Skip to content

Index lookup in python-dev-tca, and the README's business stakes - #5

Merged
kylejtobin merged 4 commits into
mainfrom
docs/index-lookup-and-readme
Oct 7, 2026
Merged

kylejtobin merged 4 commits into
mainfrom
docs/index-lookup-and-readme

Conversation

@kylejtobin

Copy link
Copy Markdown
Owner

What

  • python-dev-tca: the index lookup. After payment, the shop recommends related products through a Qdrant grouped recommend query. The action carries the shop's rules (closeness, one per aisle, the order's market, the limit), one interpreter sends them as the query and reads the reply once, the hits cross into the domain as members, and the reply's shape picks Recommended, NothingRelated or Unavailable. The collection page names how a collection crosses a layer; SKILL.md gains the rule and the diagram.
  • README. The fill example now carries the desk's stakes: what each bug costs, the unions' docstrings, what can no longer happen, and the question each construct settles.
  • AGENTS.md. The modeling loop, with its flowchart.
  • Lint. Ruff's TC001-TC003 are ignored: they require if TYPE_CHECKING:, which TCA forbids.

Checked

  • The lookup was modeled in full and run against a live Qdrant: close matches in two aisles, the currency filter, nothing related, and the index's own 404.
  • Every line of the lookup's code on the skill pages comes from that run; all 40 Python blocks pass ruff format; the diagram renders and every arrow names a box.
  • just check gates: ruff, basedpyright, smell-check and pytest pass.

You are modeling a domain, not building a program. Reread the
python-dev-tca skill, climb the layers bottom up from Scalar to
Crossing, name each layer's nouns, model one file at a time, and after
each file ask against the skill which noun is missing or misplaced,
until no layer has one; with a flowchart of the loop.
TC001-TC003 demand `if TYPE_CHECKING:` blocks, and TCA forbids `if`. On
Python 3.14 annotations are evaluated lazily, so the rules buy nothing.
The shop recommends related products after payment through a Qdrant
grouped recommend query, modeled and run against a live Qdrant.

- The action carries every rule the lookup must follow as scalars:
  closeness, one per aisle, the order's market and the limit.
- One interpreter sends them as the index's query and reads its reply
  once; the hits cross into the domain as members of a collection.
- The reply's shape picks the outcome: Recommended, NothingRelated or
  Unavailable.
- The collection page names how a collection crosses a layer, and
  SKILL.md gains the rule that an action carries its effect's rules.
The fill example now reads on two levels at once: each bug carries its
consequence for the desk, the unions show their docstrings, the table
says what can no longer happen, the "you are about to" table names the
question each construct settles, and each failure mode has an example.
@kylejtobin
kylejtobin force-pushed the docs/index-lookup-and-readme branch from d3ec89b to 13d6379 Compare October 7, 2026 21:09
@kylejtobin
kylejtobin merged commit 4fd7d76 into main Oct 7, 2026
2 checks passed
@kylejtobin
kylejtobin deleted the docs/index-lookup-and-readme branch October 7, 2026 21:10
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant