메뉴
Scaffold a complete extraction pipeline for a new domain — Pydantic model, field definitions, example script, and tests. Use when building a new extraction use case like medical records, invoices, or reviews.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Update LLM model pricing in Prompture. Pricing resolves through a pluggable source registry — by default local KB JSON files (primary, curated) then models.dev (fallback). Use when model prices change, new models launch, models.dev data is stale, or you need to add a custom pricing source.
Scaffold a new LLM provider driver for Prompture. Creates sync + async driver classes, registers them in the driver registry, adds settings, env template, setup.py extras, package exports, discovery integration, and models.dev pricing. Use when adding support for a new LLM provider.
Create a new Prompture usage example script. Follows project conventions for file naming, section structure, docstrings, and output formatting. Use when demonstrating extraction use cases or provider integrations.
Create reusable persona system prompts for Prompture. Covers Persona dataclass, template variables, composition, trait registry, global registry, serialization, and Conversation integration. Use when defining system prompts for extraction or agent behavior.
Add function-calling tools to Prompture agents. Covers ToolDefinition, ToolRegistry, tool_from_function, decorator patterns, serialization formats, and Conversation integration. Use when adding callable tools for LLM function calling.
Add predefined field definitions to the Prompture field registry. Handles field structure, categories, template variables, enum support, and thread-safe registration. Use when adding reusable extraction fields.
| name | scaffold-extraction |
| description | Scaffold a complete extraction pipeline for a new domain — Pydantic model, field definitions, example script, and tests. Use when building a new extraction use case like medical records, invoices, or reviews. |
| metadata | {"author":"prompture","version":"1.0"} |
Builds the full pipeline for a new extraction domain: model, fields, example, and tests.
Ask the user:
ollama/gpt-oss:20b| Scenario | Method |
|---|---|
| Simple model, < 8 fields | extract_with_model (1 LLM call) |
| Complex model, 8+ fields | stepwise_extract_with_model (N calls, more accurate) |
| No Pydantic model, raw schema | extract_and_jsonify |
| Structured input (CSV, JSON) | extract_from_data (TOON, 45-60% token savings) |
| DataFrame input | extract_from_pandas |
| Non-JSON output | render_output |
from pydantic import BaseModel, Field
class InvoiceData(BaseModel):
vendor_name: str = Field(description="Company that issued the invoice")
total_amount: float = Field(description="Total amount due")
currency: str = Field(default="USD", description="ISO 4217 currency code")
Rules:
Field(description=...) on every field — these become LLM instructions"", 0, 0.0, []Optional[T] = None only when a field genuinely might not existRegister in prompture/field_definitions.py only if fields are general-purpose. Use the add-field skill.
Create examples/{domain}_extraction_example.py. Use the add-example skill.
Unit test for the Pydantic model + integration test for end-to-end extraction. Use the add-test skill.
from prompture import extract_with_model
result = extract_with_model(
model_cls=InvoiceData,
text=invoice_text,
model_name="ollama/gpt-oss:20b",
instruction_template="Extract all invoice details from the following document:",
)
invoice = result["model"] # Pydantic instance
usage = result["usage"] # Token counts and cost