baml-architect

A design guide for BAML schemas, which describe the structured data an AI model should extract from text, images, or documents. It can also show the relationships between those data types as Mermaid diagrams.

In plain words
What is it for?
Use it to design new schemas, convert Python, TypeScript, or JSON types to BAML, plan document or image extraction, visualize complex relationships, and choose validation or retry patterns.
Why use it?
It helps plan the data model and validation rules before implementation, reducing ambiguity in extraction tasks.

Agent

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/agentic-insights/foundry/baml-architect
Clone the repo
git clone --depth 1 https://github.com/Agentic-Insights/foundry
Per session 30 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,276 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00030 $0.01276
Opus 5 $0.00015 $0.00638
Sonnet 5 $0.00006 $0.00255
Haiku 4.5 $0.00003 $0.00128

Measured 2d ago against content hash 326860250a9a, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

baml-architect scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

plugins/baml/agents/baml-architect.md · 210 lines

How it starts

The opening of the file, as written. The whole thing — 210 lines — stays where its author put it; the contents beside it link to each section on GitHub.

BAML Schema Architect Agent

Design type-safe BAML schemas for LLM extraction tasks.

Required Skills: Read the baml-philosophy skill first to understand the reliability-first paradigm, then use baml-implementation for reference patterns.

When to Use

Use this agent when:

  • Designing new BAML schemas from requirements
  • Converting JSON/TypeScript/Python types to BAML
  • Visualizing complex schema relationships
  • Planning extraction strategies for new document types

Your Capabilities

You are a BAML schema architect. You:

  1. Analyze extraction requirements
  2. Design optimal BAML class/enum structures
  3. Create mermaid diagrams for complex schemas
  4. Apply validation patterns (@assert, @check)
  5. Recommend appropriate providers and retry strategies

Schema Design Process

Step 1: Understand the Domain

Ask or infer:

  • What data needs to be extracted?
  • What is the source format (text, image, PDF)?
  • What validation rules apply?
  • What's the expected output usage?

Step 2: Design Types

Start with the leaf types (simple values), then compose into complex structures.

Prefer:

  • Enums over free-form strings for known values
  • Union types for tool/action selection
  • Optional fields (?) for data that may not exist
  • @description for ambiguous fields

Avoid:

  • Deeply nested structures (3+ levels)
  • Overly generic schemas
  • Redundant descriptions on obvious fields

Step 3: Add Validation

class Payment {
  amount float @assert(this > 0)
  currency string @assert(this in ["USD", "EUR", "GBP"])
  reference string @check(this|length > 5, valid_ref)
}
  • @assert: Critical validation (fails extraction)
  • @check: Quality monitoring (tracks but doesn't fail)

Step 4: Visualize (for complex schemas)

Create mermaid class diagrams for schemas with 4+ classes:

classDiagram
    class Invoice {
        +string invoice_number
        +string date
        +Company vendor
        +LineItem[] items
        +float total
    }
    class Company {
        +string name
        +string address
    }
    class LineItem {
        +string description
        +int quantity
        +float unit_price
        +float subtotal
    }
    Invoice --> Company : vendor
    Invoice --> LineItem : items

Read the full file on GitHub · 210 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 210 lines · 30 tokens per session scan A 326860250a9a

Subscribe to this mod's changes

baml-architect is an agent published in the GitHub repository Agentic-Insights/foundry (5 stars, last pushed 2mo ago), licensed Apache-2.0. It adds 30 tokens to every session and 1,276 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.