Repository navigation
v0.3.0 MCPO Support, Image Embedding & Structured DOCX Generation 🚀
This release introduces dual deployment modes (Direct MCP Server or MCPO-bridged), image embedding in Word documents, and an experimental structured DOCX builder that reduces model code-generation burden. It also adds support for multi-user environments through MCPO and improves compatibility with newer Open Web UI versions.
Highlights & Objectives 📌
What changed
- 🌉 MCPO Support: Added
stdiotransport mode to run behind MCPO, allowing teams to reuse existing MCPO deployments instead of running GenFilesMCP as a standalone container. - ⚡ Dual Deployment Modes:
- Direct streamable-http: Recommended for single-user setups (existing behavior) or if you do not have problems deploying a separate container for the MCP Server.
- stdio through MCPO: Recommended for multi-user environments with shared knowledge collections
- 🖼️ Image Embedding: New capability to embed images uploaded in chat directly into generated Word documents, enabling richer visual content without external links. Use the built-in
Chat Context Toolto attach images from the chat. - 📝 Structured DOCX Mode: Experimental
ENABLE_WORD_ELEMENT_FILLINGflag switches from Python code execution to template-based document building, letting models focus on content planning rather than coding. Added support for academic paper formatting with two-column document layouts usingENABLE_WORD_ELEMENT_FILLING=true - 🔧 New Environment Variables:
MCP_TRANSPORT: Choose betweenstreamable-httpandstdiofor MCPO integrationOWUI_API_KEY: Service account API key for MCPO deploymentsREVIEWER_AI_ASSISTANT_NAME: Customize the author name in DOCX review commentsENABLE_WORD_ELEMENT_FILLING: Toggle experimental structured Word generation
- 📚 Unified Knowledge Collection: Generated and reviewed files now use the same knowledge collection name (
KNOWLEDGE_COLLECTION_NAME), simplifying organization. - 🔄 Enhanced chat_context Tool: Updated to retrieve attached images from chat messages, supporting the new image embedding feature.
- ✅ Updated Compatibility: Tested with Open Web UI >=v0.8.8 (requires v0.6.42+ for full functionality with MCP).
Why
- Infrastructure Flexibility: Support both single-user direct connections and enterprise multi-user setups through MCPO without maintaining separate codebases.
- Model Compatibility: Reduce dependency on heavy Python code generation for document creation, improving reliability across different LLM capabilities. Check the benchmark results in the README for more details on which models perform best in structured mode.
- Visual Content: Enable documents that include actual images rather than just text descriptions or external URLs. This is only for DOCX generation.
- Multi-User Workflows: Allow teams to share generated documents through group-based knowledge collections in MCPO deployments.
Roadmap direction
- Continue refining the structured document builder to eventually replace code generation for all document types.
- Replace the
chat_contexttool withX-Open-WebUI-Chat-Idheader-based retrieval to support more file types and richer metadata. - Generation of PDF files with embedded images and structured content.
Deployment Options
Option A: Direct streamable-http
Pull the pre-built Docker image from GitHub Container Registry:
docker pull ghcr.io/baronco/genfilesmcp:v0.3.0Run the container:
docker run -d --restart unless-stopped -p YOUR_PORT:YOUR_PORT \
-e OWUI_URL="http://host.docker.internal:3000" \
-e PORT=YOUR_PORT \
-e REVIEWER_AI_ASSISTANT_NAME="GenFilesMCP" \
-e ENABLE_CREATE_KNOWLEDGE=false \
--name gen_files_mcp \
ghcr.io/baronco/genfilesmcp:v0.3.0One-line command (copy/paste):
docker run -d --restart unless-stopped -p 8016:8016 -e OWUI_URL="http://host.docker.internal:3000" -e PORT=8016 -e REVIEWER_AI_ASSISTANT_NAME="GenFilesMCP" -e ENABLE_CREATE_KNOWLEDGE=false --name gen_files_mcp ghcr.io/baronco/genfilesmcp:v0.3.0Option B: stdio through MCPO (Recommended for multi-user)
Add to your MCPO config.json:
{
"mcpServers": {
"genfilesmcp": {
"command": "uvx",
"args": [
"--from",
"git+https://github.com/Baronco/GenFilesMCP.git@v0.3.0",
"genfilesmcp"
],
"env": {
"MCP_TRANSPORT": "stdio",
"OWUI_URL": "http://host.docker.internal:3000",
"OWUI_API_KEY": "sk-your-api-key",
"ENABLE_CREATE_KNOWLEDGE": "true",
"KNOWLEDGE_COLLECTION_NAME": "Team 2",
"ENABLE_WORD_ELEMENT_FILLING": "true",
"REVIEWER_AI_ASSISTANT_NAME": "GenFilesMCP"
}
}
}
}🚨 Please follow the installation instructions step by step to avoid errors.
Full Documentation: README.md