🚀 Alpha Release - Core Features Working & Tested 🚀
Bridge the context gap between AI conversations and long-term projects.
ConvoCanvas transforms exported AI conversations into actionable content ideas and context summaries, solving the memory limitation problem that technical professionals face when working on complex, multi-session projects.
Alpha Release - Core Features Working! The MVP is functional with conversation analysis, decision tracking, and visualization capabilities tested and working. Still evolving with regular improvements. Suitable for experimentation and early adoption.
As a network engineer transitioning into DevOps and automation, I've broken my Linux installs more times than I can count building local LLMs and complex infrastructure. But the real wall I hit wasn't technical—it was context window limitations.
AI services like Claude and Gemini forget everything between sessions. For any long-term technical project, their memory is too short. I found myself constantly re-explaining complex setups, architectural decisions, and troubleshooting context.
ConvoCanvas is my external memory for AI conversations. It processes exported chats and generates:
- Context summaries to re-prime the next AI session
- Technical decision logs from conversation history
- Content ideas (LinkedIn posts, blog topics) extracted from learning processes
- Conversation themes to track project evolution
Real Impact: Recently used ConvoCanvas to maintain context across a 3-day MPLS automation pipeline troubleshooting session—something impossible with standard AI chat limits.
ConvoCanvas is evolving into a comprehensive productivity engine combining AI conversation processing with knowledge management:
- Enhanced Analysis: NLP-powered decision extraction with confidence scoring
- Interactive Mindmaps: Plotly-based decision flow visualizations
- Sentiment Analysis: Emotional context detection around decisions
- Technical Domain Classification: Auto-categorize conversations (AI/ML, DevOps, etc.)
- Feature Flags System: Environment-based feature toggles
- Conversation Processing: Parse and analyze AI chat exports
- GPU Acceleration: Optional GPU processing for high-end cards (RTX 4080/4090 tested)
- Local AI Integration: LM Studio compatibility for local models
- Canvas Generation: Visual output creation from conversations
- Automated Organization: Smart file categorization based on content
- Metadata Generation: Automatic tagging and frontmatter creation
- Obsidian Integration: Direct vault compatibility for knowledge bases
- GPU features require NVIDIA cards with 12GB+ VRAM (but gracefully disabled if unavailable)
- Alpha software - expect UI/UX improvements in future releases
- Some advanced features still experimental
- Best suited for technical users familiar with API endpoints
ConvoCanvas uses environment variables to control feature availability:
ENABLE_ENHANCED_ANALYSIS=true # Enable advanced NLP analysis
ENABLE_CANVAS_GENERATION=true # Enable visual output generation
ENABLE_NLP=true # Enable natural language processing
DISABLE_GPU=true # Force disable GPU acceleration- Export conversations using the SaveMyPhind browser extension by Hugo Collin
- Upload to ConvoCanvas backend (FastAPI + Python)
- Process conversations to extract key decisions, technical concepts, and learning moments
- Generate content ideas and context summaries for future sessions
---
config:
theme: redux-dark
look: neo
---
flowchart TB
subgraph Input["📥 Input Layer"]
A("🤖<br>AI Conversations")
B("📄<br>SaveMyPhind<br>Export")
C("📋<br>Markdown/TXT<br>Files")
end
subgraph API["🚀 ConvoCanvas API (FastAPI)"]
D("🎛️<br>Feature Flags<br>System")
E("📡<br>/api/conversations/<br>upload")
F("🧠<br>/api/v2/conversations/<br>analyze-enhanced")
G("⚡<br>/api/v3/conversations/<br>gpu-accelerated")
end
subgraph Processing["🔄 Analysis Engine"]
H("📊<br>Enhanced Content<br>Analyzer")
I("💭<br>Decision<br>Extraction")
J("😊<br>Sentiment<br>Analysis")
K("🏷️<br>Technical Domain<br>Classification")
L("🧮<br>NLP Pipeline<br>spaCy + TextBlob")
end
subgraph Visualization["📈 Visualization Layer"]
M("🗺️<br>Interactive<br>Mindmaps")
N("📊<br>Plotly<br>Visualizations")
O("🎨<br>Decision Flow<br>Networks")
end
subgraph Output["📤 Output Layer"]
P("📋<br>Decision<br>Analysis")
Q("💡<br>Content<br>Ideas")
R("📊<br>JSON API<br>Response")
S("🎯<br>Confidence<br>Scores")
end
subgraph Features["🎛️ Feature Management"]
T("🔧<br>Environment<br>Variables")
U("⚙️<br>Optional GPU<br>Acceleration")
V("🧪<br>Experimental<br>Features")
end
%% Flow connections
A --> B --> C
C --> E
C --> F
C --> G
D --> E
D --> F
D --> G
E --> H
F --> H
G --> H
H --> I
H --> J
H --> K
H --> L
I --> M
J --> M
K --> M
L --> N
M --> O
N --> O
O --> P
O --> Q
O --> R
O --> S
T --> D
U --> D
V --> D
%% Styling
style A fill:#2563eb,stroke:#1d4ed8,stroke-width:2px,color:#ffffff
style H fill:#059669,stroke:#047857,stroke-width:2px,color:#ffffff
style M fill:#dc2626,stroke:#b91c1c,stroke-width:2px,color:#ffffff
style D fill:#7c3aed,stroke:#6d28d9,stroke-width:2px,color:#ffffff
style P fill:#ea580c,stroke:#c2410c,stroke-width:2px,color:#ffffff
ConvoCanvas follows a modern microservices-inspired architecture with clear separation of concerns:
- AI Conversations: Raw chat exports from Claude, ChatGPT, etc.
