THE AASHRAI PLAYGROUND

A conversation. A little clarity.

Talk naturally. Discover what happens behind every answer.

01 / EXPLORE

Your insurance companion

Simple answers, in your language

Ready

LET’S MAKE IT SIMPLE

Big questions.
Clearer answers.

Start with what matters to you.
We’ll take it one conversation at a time.

Voice & language English · తెలుగు · தமிழ்

Language Hands-free: pause to send Read answers aloud

Click Microphone, speak, then Stop mic. Audio is sent to OpenAI to create editable text.

AI-generated voice. Reply text is sent to OpenAI when you listen.

OBSERVABILITY, MADE HUMAN

Under the hood

A little less magic. A lot more understanding.

Your first request starts the story.

CHAT TOKENS—

LLM CALLS—

RESPONSE TIME—

Input — Output —

Execution timeline

No request yet

Send a message to reveal the actual steps. A request may collect details, search documents, call a model, or use a fallback.

This panel shows reported execution events, not the model’s private reasoning.

Inspect reported JSON

No trace received.

Reported chat tokens since page load: 0

Chat totals exclude voice APIs and Codex coding tokens. Missing usage is shown as unknown, never as zero.

CONCEPTUAL MAP · NOT A LIVE TRACE

You→App→Backend

Context comes first

The backend uses the conversation and saved details to decide what to do next. Not every message needs a model call.

MCP connects the tools

For document questions, a policy-search tool can retrieve relevant passages. The trace shows whether a tool was actually called.

Skills guide the workflow

A skill is a set of instructions, not another model. Its availability does not prove it ran. Check the backend’s reported skill information.

The LLM builds an answer

A model receives context and produces text. Input and output tokens measure that work; the model name appears only when reported.

Try a small experiment

Ask “hi”, then ask about a policy’s benefits. Compare the reported LLM calls, search activity, and tokens.

How does voice travel?

Microphone → transcription API → chat request. Listen → speech API → audio playback. These are separate API operations; their usage is not included in chat totals.

A conversation. A little clarity.

Big questions.Clearer answers.

Big questions.
Clearer answers.