Connecting to Your Resources

Every resource in Featrix — foundation models, predictors, projects, jobs, endpoints — has a unique identifier. The Featrix UI shows you the exact code to access any resource you're looking at, so you can copy-paste it directly into your notebook, script, or agent prompt and know you're working with the right object.

The Problem This Solves

When you're staring at a model in the UI and you want to use it in code, you need to answer: which model is this, exactly? Names can be ambiguous — you might have "churn-model" in three different projects. UUIDs are unambiguous, but they're hard to type from memory.

Featrix solves this by putting copy-pasteable code snippets directly in the UI, next to every resource. You see a foundation model? The UI shows you:

fm = featrix.foundational_model("550e8400-e29b-41d4-a716-446655440000")

Copy it. Paste it into your notebook. You're now guaranteed to be talking to the exact model you were looking at. No name lookups, no guessing, no "wait, which one was it?"

---

What the UI Shows You

On Every Resource Page

Each resource page in the Featrix UI includes a Connect panel with:

• The UUID — click to copy to clipboard
• Python code — ready-to-paste featrixsphere SDK calls
• TypeScript code — equivalent featrixsphere-ts calls
• curl command — raw REST API call
• Deep link — shareable URL that takes anyone directly to this resource

Foundation Model Page

# Python — copy from the UI
from featrixsphere import FeatrixSphere
featrix = FeatrixSphere()
fm = featrix.foundational_model("550e8400-e29b-41d4-a716-446655440000")

// TypeScript — copy from the UI
import { FeatrixSphere } from 'featrixsphere-ts';
const featrix = await FeatrixSphere.create();
const fm = await featrix.foundationalModel("550e8400-e29b-41d4-a716-446655440000");

# curl — copy from the UI
curl -H "Authorization: Bearer $FEATRIX_API_KEY" \
  https://sphere-api.featrix.com/compute/session/550e8400-e29b-41d4-a716-446655440000

Predictor Page

predictor = featrix.predictor(session_id="550e8400-...", predictor_id="7f3a2b1c-...")

# Make a prediction
result = predictor.predict({"age": 35, "income": 75000})

Published Endpoint Page

published = featrix.published_predictor(
    org="acme",
    name="churn-v3",
    api_key="your-production-key"
)
result = published.predict({"customer_id": "C-1234"})

The UI fills in the real IDs, org names, and endpoint names for you. You just copy and run.

---

Deep Links: Share What You're Looking At

Every resource has a shareable URL:

https://app.featrix.com/goto/<type>/<identifier>

Type	Example
Organization	https://app.featrix.com/goto/org/acme
Project	https://app.featrix.com/goto/project/sales-forecasting
Foundation Model	https://app.featrix.com/goto/foundation/550e8400-...
Model (Predictor)	https://app.featrix.com/goto/model/churn-predictor
Job	https://app.featrix.com/goto/job/7f3a2b1c-...
Prediction Endpoint	https://app.featrix.com/goto/prediction/churn-endpoint

You can use names or UUIDs as the identifier — Featrix resolves either. If the resource belongs to a different organization than your current one, your active org is switched automatically.

Use case: Paste a deep link into Slack, a Jira ticket, a notebook comment, or an agent prompt. Anyone who clicks it lands on exactly the resource you meant.

---

UUIDs: The Unambiguous Identifier

Every Featrix resource has a UUID (e.g., 550e8400-e29b-41d4-a716-446655440000). UUIDs are:

• Immutable — they never change, even if you rename the resource
• Globally unique — no two resources share a UUID, across all orgs
• Copy-pasteable — click the UUID in the UI to copy it to your clipboard

Always use UUIDs in automation

Names are for humans. UUIDs are for code. When writing scripts, CI/CD pipelines, or agent prompts, always use the UUID. Names can change; UUIDs can't.

Searching by UUID

You can paste a UUID into the Featrix UI search bar to jump directly to any resource. This is especially useful when you're debugging — you see a UUID in a log file, paste it into search, and immediately see what it is.

