variables:

   OWUI_URL=https://myAI.myDomain.com
   JWT_SECRET=your_api_key_or_jwt_token

   • OWUI_URL: Your Open WebUI instance URL (e.g., http://localhost:3000 or https://openwebui.yourdomain.com)
   • JWT_SECRET: Generated from User Settings > API Keys in Open WebUI

3. **Add this section in model prompt (next to current model prompt about FileGenTool):

### Office document revision (.docx or .xlsx or .pptx)
If the user asks you to review a Word document with comments:
**Review workflow (mandatory):**
  1. Call `tool_full_context_document_post` → to retrieve element indices.
  2. Call `tool_review_document_post` → pass the list of tuples `(element_index, comment)`.
    Never add extra content in step 3.
    For XLSX files, use the "index" field (e.g., "B3") to reference cells in `tool_review_document_post`.

---

🛠️ Improvements

🔧 Code Refactoring & Tool Descriptions

• Full codebase cleanup and modularization
• Optimized and clearer tool descriptions for better LLM understanding
• Improved tool naming and parameter clarity

📊 Enhanced Logging

• Added version tracking: tool_version: v0.7.0 in all logs
• More detailed logs for:
  • File upload/download
  • Revision workflow steps
  • API communication with Open WebUI
• Better debugging and monitoring

---

🐞 Bug Fixes

• ✅ Fixed incorrect URL formatting in output — now properly returned in https://... format
• ✅ Corrected incomplete/inaccurate logs — all key events now logged with full context
• ✅ Resolved edge cases in file handling during revision

---

📌 Future Direction

🔜 Next: Advanced Document Modification
We’re already working on direct in-place editing of document content — not just comments.
This will allow AI to:

• Rewrite sections
• Update tables and charts
• Reformat entire documents
• Apply consistent styling

Stay tuned — the next release will transform how you interact with AI-generated documents.

---

📌 In the Pipeline (No Release Date Yet)

• 📚 Enhanced documentation — actively being built
• 📄 Refactoring of PDF construction — aiming for better layout, font handling, and performance

---

🙌 Thank You

Huge thanks to:

• @Baronco (GitHub) for creating the foundational GenFilesMCP tool that inspired this release
• @MarouaneZhani (GitHub) for extensive testing and feedback
• The Open WebUI community on Reddit, GitHub, and Discord — your input drives innovation

---

📌 Ready to Upgrade?

👉 Check the full changelog: GitHub v0.7.0
👉 Join Discord for early feedback and testing
👉 Open an issue or PR if you have suggestions!

---

© 2025 MCP_File_Generation_Tool | MIT License

v0.6.1

What's Changed

• 0.6.1 by @GlisseManTV in #26

Prompt optimization
Tool call more clear for LLM.

Full Changelog: v0.6.0...v0.6.1

v0.6.0

🚀 Release Notes – v0.6.0

🔥 Major Release: Smarter, Faster, More Powerful

We’re excited to announce v0.6.0 — a major leap forward in performance, flexibility, and usability for the MCPO-File-Generation-Tool. This release introduces a streaming HTTP server, a complete tool refactoring, Pexels image support, native document templates, and significant improvements to layout and stability.

---

✨ New Features

📦 Docker Image with SSE Streaming (Out-of-the-Box HTTP Support)

Introducing:
👉 ghcr.io/glissemantv/file-gen-sse-http:latest

This new image enables streamable, real-time file generation via SSE (Server-Sent Events) — perfect for interactive workflows.

✅ Key benefits:

• Works out of the box with OpenWebUI 0.6.31
• Fully compatible with MCP Streamable HTTP
• No need for an MCPO API key (the tool runs independently)
• Still requires the file server (separate container) for file downloads

---

🖼️ Pexels as an Image Provider

Now you can generate images directly from Pexels using:

• IMAGE_SOURCE: pexels
• PEXELS_ACCESS_KEY: your_api_key (get it at https://www.pexels.com/api)

Supports all existing prompt syntax: ![Recherche](image_query: futuristic city)

---

📄 Document Templates (Word, Excel, PowerPoint)

We’ve added professional default templates for:

• .docx (Word)
• .xlsx (Excel)
• .pptx (PowerPoint)

📍 Templates are included in the container at the default path:
/app/templates/Default_Templates/

🔧 To use custom templates:

1. Place your .docx, .xlsx, or .pptx files in a shared volume
2. Set the environment variable:

   DOCS_TEMPLATE_DIR: /path/to/your/templates

✅ Thanks to @MarouaneZhani (GitHub) for the incredible work on designing and implementing these templates — they make your outputs instantly more professional!

---

🛠️ Improvements

🔧 Complete Code Refactoring – Only 2 Tools Left

We’ve reduced the number of available tools from 10+ down to just 2:

• create_file
• generate_archive

✅ Result:

• 80% reduction in tool calling tokens
• Faster execution
• Cleaner, more maintainable code
• Better compatibility with LLMs and MCP servers

📌 This change is potentially breaking — you must update your model prompts accordingly.

---

🎯 Improved Image Positioning in PPTX

Images now align perfectly with titles and layout structure — no more awkward overlaps or misalignment.

• Automatic placement: top, bottom, left, right
• Dynamic spacing based on content density

---

⚠️ Breaking Change

🔄 Tool changes require prompt updates
Since only create_file and generate_archive are now available, you must update your model prompts to reflect the new tool set.
Old tool names (e.g., export_pdf, upload_file) will no longer work.

