langextract-claude

Anthropic Claude provider for LangExtract.

LangExtract ships with Gemini, Vertex AI, OpenAI, and Ollama providers. This package adds Claude support as a third-party provider plugin.

Installation

pip install langextract-claude

Usage

import langextract as lx

results = lx.extract(text, schema, model_id="claude-sonnet-4-6")

Any claude-* model ID is automatically routed to this provider. The default model is claude-sonnet-4-6.

Configuration

The API key is resolved in order:

1. Explicit api_key parameter
2. ANTHROPIC_API_KEY environment variable
3. LANGEXTRACT_API_KEY environment variable

from langextract_claude import AnthropicLanguageModel

model = AnthropicLanguageModel(
    model_id="claude-sonnet-4-6",
    api_key="sk-ant-...",       # or set ANTHROPIC_API_KEY
    temperature=0.0,
    max_tokens=8192,
    max_workers=5,              # parallel inference threads
)

Features

• Multi-block JSON normalization — Claude sometimes splits large extractions across multiple fenced code blocks. This provider automatically merges them into a single block that LangExtract can parse.
• Parallel inference — batch prompts are processed concurrently via a thread pool (configurable max_workers, default 5).
• Auto-discovery — any model ID matching ^claude is routed here. No manual provider registration needed.

License

Apache-2.0