🤖 programming language

ai.play

v0.5 ● active Python reference impl

A programming language for building AI systems. Write .aip files — model selection, web search, vision, voice, and memory in one place. No build steps, no config. Just run.

🔗 GitHub ⬇️ Download
Capabilities
🤖

Model Selection

Choose from factual, fun, thinking, or custom modes. Tune the personality to your use case in a single line.

🔍

Web Search

Built-in DuckDuckGo integration. Your AI can search the web in real-time without any extra wiring or API keys.

👁️

Vision Input

Single image or live webcam feed. Pass visual context to your model with no external API boilerplate.

🖼️

Image & Video Generation

Generate images and text-to-video output from within your .aip file. First-class output, not an afterthought.

🎤

Voice I/O

Speech input and output for phone-call style interactions. Deploy conversational AI without complex pipelines.

📄

Data Training

Feed native pairs, JSON, CSV, PDF, DOCX, or plain text. The pipeline handles tokenisation, TF-IDF embedding, and semantic retrieval.

Memory Architectures
🔮

Rule-Based

Pattern matching against defined rules. Predictable, fast, and deterministic — best for structured domains.

pattern matching
🧠

Generative

Concept extraction and generative memory. The system learns and adapts from interactions over time.

concept extraction
💾

Persistent

User memory that survives sessions. Sync across ten machines on a network mount — the memory follows the user.

cross-machine sync
Deployment
💻

Local Dev

Built-in test UI. Run your .aip file locally and iterate fast — no server setup needed.

🌐

HTTP Server

out.in() launches an HTTP server with user persistence and optional remote memory synchronisation.

🔗

Windows Installer

Graphical installer registers .aip file associations system-wide. One install, ready to go.

Quick Start
example.aip 
# choose your model mode
model  factual

# enable capabilities
search enabled
vision disabled

# set memory architecture
memory persistent

# define a system personality
system "You are a helpful coding assistant. You remember your users."

# train on custom data
train  "docs/knowledge.json"

# load a module from the registry
module "code-assist"

# launch the HTTP server
out.in()
View full docs on GitHub →