Skip to content

Quick Start

Track your first AI interaction in under 5 minutes.

Prerequisites

Step 1: Install

pip install atheon-codex

Step 2: Initialise & Track (Backend)

Call atheon.init() once at startup, then atheon.track() after every LLM call. Both are non-blocking — your API response time is unaffected.

import os
import atheon

# Call once at application startup
atheon.init(os.environ["ATHEON_API_KEY"])

# Inside your chat handler
user_query  = "How do I center a div?"
llm_output  = "You can use Flexbox..."

interaction_id = atheon.track(
    provider="openai",
    model_name="gpt-4o",
    input=user_query,
    output=llm_output,
    tokens_input=14,
    tokens_output=32,
    finish_reason="stop",
)

# Return the interaction_id to your frontend
return {
    "reply": llm_output,
    "interaction_id": str(interaction_id),
}

Step 3: Render (Frontend)

Load the Atheon script and wrap your output in <atheon-container>, passing the interaction_id returned by your backend.

<script
  data-atheon-publisher-key="YOUR_PUBLISHER_KEY"
  src="https://js.atheon.ad/atheon.js"
  defer
></script>

<atheon-container id="chat-bubble">
  <div id="text-content"></div>
</atheon-container>

<script>
  const { reply, interaction_id } = backendResponse;

  document.getElementById('text-content').innerText = reply;
  document.getElementById('chat-bubble').setAttribute('interaction-id', interaction_id);
</script>

Step 4: Shut Down Gracefully

Call atheon.shutdown() when your process exits to flush any remaining queued events.

atheon.shutdown()

That's it — your interaction is now tracked, fingerprinted, and visible in the Atheon Dashboard.

Next Steps