Skip to content

Releases: timerise-ai/ecommerce-process-mining

v0.1.7

Choose a tag to compare

@sowamateusz sowamateusz released this 29 Sep 13:07

Wording release, from scoring the prompt-1 agent eval runs against 0.1.6. The
templates and the suite are unchanged.

Changed

  • SKILL.md quick start, step 1, and references/adaptation.md: no session is
    signed out, in every environment. A request without a session or a tenant is
    a 401 or a redirect to sign-in, never a development user, a sandbox tenant
    or a first-tenant default.
  • SKILL.md framing paragraph tightened to stay within the line budget;
    references/provenance.md records the addition.

v0.1.6

Choose a tag to compare

@sowamateusz sowamateusz released this 29 Sep 12:21

Wording release, from scoring the prompt-1 agent eval runs against 0.1.5. The
templates and the suite are unchanged.

Changed

  • SKILL.md fourth hard rule, the fourth non-negotiable in README.md and
    references/extension-auth.md: the extension-id allowlist is configuration,
    empty by default, with no id written into code. The key pair that pins the id
    belongs to the operator: code built from the skill ships the step that makes
    it and never commits a generated pair.
  • SKILL.md quick start and references/data-model.md: host code goes in files
    of its own, the host's SQL included, for example
    db/process-mining/host.supabase.sql, rather than appended to a template.
  • references/provenance.md records the additions.

v0.1.5

Choose a tag to compare

@sowamateusz sowamateusz released this 29 Sep 11:36

Wording release, from scoring the prompt-1 agent eval runs against 0.1.4. The
templates and the suite are unchanged.

Changed

  • SKILL.md fifth hard rule, the fifth non-negotiable in README.md and
    references/ingest.md: memory is not storage. Without a database, ingest
    answers an error and the extension keeps its queue; an in-process fallback
    store answers success for events it loses and reads consent nobody saved.
  • SKILL.md quick start, step 9: the pinned extension id, the untested capture
    path and the legal basis close the final summary, whatever else it says.
  • SKILL.md quick-start preamble and step 2 tightened to stay within the line
    budget; references/provenance.md records the additions.

v0.1.4

Choose a tag to compare

@sowamateusz sowamateusz released this 29 Sep 10:56

Wording release, from scoring the prompt-1 agent eval runs against 0.1.3. The
templates and the suite are unchanged.

Changed

  • SKILL.md quick start: every // file: block is written as shipped, with the
    renames in references/adaptation.md the only edits, and stricter privacy
    goes in the host's wiring rather than inside a template.
  • SKILL.md quick start and references/testing.md: zod, vitest,
    happy-dom and @testing-library/react are installed from the package
    registry, which is not an external service, and the nine suites run as
    written under npx vitest run, 76 cases; a suite is never ported to another
    runner.
  • SKILL.md quick start and references/screenshots.md: with screenshots off,
    screenshot.ts and its suite still ship, because the mode is a setting read
    by the gate, not a build option.
  • SKILL.md quick start ends with the handover: pin the extension id and add it
    to the allowlist, plan the browser spike, and do not treat the consent toggle
    as a legal basis. Two quick-start steps merged to stay within the line budget.
  • references/provenance.md records these under Added.

v0.1.3

Choose a tag to compare

@sowamateusz sowamateusz released this 28 Sep 14:54

Wording release. The skill content is unchanged from 0.1.2.

Changed

  • The six hard rules in SKILL.md are now the same list, in the same order, as
    the six non-negotiables in README.md: the sixth rule is that a manager never
    sees who opted in, and each rule states the reason it holds. Carrying the
    [P], [D] and [A] tags remains an instruction in the framing
    paragraph.
  • The file table in README.md and the structure in CLAUDE.md list evals/
    and the agent eval workflow, and CLAUDE.md records that evals are not skill
    content.

v0.1.2

Choose a tag to compare

@sowamateusz sowamateusz released this 21 Sep 19:26

Wording release. The skill content is unchanged from 0.1.1.