---

Programmatic Access: Full Reference

Python SDK (featrixsphere)

pip install featrixsphere

from featrixsphere import FeatrixSphere

# Reads API key from ~/.featrix or FEATRIX_API_KEY env var
featrix = FeatrixSphere()

What you want	Code
Verify identity	featrix.whoami()
Get a foundation model	featrix.foundational_model("uuid")
List all foundation models	featrix.list_sessions()
Filter by name	featrix.list_sessions(name_prefix="sales")
Get a predictor	featrix.predictor(session_id="uuid")
List predictors on a FM	fm.list_predictors()
Set active project	featrix.set_current_project("project-name")
Get model card	fm.get_model_card()
Make a prediction	predictor.predict({"col": "value"})
Batch predict	predictor.batch_predict([{"col": "val"}, ...])
Explain a prediction	predictor.explain({"col": "value"})
Submit ground truth	featrix.prediction_feedback(prediction_uuid="uuid", ground_truth="label")
Access published model	featrix.published_predictor(org="acme", name="model", api_key="key")
Create API endpoint	predictor.create_api_endpoint(name="endpoint-name")
Wait for training	fm.wait_for_training() or predictor.wait_for_training()

TypeScript SDK (featrixsphere-ts)

import { FeatrixSphere } from 'featrixsphere-ts';
const featrix = await FeatrixSphere.create({ apiKey: 'your-key' });

What you want	Code
Verify identity	await featrix.whoami()
Get a foundation model	await featrix.foundationalModel("uuid")
List all foundation models	await featrix.listSessions()
Get a predictor	await featrix.predictor("session-uuid")
Set active project	await featrix.setCurrentProject("project-name")
Make a prediction	await predictor.predict({col: "value"})
Batch predict	await predictor.batchPredict([{col: "val"}, ...])
Explain a prediction	await predictor.explain({col: "value"})

---

End-to-End: UI to Code to Production

1. You train a foundation model in the UI. When it's done, you see the model page with its UUID, metrics, and code snippets.

2. You copy the code snippet into your notebook:

fm = featrix.foundational_model("550e8400-e29b-41d4-a716-446655440000")
predictor = fm.create_binary_classifier(
    name="churn-v3",
    target_column="churned",
)
predictor = predictor.wait_for_training()
print(f"AUC: {predictor.auc}, F1: {predictor.f1}")

3. You share the predictor link with your team for review:

https://app.featrix.com/goto/model/churn-v3

4. Your teammate opens the link, sees the predictor page, copies the prediction code, and tests it:

predictor = featrix.predictor(session_id="550e8400-...", predictor_id="7f3a2b1c-...")
result = predictor.predict({"tenure_months": 3, "monthly_spend": 29.99})

5. You deploy to production and give agents the endpoint:

endpoint = predictor.create_api_endpoint(name="churn-v3-prod")

# An agent uses the published predictor
published = featrix.published_predictor(org="acme", name="churn-v3-prod", api_key="agent-key")
result = published.predict(customer_record)

At every step, you copied code from the UI. At every step, the UUID guaranteed you were working with the exact right object.

---

Safety and Access Control

• Authentication required — all deep links require login; API access requires a valid API key
• Organization scoping — resources belong to orgs; API keys can only access their own org's resources
• Verify before automating — use featrix.whoami() to confirm you're in the right org before running batch operations
• UUIDs over names — names can change and collide across projects; UUIDs are permanent and unique
• Prediction feedback — use prediction_feedback() with the prediction UUID to close the monitoring loop and detect drift

# Always verify you're in the right place
identity = featrix.whoami()
print(f"Org: {identity['org_name']}, User: {identity['user_id']}")

---

Tips

• Click any UUID in the UI to copy it to your clipboard
• Paste a UUID into UI search to jump to any resource
• Names are case-insensitive in deep links — /goto/project/My%20Project and /goto/project/my%20project both work
• Store your API key in ~/.featrix or FEATRIX_API_KEY env var so you don't pass it every time
• Use %20 for spaces in deep link URLs (e.g., /goto/project/Sales%20Model)