Knowledge Packs — Stunspot’s Guide to Gastronomic Engineering

The repository contains three operational corpus formats. They all represent the same canon, but they are optimized for different upload, retrieval, and reading conditions.

The canonical individual reports live in knowledge-packs/by-report/. The compiled packs and omnibus are convenience formats generated from those reports.


Use the compiled packs unless you have a specific reason not to.

They are the best default for AI Projects, RAG workspaces, NotebookLM-style tools, and long-context systems with file-count limits. Three files are usually easier for a model to route than ten separate reports, while still preserving the major canon arcs better than one giant omnibus.


Format Selection Matrix

Pack Type Files Path Best Use Trade-Off
Source reports 10 knowledge-packs/by-report/ Precise retrieval, selective upload, exact citation, editing, and inspection. More files to manage; requires better routing.
Compiled packs 3 knowledge-packs/compiled-packs/ Recommended default: grouped coverage with manageable file count. Less granular than source reports.
Omnibus 1 knowledge-packs/omnibus/ One-file import, local archive, whole-corpus search, strong long-context systems. Large single file; weaker for pinpoint retrieval in some tools.

Use Source Reports When…

Choose knowledge-packs/by-report/ when the task requires precision.

Good use cases:

Source reports:


Use Compiled Packs When…

Choose knowledge-packs/compiled-packs/ for most model-facing workflows.

Good use cases:

Compiled packs:

  1. Gastronomic Engineering — Vol. 1 A-C — Fundamentals and Worldview
  2. Gastronomic Engineering — Vol. 2 D-F — Constraint Layers for Specialty Dietary Needs I
  3. Gastronomic Engineering — Vol. 3 G-J — Constraint Layers for Specialty Dietary Needs II

Recommended model instruction:

Use these compiled packs as the active gastronomic engineering knowledge substrate. Preserve report letters and source paths where possible. Route formulation questions through phase-state analysis, lost-function mapping, constraint identification, and control-variable diagnosis before recommending substitutions or process changes.

Use the Omnibus When…

Choose knowledge-packs/omnibus/ when you need the entire corpus in one file.

Good use cases:

Omnibus:

Use the omnibus with caution in weaker long-context environments. If the model starts answering generically or loses report boundaries, switch back to compiled packs.


RAG Ingestion Recommendations

For retrieval systems, treat the corpus as a structured doctrine set, not a flat pile of Markdown.

Minimum metadata to preserve:

Recommended routing hints:

Domain First Report Route
food physics, phase states, control variables A
closed-loop formulation, predictive control B
experimental protocols, practicums C
low-carb, keto, glycemic control, sugar-free D
high-protein, satiety, protein fortification E
low-calorie, volumetric, energy density F
whole-food, paleo, Whole30, clean-label G
vegan, vegetarian, dairy-free, egg-free H
allergen-free, gluten-free, multi-allergen I
low-FODMAP, microbiome, reflux, digestive tolerance J

Chunking guidance:


Directory Boundary

docs/ is only for navigation and usage guidance. Do not create docs/reports/ or mirror the corpus into the docs tree. The corpus belongs under knowledge-packs/.