Paste a model card, pick a teacher, and let ALAIN assemble outline JSON, runnable sections, and validators so launch and enablement teams stay in lockstep.
Get started with any model by pasting its Hugging Face repo (owner/model) or full URL.
Research Scout, Lesson Architect, and Outline Builder lock model context into deterministic JSON before a single section is drafted.
Semantic Reviewer and Quality & Colab Fixer catch placeholders, markdown/code drift, and runtime issues before anything ships.
ALAIN works with Poe, OpenAI-compatible gateways, and local Ollama or vLLM so the same lesson contract follows your deployment.
ALAIN assembles markdown, runnable cells, validators, and analytics metadata before you ever download the `.ipynb`. Review it, edit it, or rerun any stage without losing the thread.
Paste a model card or spec. ALAIN extracts guardrails, objectives, and constraints into structured JSON so every downstream stage stays aligned.
Editors can pause here, adjust the outline, and replay the remaining stages with a single command.
from alain_kit.runtime import client
client.health_check(model="gpt-oss-20b")
outline, sections = client.generate_notebook(ref="TheBloke/gpt-oss-20b")
print(f"Outline steps: {len(outline['steps'])}")Eight observable stages move a messy model card into a reviewed lesson. Every checkpoint stores artifacts and timings so you can replay just the part that needs love.
Digest cards, specs, and community notes into a structured brief you can reuse.
Define learners, objectives, and assessments so downstream stages stay aligned.
Capture titles, references, and sequenced steps in deterministic outline JSON.
Expand each step into balanced markdown, runnable code, and reproducibility callouts.
Catch placeholders, missing next steps, or uneven pacing before reviewers ever look.
Audit clarity, terminology, and completeness with targeted repair notes.
Apply Colab runtime fixes automatically and capture expected runtimes for handoff.
Assemble the notebook, validation summary, and artifacts so you can export with confidence.
Swap teachers and runtimes without rewriting your docs. The same prompts and validators travel from hosted Poe to OpenAI-compatible APIs to local Ollama or vLLM runs.
Fastest path to GPT-OSS teachers—just add `POE_API_KEY`.
Point at existing enterprise infra and reuse your gateway policies.
Keep lessons fully offline with the same prompts and validations.