---

📌 In the Pipeline (No Release Date Yet)

• 📚 Enhanced documentation — now being actively built
• 📄 Refactoring of PDF generation — aiming for better layout, font handling, and performance

---

🙌 Thank You

Huge thanks to:

• @MarouaneZhani for the stunning template design and implementation
• The OpenWebUI community on Reddit, GitHub, and Discord for feedback and testing
• Everyone who helped shape this release through real-world use

---

📌 Don’t forget to run the file server separately for downloads.

---

📌 Ready to upgrade?

👉 Check the full changelog: GitHub v0.6.0
👉 Join Discord for early feedback and testing
👉 Open an issue or PR if you have suggestions!

---

© 2025 OWUI_File_Gen_Export | MIT License

v0.5.0

Release Notes - v0.5.0

🚀 Overview

This release marks a major leap forward in AI-powered file generation, introducing native DOCX support, local image generation via Automatic1111 Stable Diffusion UI, and significant improvements to data handling, logging, and extensibility. Designed for developers, power users, and teams building intelligent document workflows.

---

✨ New Features

🔹 DOCX (Microsoft Word) Support

• Full integration for generating .docx files using python-docx.
• Supports rich text, tables, headings, lists, and custom styles.
• Ideal for professional reports, contracts, and structured documentation.

🔹 Local Image Generation via Automatic1111 Stable Diffusion UI

• Add local AI image generation as a provider directly in the workflow.
• Configured via environment variables:
  • IMAGE_SOURCE: "unsplash" to use Unsplash or "local_sd" to use your local Stable Diffusion instance (default is "unsplash")
  • LOCAL_SD_URL: URL of your local Stable Diffusion instance (if using local_sd) (no default value, mandatory if local_sd is used above)
  • LOCAL_SD_USERNAME: Username of your local Stable Diffusion instance (if any) (no default value, not mandatory)
  • LOCAL_SD_PASSWORD: Password of your local Stable Diffusion instance (if any) (no default value, not mandatory)
  • LOCAL_SD_DEFAULT_MODEL: Default model to use (if any) (default sd_xl_base_1.0.safetensors, not mandatory)
  • LOCAL_SD_STEPS: Number of steps to use (default 20, not mandatory)
  • LOCAL_SD_WIDTH: Width of the image to generate (default 512, not mandatory)
  • LOCAL_SD_HEIGHT: Height of the image to generate (default 512, not mandatory)
  • LOCAL_SD_CFG_SCALE: CFG scale to use (default 1.5, not mandatory)
  • LOCAL_SD_SCHEDULER: Scheduler to use (default Karras, not mandatory)
  • LOCAL_SD_SAMPLE: Sampler to use (default Euler a, not mandatory)
• Enables privacy-first, offline image generation with full control over content and quality.

🔹 Shared Output Path Logic

• Unified file output path across all providers (local, remote, Unsplash, SD).
• Ensures consistent file linking and reduces configuration complexity.
• Works seamlessly in Docker, local, and hybrid setups.

---

🛠️ Improvements

🔹 Enhanced Data Processing Logic

• Refactored core data pipeline for better performance, scalability, and error resilience.
• Improved handling of nested structures, metadata, and file dependencies.

🔹 Debug-Level Logging

• Added detailed debug logs throughout the system:
  • Request flow tracing
  • Image generation status
  • File export lifecycle
  • Provider-specific diagnostics
• Helps identify issues early during development and testing.

🔹 Multilingual Log Improvements

• Completed missing translations for key log messages.
• Supports consistent, clear feedback in both French and English (and future languages).

---

🐞 Bug Fixes

• ✅ Fixed incorrect error message in PDFs when Unsplash connection failed.
  • Now displays a clear, actionable error instead of a cryptic internal message.
  • Improved error handling at the integration layer.

---

📌 Future Roadmap (In Progress)

• 🔮 Support for additional image generation providers – Community poll upcoming to decide next integrations.
• 🔗 Out-of-the-box compatibility with third-party OpenAPI MCP servers – Under analysis and testing.
• 📝 Custom document templates – Design phase in progress (Markdown + Jinja2 support planned).
• 📚 Expanded documentation – Comprehensive guides, API reference, and best practices in development.

---

📦 Supported File Types

• ✅ .xlsx (Excel)
• ✅ .pdf
• ✅ .csv
• ✅ .zip
• ✅ .tar.gz
• ✅ .7z
• ✅ .docx — NEW
• ✅ Any other file type via custom export

---

📌 Migration Notes

• Existing configurations remain compatible.
• Update your .yaml file to include (if using sdUI):

      - IMAGE_SOURCE=local_sd
      - LOCAL_SD_URL=http://localhost:7860
      - LOCAL_SD_USERNAME=local_user
      - LOCAL_SD_PASSWORD=local_password
      - LOCAL_SD_DEFAULT_MODEL=sd_xl_base_1.0.safetensors
      - LOCAL_SD_STEPS=20
      - LOCAL_SD_WIDTH=512
      - LOCAL_SD_HEIGHT=512
      - LOCAL_SD_CFG_SCALE=1.5
      - LOCAL_SD_SCHEDULER=Karras
      - LOCAL_SD_SAMPLE=Euler a

• Ensure your Stable Diffusion server is running and accessible at the specified URL.

---

📢 Feedback & Contribution

We’re building this tool with the community — your input shapes the future!

---

© 2025 MCPO-File-Generation-Tool | MIT License