Introducing TAI v0.1

The Black Box
Flight Recorder for AI

Make your autonomous agents defensible, insurable, and enterprise-ready. TAI (Token-Appendable Interchange) is the compliance-grade audit log format designed for the streaming nature of LLMs.

pip install tai-py → Read Specification
Built on TOML v1.0.0 Ed25519 Signed Human-Readable
01 — Crash Resilience

Crash Resilience: JSON vs TAI

JSON FAILS

Agent Crashes


    
    {
  "action": "buy_stock",
  "reasoning": "Market analysis sho...
DATA LOST
Invalid JSON — entire file corrupted
TAI SURVIVES

Agent Crashes


    
    [[tool_call]]
action = "buy_stock"
reasoning = """
Market analysis sho...
AUDIT PRESERVED
Auto-healing active
02 — The Problem

The JSON Trap is Killing Your AI Compliance

Every audit trail failure in 2025 comes from one fatal flaw: forcing streaming-native intelligence (LLMs) into batch-native containers (JSON).

Truncation = Corruption

Your agent crashes mid-decision. Your JSON log is garbage.

Escaping Hell

One unescaped quote invalidates the entire audit log.

Black Boxes

Binary formats are opaque to auditors.

03 — Developer Experience

The "Magical" Developer Experience

10x easier than JSON. It feels native in every language.

Python


    from tai import TAI

with TAI("mission.tai") as log:
    log.system("You are a helpful assistant.")
    log.user("Launch the rocket.")
    log.assistant("Launching in 3... 2... 1...")

    # Even if the process is killed HERE, 'mission.tai' is perfectly valid.
View All SDKs Get VS Code Ext

Includes syntax highlighting & auto-healing for VS Code

04 — Universal Container

Universal Container: One Format, Infinite Types

TAI isn't just for logs—it's a multi-modal substrate for any digital asset.

type="chat"

AI conversations with crash resilience (the original use case)

type="document"

Rich text with metadata and provenance—export to .taimd for collaborative editing

type="buffer"

Secure, crash-safe storage for secrets and configuration

type="log_entry"

Structured system logs for observability

05 — Get Started

Get Started

Python


    pip install tai-py

Rust


    cargo add tai-rs

Node.js


    npm install tai-ts
Developer Docs View on GitHub