- SaveMyPhind Integration: Seamless browser extension workflow
- File Support: Markdown (.md) and text (.txt) formats
- Feature Flags System: Environment-controlled feature toggles
- Tiered Endpoints: Basic → Enhanced → GPU-accelerated processing
- RESTful Design: Standard HTTP methods with JSON responses
- Enhanced Content Analyzer: Core NLP processing pipeline
- Decision Extraction: AI-powered decision point identification
- Sentiment Analysis: TextBlob-based emotional context detection
- Domain Classification: spaCy-powered technical categorization
- Interactive Mindmaps: Plotly-based decision flow networks
- Dynamic Layouts: Force-directed graph positioning
- Responsive Design: Scales with conversation complexity
- Structured Analysis: JSON responses with confidence scores
- Content Ideas: Actionable insights for LinkedIn/blog posts
- Decision Tracking: Historical decision patterns and outcomes
# Clone and setup
git clone https://github.com/rduffyuk/convocanvas.git
cd convocanvas
# Backend (Python/FastAPI)
cd backend
python -m venv venv
source venv/bin/activate # or `venv\Scripts\activate` on Windows
pip install -r requirements.txt
uvicorn app.main:app --reload --host 0.0.0.0 --port 8000
# Test the API
curl -X GET http://localhost:8000/docker-compose up --build# Test the API endpoints
curl -X GET http://localhost:8000/
# Check feature flags
curl -X GET http://localhost:8000/feature-flags
# Upload a conversation file (basic parsing)
curl -X POST "http://localhost:8000/api/conversations/upload" \
-H "accept: application/json" \
-H "Content-Type: multipart/form-data" \
-F "[email protected]"
# Enhanced analysis with decision tracking and mindmaps
curl -X POST "http://localhost:8000/api/v2/conversations/analyze-enhanced" \
-H "accept: application/json" \
-H "Content-Type: multipart/form-data" \
-F "[email protected]"- ✅ Enhanced Analysis API: Decision extraction with confidence scoring
- ✅ Interactive Mindmaps: Plotly-based decision flow visualizations
- ✅ Sentiment Analysis: Emotional context detection around decisions
- ✅ Technical Domain Classification: Auto-categorize conversations by domain
- ✅ Feature Flags System: Environment-based feature toggles
- ✅ API Foundation: Multiple endpoint tiers (basic, enhanced, GPU)
- ✅ SaveMyPhind Integration: Parse exported conversation formats
- ✅ LibreChat Integration: Full local AI deployment with web search
- ✅ Universal File Organization: Automated Obsidian vault management
- ✅ Smart Tagging System: YAML frontmatter automation across 56+ files
- ✅ Local-First Architecture: No external API dependencies for core AI
- ✅ Perplexity-Style Search: Real-time web search with citations
- ✅ Auto-Documentation: Self-generating session logs and system updates
- ✅ Error Handling: Graceful degradation for missing dependencies
- ✅ Testing Framework: Comprehensive API endpoint testing completed
- ✅ Production Config: Docker + nginx setup with deployment guides
- 🚧 Web Interface: API-based (UI planned for future release)
- 📋 Advanced Features: Context summarization, knowledge graphs (future)
- 📋 Monitoring: Production logging and monitoring (future)
- Backend: FastAPI + Python 3.12+
- NLP Processing: spaCy + TextBlob + scikit-learn
- Visualization: Plotly + NetworkX for interactive mindmaps
- Feature Management: Environment-based feature flags
- Production: Docker + nginx + comprehensive testing
- Integration: SaveMyPhind browser extension support
- Conversation file parser
- Basic content extraction
- API endpoints
- SaveMyPhind format support
- Context summarization for session continuity
- Technical decision tracking
- Multi-conversation thread analysis
- Web interface for easier testing
- Browser extension integration
- Automated content generation
- Export to content platforms
- Knowledge graph visualization
For Technical Professionals:
- Maintain context across multi-day debugging sessions
- Extract learning insights from AI-assisted problem solving
- Generate technical content from real troubleshooting experiences
For Content Creators:
- Transform technical conversations into blog post ideas
- Generate LinkedIn posts from learning moments
- Track technical learning journey over time
For Career Transitioners:
- Document skill development through AI conversations
- Create portfolio content from learning processes
- Bridge knowledge gaps between domains
Note: This project is in early development. While contributions are welcome, expect frequent changes to the codebase as I build toward the first stable release.
This started as a personal tool for my Windows→Linux→DevOps journey, but it's built to help anyone facing the AI context window problem.
Built with conversations that ConvoCanvas now analyzes - the entire project was planned in the very AI chats that it processes.
- Phase 1 (Current): Core parsing and basic content extraction
- Phase 2: Robust error handling, testing, and web interface
- Phase 3: Advanced AI features and production readiness
- v1.0.0: First stable release
- Python 3.12+
- FastAPI
- SaveMyPhind browser extension for conversation exports
- SaveMyPhind Extension by Hugo Collin - Essential for exporting conversations
- Obsidian - Knowledge management app for manual workflow integration
MIT License - See LICENSE for details.
From network engineering to automation, one conversation at a time. 🌐→🤖