Changed

  • The Not this table in README.md and the When NOT to use table in
    SKILL.md name the sibling
    browser-extension-connector
    skill, so a request to pull data out of a service with no API lands there
    rather than here. That extension acts inside the user's session; this one
    only observes what the employee does.

v0.1.1

Choose a tag to compare

@sowamateusz sowamateusz released this 21 Sep 19:15

Wording release. The skill content is unchanged from 0.1.0.

Added

  • SKILL.md closes with a line linking the
    Timerise Skills index, so an agent that
    has the skill loaded can find the sibling skills for neighbouring modules without
    leaving the entry point.

Changed

  • CLAUDE.md records the closing line in the SKILL.md layout, and the line budget
    it states holds that line aside.

v0.1.0

Choose a tag to compare

@sowamateusz sowamateusz released this 21 Sep 19:02

Initial release of the ecommerce-process-mining skill: consented DOM-event capture
from the admin tools an e-commerce back office already uses, a gap-capture form for the
work a browser cannot see, and a batch pipeline that turns both into per-role SOPs and a
ranked automation shortlist.

Added

  • SKILL.md entry point: the architecture diagram, ten critical facts, six hard rules,
    the quick-start order, and the reference directory table mapping trigger keywords to
    references/.
  • references/adaptation.md: the seam contract with the host app, the capability table,
    the canonical rename table, the background jobs, module placement shared by the app and
    the extension, and the host probe.
  • references/consent-and-privacy.md: the employee and the customer on the screen, the
    four capture layers in order, pause against revoke against disconnect, the
    excluded-domain module shared by both sides, what the employee must be told, and the
    legal layer per jurisdiction.
  • references/data-model.md: the neutral TypeScript shapes, the Postgres schema with
    row-level security, the append-only consent ledger and its immutability trigger, the
    suppressing enrollment function, the Supabase wiring and the Firestore differences.
  • references/console-surfaces.md: the employee page and the manager page, the
    server-confirmed consent toggle, the gap-capture form and its validation, automation
    ranking, the help page, and their suites.
  • references/ingest.md: the ingest contract, the consent re-read per batch, per-event
    validation, the age window, idempotency, 207 with retry indexes, counters and
    retention.
  • references/pii-scrubber.md: the scrubber, the card-against-barcode rule, IBAN, JWT
    and email masking, sensitive labels, credential URLs, per-deployment configuration and
    the known limits.
  • references/extension-auth.md: the PKCE handshake, the extension-id allowlist, the
    single-statement claim, minting the scoped key, and revoking it.
  • references/extension.md: Manifest V3 with its corrections, the tested control logic
    for pause, budget, queue and session rotation, label resolution and dom_confidence,
    the Chrome wiring, and the assumptions to settle in a spike.
  • references/screenshots.md: the screenshot modes, the capture gate and throttle,
    viewport and crop perceptual hashing, bounding boxes, upload and retention.
  • references/ai-pipeline.md: the three batch stages, routing on dom_confidence,
    embeddings, SOP output and the guardrails.
  • references/ecommerce-playbook.md: which tools to capture in a shop, which processes
    to mine first, the default gap-form categories, the rollout order and the failure
    modes.
  • references/testing.md: what is verified and how, the vitest suites with 76 cases, the
    24 SQL checks in verify.sql, and what to add in the host.
  • references/provenance.md: the engineering ledger, the maturity tags, what the audit
    changed and how the templates verify it, what was kept deliberately, what was designed
    here and has never run, the unverified list, and the fix order for anyone maintaining
    the earlier implementation.
  • Templates for lib/process-mining/*, db/process-mining/schema.sql and verify.sql,
    and the extension/ tree, written to compile under strict and
    noUncheckedIndexedAccess against zod 3 and zod 4.
  • The properties the templates hold: consent per person and per tenant, re-read on every
    batch; one exclusion matcher imported by the extension and the ingest route; values
    that would need scrubbing never read; a scrubber that separates a card number from a
    barcode; a one-time code redeemable exactly once; a consent history nobody can rewrite;
    an ingest response that never claims an unstored event; and an enrollment count
    withheld when it would name the people.
  • README.md: install, activation, the file table, the six non-negotiables,
    requirements, verification, the Not this table and the contributing conventions.
  • CLAUDE.md: editing conventions for this repository.
  • LICENSE: MIT.