WA-Construct AI — AI-Powered Construction Assistant
AI / Machine Learning
Project Overview
An AI-powered construction assistant built with FastAPI, LangChain, LangGraph, and Weaviate. The project combines LLM workflows, retrieval-based knowledge access, and backend services to support intelligent construction-related assistance through contextual information retrieval and AI-driven processing.
Tech Stack
The Problem
Construction-related workflows can involve large amounts of technical information and domain-specific knowledge. Accessing relevant information efficiently and building intelligent workflows around that information requires a backend architecture capable of supporting LLM integration, retrieval, asynchronous processing, and secure access control.
The Solution
Developed an AI-powered construction assistant using FastAPI and Python, integrating LangChain and LangGraph for LLM orchestration and agent workflows. Incorporated Weaviate for vector-based information retrieval, PostgreSQL for relational data management, and Redis with ARQ for asynchronous background processing. Casbin supports authorization and access-control requirements
Key Features
- AI-powered construction assistance
- LLM workflow orchestration with LangChain
- Agent workflow development with LangGraph
- Vector-based information retrieval with Weaviate
- REST API development with FastAPI
- PostgreSQL integration using SQLAlchemy
- Asynchronous task processing with Redis and ARQ
- Authorization and access control with Casbin
- Local LLM integration using Ollama
Challenges
- Designing a backend architecture that integrates multiple AI and infrastructure components
- Orchestrating LLM workflows and multi-step agent execution
- Connecting vector-based retrieval with application data and AI processing
- Managing asynchronous tasks and backend processing
- Implementing authorization within backend workflows
Learnings
- Building Python backend services with FastAPI
- Orchestrating LLM applications using LangChain and LangGraph
- Working with vector databases for semantic retrieval
- Integrating relational databases with SQLAlchemy
- Implementing asynchronous processing with Redis and ARQ
- Applying authorization policies with Casbin
- Integrating local LLMs using Ollama
Case Study
WA-Construct AI — AI-Powered Construction Assistant
Overview
WA-Construct AI is an AI-focused construction software project built around modern Python backend technologies and LLM orchestration. It combines AI workflows, information retrieval, database services, and asynchronous processing to support construction-related assistance.
Problem
Construction workflows often involve domain-specific information and complex requirements. Making that information accessible through intelligent assistance requires more than a simple chatbot: the application needs a structured backend, retrieval capabilities, orchestrated AI workflows, and appropriate access controls.
Solution
I worked on building the backend architecture and AI components using FastAPI, LangChain, and LangGraph. The architecture integrates vector-based retrieval through Weaviate, relational data management through PostgreSQL and SQLAlchemy, and asynchronous processing through Redis and ARQ.
Ollama provides integration with locally hosted language models, while Casbin supports authorization and access-control policies.
Technical Architecture
- Backend: FastAPI and Python for API development.
- AI orchestration: LangChain and LangGraph for LLM workflows and agent orchestration.
- Vector retrieval: Weaviate for vector-based information retrieval.
- Database: PostgreSQL and SQLAlchemy for relational persistence.
- Background processing: Redis and ARQ for asynchronous task execution.
- Authorization: Casbin for access-control policies.
- LLM runtime: Ollama for local model integration.
Key Engineering Areas
The project brings together AI orchestration and conventional backend engineering. Its architecture addresses the need to coordinate LLM workflows, retrieve relevant context, manage application data, process background tasks, and enforce authorization.
Key Takeaways
WA-Construct AI strengthened my experience in Python backend development, LLM integration, agent orchestration, vector retrieval, and AI application architecture. It also provided practical experience combining AI frameworks with databases, asynchronous infrastructure, and access-control systems.
Technologies
Python, FastAPI, LangChain, LangGraph, Weaviate, PostgreSQL, SQLAlchemy, Redis, ARQ, Ollama, and Casbin.