Build Phases

Progress

0/5

◆ Blueprint — read this first

What you'll build, what you need, what it costs

Setup time

~25 min

Build time

~90 min

Monthly cost

$0 dev

You're shipping a production-shape RAG service: source docs → indexed vector store → retriever → reranker → grounded LLM answer, exposed behind a FastAPI endpoint. Everything below is required before Phase 1.

1. Prerequisites

Python 3.10+Required
Core runtime for the LangChain + Qdrant stack used in this build.
Get it ↗
Docker DesktopRequired
Spins up Qdrant locally in seconds — no cloud signup needed for dev.
Get it ↗
OpenAI API keyRequired
Used for embeddings (text-embedding-3-small) + generation (gpt-4o-mini).
Get it ↗
Embeddings & vectors — read firstOptional
Skim the Read mode if "embedding space" or "cosine similarity" feel fuzzy.
A REST client (curl / Insomnia)Optional
For poking the FastAPI service once Phase 5 is up.
~2 GB free diskOptional
Qdrant volume + embedding cache. Negligible for ~500 docs.

2. System architecture — the finished system

Hover a phase to highlight its components

3. Cost reality check

Service	Local / free	Production
OpenAI embeddings	—	$0.02 / 1M tokens
OpenAI generation (gpt-4o-mini)	—	$0.15 / $0.60 per 1M in/out
Qdrant	Local Docker · free forever	$25/mo cloud cluster
Hosting (FastAPI)	Localhost	$5–20/mo on Fly/Railway

Indexing 1,000 typical docs (~500k tokens) costs about $0.01 in embeddings. A live demo answering 100 queries/day with gpt-4o-mini lands under $1/month.

🛠️

Phase 1 of 5

Environment Setup

Install dependencies and configure your environment for building a production RAG system.

Tasks

Install qdrant-client, langchain, openai, python-dotenv

Start Qdrant vector database (local or cloud)

Set OPENAI_API_KEY in .env

Verify embedding model access

🎯 What You'll Learn

▸Why each dependency is needed in the RAG stack
▸How to run a local vector database
▸Environment variable management for API keys

💡 Pro Tips

⚡Use a virtual environment (python -m venv .venv) to isolate deps
⚡Qdrant Cloud offers a free tier — great for testing without Docker
⚡Never commit .env files — add to .gitignore immediately

Code

Textrequirements.txt
1 qdrant-client>=1.8.0
2 langchain>=0.2.0
3 langchain-openai>=0.1.0
4 langchain-community>=0.2.0
5 openai>=1.25.0
6 python-dotenv>=1.0.0
7 pypdf>=4.0.0

YAMLdocker-compose.yml
1 services:
2   qdrant:
3     image: qdrant/qdrant:latest
4     ports:
5       - "6333:6333"
6       - "6334:6334"
7     volumes:
8       - qdrant_storage:/qdrant/storage
9  
10 volumes:
11   qdrant_storage:

Bash.env
1 OPENAI_API_KEY=sk-...
2 QDRANT_URL=http://localhost:6333
3 COLLECTION_NAME=rag_knowledge_base

✅ Expected Output

$ docker compose up -d
✅ qdrant started on port 6333
$ pip install -r requirements.txt
Successfully installed 12 packages

Complete all tasks above, then continue

Service

Local / free

Production

OpenAI embeddings

—

$0.02 / 1M tokens

OpenAI generation (gpt-4o-mini)

—

$0.15 / $0.60 per 1M in/out

Qdrant

Local Docker · free forever

$25/mo cloud cluster

Hosting (FastAPI)

Localhost

$5–20/mo on Fly/Railway

1 services: 2 qdrant: 3 image: qdrant/qdrant:latest 4 ports: 5 - "6333:6333" 6 - "6334:6334" 7 volumes: 8 - qdrant_storage:/qdrant/storage 9 10 volumes: 11 qdrant_storage:

Build a RAG System — Step by Step

What you'll build, what you need, what it costs

1. Prerequisites

2. System architecture — the finished system

3. Cost reality check

Environment Setup

Tasks

🎯 What You'll Learn

💡 Pro Tips

Code

Discussion

Build a RAG System — Step by Step

What you'll build, what you need, what it costs

1. Prerequisites

2. System architecture — the finished system

3. Cost reality check

Environment Setup

Tasks

🎯 What You'll Learn

💡 Pro Tips

Code

Discussion

1	qdrant-client>=1.8.0
2	langchain>=0.2.0
3	langchain-openai>=0.1.0
4	langchain-community>=0.2.0
5	openai>=1.25.0
6	python-dotenv>=1.0.0
7	pypdf>=4.0.0

1	services:
2	qdrant:
3	image: qdrant/qdrant:latest
4	ports:
5	- "6333:6333"
6	- "6334:6334"
7	volumes:
8	- qdrant_storage:/qdrant/storage
9
10	volumes:
11	qdrant_storage:

1	OPENAI_API_KEY=sk-...
2	QDRANT_URL=http://localhost:6333
3	COLLECTION_NAME=rag_knowledge_base