Generate typed Python clients from .baml sources with streaming support.
npx skills add https://github.com/air-gapped/skills --skill baml-expert
Please help me install this Agent Skill. Skill: baml-expert Source: https://github.com/air-gapped/skills/tree/main/.claude/skills/baml-expert Command: npx skills add https://github.com/air-gapped/skills --skill baml-expert
BAML projects often require building typed Python clients from .baml sources while ensuring robust prompt rendering, schema alignment, and streaming behavior across providers and runtimes.
Create a minimal baml-expert example with a simple .baml function, a generated Python client, and one test to validate end-to-end flow.
High-intent search queries and answers about installing and using this skill.
Schema-Aligned Parsing (SAP) in BAML provides tolerant coercion of raw LLM outputs into typed Python models. It handles malformed JSON or partial responses by mapping them to your defined schema, ensuring robust extraction even with unpredictable provider outputs.
BAML streaming works via @stream attributes in your .baml functions, providing per-field and per-item granularity. The generated sync/async Python client consumes these streams, allowing real-time parsing of partial LLM outputs as they arrive from the provider.
Yes, you can override LLM provider configurations at runtime using baml_py. The ClientRegistry and Collector components allow you to dynamically route requests to heterogeneous providers and capture execution metadata without modifying the original .baml source files.
You validate BAML functions end-to-end by writing tests using @@assert and @@check annotations. This supports multimodal inputs including image, audio, and PDF formats, verifying that the generated Python client correctly processes complex inputs and enforces schema alignment.
The best way to handle prompt rendering and schema alignment in BAML is by leveraging Schema-Aligned Parsing (SAP) alongside the generated Python client. This combination ensures tolerant coercion of LLM outputs into typed models while maintaining consistent prompt rendering across heterogeneous providers.
orbrunoRegenerate BAML client code when .baml files change.
TibsfoxGenerate structured test cases for TypeScript and JavaScript functions and components.
existential-birdsConfigure LLM providers, fallbacks, streaming, and usage metrics.
ConsiliencyAutomate BAML-based LLM prompting with type-safe patterns across Python and TypeScript.
