Langflow AI Prototyping Environment

Overview

This project provides a Langflow-based environment for rapid AI prototyping. It leverages a set of containerized services to facilitate experimentation with Langflow, Weaviate for vector storage, and various monitoring tools. The setup does not include an inference service. It includes Weaviate, Postgres and Redis.

NOTE: Langflow now has a desktop client, simplifying setup. The default docker-compose.yml now excludes self-hosted Langflow and Postgres. Use docker-compose-langflow.yml for a fully self-hosted browser-based setup.

Components

Core Services

• Langflow – Main UI for building AI workflows

• PostgreSQL – Database backend for Langflow

• Weaviate – Vector database for AI memory storage (available at endpoint: /v1/graphql)

• Weaviate Embeddings – Handles local text embedding with BAAI/bge-base-en-v1.5

• Redis – Caching layer first used instead of chat history shoveling but could be useful elsewhere

• RedisInsight – Web-based Redis management UI (for database inspection)

Embeddings Setup

• The system uses Weaviate’s module with BAAI/bge-base-en-v1.5 (ONNX, CPU-optimized). CPU was chosen to leave VRAM for model inference.
• The embeddings service runs in a dedicated container (langflow-weaviate-embeddings).

Networking & Ports

The following services are available at:

Service	External URL	Internal Container & Port
Langflow UI	http://10.0.0.11:7878	langflow:7860
Weaviate	http://10.0.0.11:8181	langflow-weaviate:8080
Weaviate gRPC	n/a	langflow-weaviate:50051
Embeddings	n/a	langflow-weaviate-embeddings:8080
PostgreSQL	n/a	langflow-postgres:5432
Redis	n/a	langflow-redis:6379
RedisInsight	http://10.0.0.11:6380	langflow-redisinsight:6380

Project Configuration & Utilities

Configuration Files

• Sample Langflow flows and Weaviate vector DB classes can be found in the config/ directory.

Schema Setup

This project includes setup_weaviate.py for a repeatable way to load a new schema into the vector database and populate it with sample data:

🛠 Commands: --add-class Add schema from JSON (e.g. Memory_v1) --add-sample Add sample data from corresponding _samples.py --clear-sample Delete objects tagged as 'sample' --clear-all Delete all objects in collection

Command	Description
--add-class <Collection>	Add schema from JSON (e.g. Memory_v1)
--add-sample <Collection>	Add sample data from corresponding samples.py
--clear-sample <Collection>	Delete objects tagged as 'sample'
--clear-all <Collection>	Delete all objects in collection

Usage

1. Start the environment

   docker-compose up -d

2. Initialize Weaviate schema

   Add a sample schema (loads config/weaviate/Memory_v1.json)

   python --add-class Memory_v1

   Add sample data (optional)

   (loads config/weaviate/Memory_v1_samples.py)

   python --add-sample Memory_v1

3. Verify Weaviate schema and data

   Verify the schema directly http://10.0.0.11:8181/v1/schema or use a GraphQL client Altair GraphQL Client (think: Postman for Graph databases). You may save collections of queries in agc format.

   Note: A collection of queries is located in config/weaviate/Memory_v1.agc

4. Setup Langflow Desktop Client

   Install the Langflow macOS desktop client or direct your browser to yoru locally hosted Langflow

   Load sample workflows located in ./config/flows

Notes

• As stated setup does not include an LLM inference service.
• Uses Docker volumes instead of host file system mapping for ease erasure
• All containers are running as root. Do not expose this project to the open web