skills.rest
sethjuarez
sethjuarezCOMMUNITY
Explore all skills by this author →
Website𝕏@sethjuarezGitHub

castia-lifecycle

Build, evaluate, and optimize Foundry hosted agents using the castia Python SDK.

Updated Sep 8, 2026
Search Tags:#agent#evaluation#fine-tuning#scaffold#foundry#python sdk#castia
One-click install
npx skills add https://github.com/sethjuarez/castia --skill castia-lifecycle-sethjuarez
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: castia-lifecycle
Source: https://github.com/sethjuarez/castia/tree/main/plugins/foundry-agent-starter-kit/skills/castia-lifecycle
Command: npx skills add https://github.com/sethjuarez/castia --skill castia-lifecycle-sethjuarez

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Building a Microsoft Foundry hosted agent involves many moving parts: correct SDK imports, uv extras, credentials, offline checks, and knowing which operations are free versus billable. This Skill guides you through the full castia Python SDK lifecycle so you scaffold the right project shape, avoid stale import paths, and never trigger billable evaluation, optimization, or fine-tuning jobs without explicit approval. ## Core Features & Use Cases - New agent golden path: Scaffold a minimal Castia app with main.py, pyproject.toml, azure.yaml, and .agent_configs baseline files instead of starting from a non-Castia sample. - Lifecycle command guidance: Covers build, observe, eval, optimize, snapshot, and finetune CLI commands with their offline dry-run modes and live approval boundaries. - Runtime and credential setup: Explains FOUNDRY_PROJECT_ENDPOINT, DefaultAzureCredential, native MCP tool integration, and candidate resolution via load_agent_config. - Use Case: A developer asks to create a new Foundry Responses agent with MCP tools; the Skill produces the canonical Castia file layout, correct imports, and a safe offline test before any live deployment. ## Quick Start Ask the assistant to scaffold a new Castia Foundry agent with a Responses endpoint and offline tests using the castia Python SDK.

Frequently Asked Questions about castia-lifecycle

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I build a Foundry hosted agent with the castia Python SDK?▼

Create a minimal app with main.py, pyproject.toml, azure.yaml, and a .agent_configs baseline, or run python -m castia build scaffold. Register an Agent, use the @app.responses() decorator, and guard app.run() with the __main__ check.

How do I install castia with uv and which extras do I need?▼

Run uv pip install --prerelease=allow "castia[deploy,optimize,test]" in a Python 3.11+ environment. The extras supply YAML tooling, the candidate resolver, and tests; Azure CLI and azd are installed separately.

Which castia operations are free versus billable?▼

Offline commands like build check, eval --dry-run, and optimize --check are free. Live operations such as eval generate/run, optimize run, and fine-tuning submission incur billable model or training costs and each requires explicit authorization.

Does castia load .env files and Azure credentials automatically?▼

No, castia does not load .env automatically; starter apps load it with python-dotenv before constructing Model. Model reads FOUNDRY_PROJECT_ENDPOINT and uses DefaultAzureCredential via Azure CLI login or managed identity.

Why did my old castia flat submodule imports stop working?▼

The former 31 flat submodules were removed without aliases. Use the short public API like from castia import Agent, Depends, Model, or capability packages such as castia.inference.model and castia.evaluation.suite.

Can castia run RFT fine-tuning directly from eval.yaml?▼

No, RFT accepts explicit JSONL splits and a grader, not eval.yaml as a universal config. RFT wire acceptance is provisional, and training plus checkpoint deployment each need separate authorization.

Related Skills

Explore More Software Engineering Skills→
organvm-i-theoriaorganvm-i-theoria

microsoft-foundry

Deploy, evaluate, and manage Microsoft Foundry agents and model deployments end-to-end.

Official
Advanced
sohamdasohamda

microsoft-foundry

Deploy, evaluate, and manage Azure AI Foundry agents and model deployments end-to-end.

Community
Advanced
obsessiondbobsessiondb
12

clickhouse-schema-design

Design ClickHouse table schemas with ORDER BY, PARTITION BY, and TTL rules.

Official
Advanced
JustinWangJPJustinWangJP
6

ms-agent-framework-python

Create AI agents and multi-agent workflows with Microsoft Agent Framework for Python.

Community
Advanced