kontraktus-ai
Local AI engine, Ollama, embeddings, RAG pipeline, ChromaDB.
Technical Roadmap
Kontraktus Implementation Planning
Planned MVP Architecture
kontraktus-web
Dashboard for uploads, Q&A, citations, and risk register.
kontraktus-docs
Optional Paperless-ngx layer for OCR and document organization.
database layer
Contracts, risks, users, audit logs, and future workflow records.
Proposed AI Workflow
1Upload
2Extract
3Chunk
4Embed
5Retrieve
6Analyze
7Cite
Planning Metrics
100%
Local-first design
0
Cloud contract uploads
3
Planned core containers
1
Approved capstone idea
16GB
RAM MVP target
i7 12th Gen
Test machine
Research Notes & Decisions
Research Focus
Local RAG systems and private AI for contract intelligence workflows.
Current Challenge
Balancing local AI performance with limited hardware resources.
Planned Stack
Ollama, ChromaDB, LangChain, Streamlit/Jekyll.
Supervisor Feedback
Focus on an achievable MVP before adding advanced workflows.
Key Decisions
Prioritize citation-grounded answers and privacy-preserving local processing.
Current Risks/Challenges
Scope creep, model latency on local hardware, and document quality variance.