Instructions
Jinja2 instruction templates for AI processing tasks
platform/registry/instructions/ contains Jinja2 templates for all AI processing tasks.
Domains
| Domain | Templates | Usage |
|---|---|---|
ocr/ |
full_page_scholarly, standard, historical, letter_*, receipt | OCR processing |
entities/ |
date, event, ner, person, place | Entity extraction |
figures/ |
detection, extraction, caption_extraction | Figure handling |
classification/ |
context_aware, voicememo, correspondence | Content classification |
transcription/ |
audio, diarization | Audio/video |
translation/ |
scholarly, correspondence | Translation |
analysis/ |
interpretation | Content analysis |
shared/ |
xml_page_schema, footnote_format, math_format | Shared components |
inqua/ |
INQUA_*, socials, objectives | INQUA-specific |
Usage
Via Python
from jinja2 import Environment, FileSystemLoader
from pathlib import Path
inst_dir = Path.home() / "ai-mac-studio/platform/registry/instructions"
env = Environment(loader=FileSystemLoader(str(inst_dir)))
tmpl = env.get_template("ocr/full_page_scholarly.j2")
rendered = tmpl.render(language="de", style="scholarly")Via MCP
render_instruction(name="ocr/full_page_scholarly.j2", variables={"language": "de"})Via Admin UI
Browse and preview at /admin/instructions.