Valkey for AI

Cookbooks, reference architectures, and runnable demos for using Valkey in AI workloads - semantic caching, vector search, conversation memory, agent state, rate limiting, RAG pipelines, and more.

Live site: valkeyforai.com

What's here

9 use-case tracks with 45+ step-by-step cookbooks:

Track	Cookbooks	What it covers
Semantic Caching	3	Cache LLM responses by meaning, not exact match
Conversation Memory	5	Chat history, session management, semantic search
Vector Search	3	HNSW indexes, KNN queries, hybrid filters
RAG Pipelines	6	Document chunking, retrieval, caching, monitoring
Rate Limiting	6	Fixed window, token-aware, hierarchical, cost-based
Feature Store	6	Real-time feature serving for ML models
Pub/Sub & Streaming	6	LLM token streaming, consumer groups, fan-out
Context Engineering	3	Memory assembly, context budgeting, production patterns
Agent Session State	Coming soon	Tool call persistence, reasoning checkpoints

5 framework integrations:

Framework	Cookbooks	Integration
Mem0	3	Native provider: "valkey" connector
LangChain / LangGraph	4	langgraph-checkpoint-aws with ValkeySaver
CrewAI	3	Custom ValkeyStorage with GLIDE client
Strands	3	strands-valkey-session-manager package
Haystack	2	valkey-haystack with ValkeyDocumentStore

Each track has interactive demos on the live site.

Quick start

git clone https://github.com/meet-bhagdev/valkeyforai.git
cd valkeyforai

# Serve locally
python -m http.server 8000
# or
npx serve .

Open http://localhost:8000 in your browser.

Contributing

Editing cookbooks

All cookbook content lives in markdown files under content/. When you push to main, a GitHub Action runs node build.js to regenerate the HTML in cookbooks/, and the site updates automatically.

content/
├── semantic-caching/
│   ├── meta.json                    # Track config (titles, nav, difficulty)
│   ├── 01-getting-started.md        # ← edit this
│   ├── 02-multiturn-caching.md
│   └── 03-production.md
├── conversation-memory/
│   ├── meta.json
│   └── *.md
├── crewai/
│   ├── meta.json
│   └── *.md
└── ... (13 tracks total)

To edit a cookbook:

1. Edit the .md file in content/<track>/
2. Use fenced code blocks with language tags (```python, ```bash)
3. Run node build.js locally to preview (optional)
4. Push to main - the GitHub Action rebuilds and Amplify deploys

To add a new cookbook to an existing track:

1. Create a new .md file in the track's content/ directory
2. Add an entry to meta.json with title, difficulty, time, and prev/next links
3. Update the prev/next links on adjacent cookbooks in meta.json
4. Push to main

To add a new track:

1. Create content/<track-name>/meta.json (copy an existing one as template)
2. Add your .md files
3. Create cookbooks/<track-name>/index.html (copy from an existing track's index)
4. Add the track to the homepage index.html
5. Push to main

meta.json format

{
  "trackName": "Semantic Caching",
  "cookbooks": [
    {
      "num": "01",
      "source": "01-getting-started.md",
      "output": "01-getting-started.html",
      "title": "Getting Started with Semantic Caching",
      "h1": "Getting Started with Semantic Caching",
      "breadcrumb": "Getting Started",
      "difficulty": "Beginner",
      "time": "15 min",
      "next": {
        "file": "02-multiturn-caching.html",
        "title": "02 - Multi-Turn Caching"
      }
    }
  ]
}

Other files

File	What it is	How to edit
index.html	Homepage	Edit directly (not generated)
cookbooks/<track>/index.html	Track landing page	Edit directly (not generated)
demo/*.html	Interactive demos	Edit directly
use-cases/*/index.html	Use-case overview pages	Edit directly
styles.css	Main site styles	Edit directly
cookbooks/cookbook.css	Cookbook page styles	Edit directly
build.js	Markdown-to-HTML builder	Generates cookbooks/<track>/*.html from content/

Running the build locally

npm install        # first time only
node build.js      # builds all tracks
node build.js semantic-caching   # build one track

Project structure

valkeyforai/
├── index.html                # Homepage
├── styles.css                # Site styles
├── script.js                 # Homepage interactions
├── build.js                  # Markdown → HTML builder
├── content/                  # Markdown source (edit these)
│   ├── semantic-caching/
│   ├── conversation-memory/
│   ├── vector-search/
│   ├── rag-pipelines/
│   ├── rate-limiting/
│   ├── feature-store/
│   ├── pubsub-streaming/
│   ├── context-engineering/
│   ├── crewai/
│   ├── langchain/
│   ├── mem0/
│   ├── haystack/
│   └── strands/
├── cookbooks/                # Generated HTML (don't edit directly)
├── demo/                     # Interactive demos
└── use-cases/                # Use-case overview pages

Tech stack

• Pure HTML/CSS/JS (no framework, no build tools beyond node build.js)
• Inter font
• highlight.js for syntax highlighting
• Hosted on AWS Amplify

License

MIT

Links

• Valkey
• Valkey docs
• valkey-py
• Valkey GLIDE