hoist-az 0.1.2

The Problem

Infrastructure-as-Code tools like ARM, Bicep, and Terraform are great at provisioning an Azure AI Search service — but they stop at the front door. The configuration inside the service — index schemas, skillsets, indexer schedules, knowledge base definitions — is what actually determines how your application behaves, and none of these tools manage it.

For traditional relational databases, this gap was filled long ago. SQL provides a standardized language for defining tables, indexes, and stored procedures, and a mature ecosystem of migration tools (Flyway, Liquibase, EF Migrations, Alembic) has evolved around it. Azure AI Search has no equivalent. It's a specialized search and vector database with a REST/JSON API, no schema definition language, and no migration framework.

In practice, this means search configurations are managed through the Azure portal or one-off scripts, which creates real problems:

• No change history — Azure doesn't track who changed an index schema or when. If a field type change breaks your application, there's no way to see what happened or roll back.
• Portal drift — The portal makes ad-hoc changes frictionless. In team environments, configurations silently diverge between services and between what's deployed and what anyone remembers deploying.
• No review process — Index schema changes, scoring profile updates, and skillset modifications go live without review, even though they can fundamentally change application behavior.
• Manual environment promotion — Copying configurations from dev to staging to production means manually exporting JSON, updating cross-resource references, and hoping nothing was missed.
• Growing stakes with agentic AI — As Azure AI Search becomes the retrieval backbone for AI agents in Azure AI Foundry, these configurations become critical infrastructure. A misconfigured knowledge base or skillset silently degrades agent quality — making version control and review more important, not less.
• Configuration locked behind APIs — There is no single place to see how your indexes, skillsets, indexers, and knowledge bases fit together. The full picture is spread across portal blades and REST endpoints, making it hard for anyone — human or AI — to reason about the service as a whole.

What Hoist Does

hoist bridges the gap by treating your search service configuration as code. It pulls resource definitions as normalized JSON files, versions them in Git, and pushes changes back:

• Version control — track who changed what, when, and why via Git history
• Code review — review index schema changes, skillset updates, and knowledge base configurations in pull requests
• Environment promotion — copy resources between services (dev → staging → prod) with automatic reference rewriting
• Drift detection — diff local files against the live service to catch manual portal changes
• AI-assisted development — with every resource definition available as a local file, AI coding tools like Claude Code, GitHub Copilot, Codex, and others can read your entire search configuration in context, understand how resources relate to each other, and help you develop, troubleshoot, and evolve your implementation — no API calls or portal access required

Quick Start

# Install
cargo install hoist-az

# Initialize a project (discovers your service via Azure CLI)
hoist init . --path search

# Pull all resources as JSON files
hoist pull --all

# Edit locally, then push changes back
hoist push --all

After init, your project looks like this:

hoist.toml                     # Project configuration
.hoist/                        # Sync state (gitignored)
search/
  search-management/
    indexes/
      hotels.json
    indexers/
      hotels-indexer.json
    data-sources/
      cosmos-hotels.json
    skillsets/
      enrichment-pipeline.json
    synonym-maps/
      hotel-synonyms.json
  agentic-retrieval/
    knowledge-bases/
      regulatory-kb.json
    knowledge-sources/
      regulatory-docs.json

Each JSON file is a normalized, deterministic representation of the resource — credentials stripped, properties in Azure's canonical order, arrays sorted by identity key.

Features

Pull & Push

Download resource definitions from Azure and upload local changes back:

# Pull everything
hoist pull --all

# Pull specific resource types
hoist pull --indexes --skillsets

# Pull a single resource by name
hoist pull --index hotels

# Push with dry-run preview
hoist push --all --dry-run

# Push a single resource
hoist push --indexer hotels-indexer

Semantic Diff

Compare local files against the live service with field-level change descriptions:

hoist diff --all

~ Index 'hotels' (modified)
    fields[3].type: Edm.String → Edm.Int32
    fields[7]: added 'rating'
    scoringProfiles[0].functions: 2 → 3 items

Cross-Service Copy

Copy resources between search services with automatic name remapping and reference rewriting:

# Copy with a suffix (auto-generates new names)
hoist push --all --target prod-search --suffix "-v2"

# Copy with interactive name prompts
hoist push --knowledgebase my-kb --recursive --copy --target staging

# Copy with a pre-built name mapping
hoist push --all --target prod-search --answers name-map.json

The --recursive flag automatically includes dependent and child resources. For example, --knowledgebase my-kb --recursive includes the knowledge base, all its knowledge sources, and their referenced indexes.

Watch Mode

Continuously poll for server-side changes:

hoist pull-watch --all --interval 30

Validation

Check local files for structural issues and referential integrity before pushing:

hoist validate

Resource Types

Preview resources (Knowledge Bases and Knowledge Sources) use the 2025-11-01-preview API and are included by default with the agentic template.

Authentication

hoist authenticates using the Azure CLI or service principal credentials:

# Option 1: Azure CLI (recommended for development)
az login
hoist pull --all

# Option 2: Service principal (for CI/CD)
export AZURE_CLIENT_ID=...
export AZURE_CLIENT_SECRET=...
export AZURE_TENANT_ID=...
hoist pull --all

Configuration

Project settings live in hoist.toml:

[service]
name = "my-search-service"
subscription_id = "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx"

[sync]
resource_dir = "search"
include_preview = true

[api]
api_version = "2024-07-01"
preview_api_version = "2025-11-01-preview"

View and modify settings with the config command:

hoist config get service.name
hoist config set sync.include_preview false

Architecture

Four crates with a clear dependency hierarchy:

hoist-az  →  hoist-core
     ↓              ↑
hoist-azent ────┘
hoist-diff  (standalone)

Building from Source

git clone https://github.com/mklab-se/hoist.git
cd hoist
cargo build --release

Requires Rust 1.82 or later.

Running Tests

cargo test              # All 226 tests
cargo test -p hoist-core   # Core crate only
cargo clippy            # Lint

License

MIT — see LICENSE.md.