First Steps

Choose one route:

Route A — Use this website (Baseline Generator)

If you want real-time metrics without writing code, use the Baseline Generator.

• Open: /baseline-generator
• What you get: live SGI + Velocity plots, context phase, per-turn telemetry, exportable JSON.
• Best for: collecting human sessions, validating the coherence region, quick demos.

Route B — Develop using the SDK (local/dev)

If you want to integrate semantic telemetry into your own app, use the SDK service endpoints.

Install

pip install -r MirrorMind/aicoevolution_sdk/requirements.txt

Minimal example (HTTP)

import time
import requests

SDK_URL = "http://localhost:8001"

def ingest(conversation_id: str, role: str, text: str):
    r = requests.post(f"{SDK_URL}/v0/ingest", json={
        "conversation_id": conversation_id,
        "role": role,
        "text": text,
        "timestamp_ms": int(time.time() * 1000),
    })
    r.raise_for_status()
    return r.json()

conv = "demo_conv_001"

ingest(conv, "user", "What is semantic orbital mechanics?")
data = ingest(conv, "assistant", "It models a conversation as an orbit around a context centroid.")

print("SGI:", data["ensemble"]["sgi_lite_mean"])
print("Velocity:", data["ensemble"]["angular_velocity_mean"])
print("Context phase:", data["by_backend"]["nomic"]["context_state_latest"])

Expected output (example)

SGI: 0.95
Velocity: 34.2
Context phase: stable

Standalone script (open source)

The reference standalone script is:

• MirrorMind/open_source/semantic_telemetry.py

It will be included in the public Paper 03 research repository.