Files
MealQuest/internal/shopping/title_test.go
T
dev_agentandClaude Opus 5.5 fdfd0123fb
CI / ci (push) Successful in 4m48s
CI / ci (pull_request) Successful in 4m53s
Add the archive_title skill and title Supply runs on generation
archive_title v1 writes a Supply run's post title for the Supply archive
from the week's dinner names in day order. Validation requires 1-48
characters on one line, no emoji and no surrounding quotation marks; a
reply breaking a rule gets the harness's one repair turn.

shopping.Generate and GeneratePushed run it, outside the generation's
transaction, whenever the stored list is still untitled, and store
shopping_lists.title only while the list is untitled and not frozen. A
title failure never fails the generation: the list stays untitled, the
failure is logged and recorded in ai_runs, and the week's next
generation tries again. A titled list keeps its title on remakes.

Maestro task fa4b94c9-6623-41ab-8d2a-0ca7d1e6401a.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-08 05:31:59 -04:00

233 lines
7.7 KiB
Go

package shopping_test
import (
"database/sql"
"encoding/json"
"fmt"
"strings"
"testing"
"code.cubecraftlabs.com/Maniacs_Mansion/Meal-Quest/internal/ai/harness"
"code.cubecraftlabs.com/Maniacs_Mansion/Meal-Quest/internal/ai/llm"
"code.cubecraftlabs.com/Maniacs_Mansion/Meal-Quest/internal/db"
"code.cubecraftlabs.com/Maniacs_Mansion/Meal-Quest/internal/shopping"
)
// titled is archive_title's reply giving title.
func titled(title string) llm.GenerateResult {
js := fmt.Sprintf(`{"title":%q}`, title)
return llm.GenerateResult{Response: llm.Response{Text: js, JSON: json.RawMessage(js)}}
}
// titleWeek stores three dinners, one with no recipe, and plans them on
// the week's first three days.
func titleWeek(t *testing.T) *env {
t.Helper()
e := newEnv(t)
e.dinner(t, "Beef Chili", ing{"ground beef", "1", "lb"})
e.dinner(t, "Fish Tacos")
e.dinner(t, "Pad Thai", ing{"rice noodles", "8", "oz"})
e.plan(t, "Pad Thai", "Fish Tacos", "Beef Chili")
return e
}
// storedTitle returns the week's list's stored title; nil when NULL.
func (e *env) storedTitle(t *testing.T) *string {
t.Helper()
var title sql.NullString
if err := e.st.Reader.QueryRow(`SELECT title FROM shopping_lists WHERE week_start = ?`, week.String()).Scan(&title); err != nil {
t.Fatal(err)
}
if !title.Valid {
return nil
}
return &title.String
}
// runStatuses returns the status of every archive_title run recorded in
// ai_runs, oldest first.
func (e *env) runStatuses(t *testing.T) []string {
t.Helper()
rows, err := e.st.Reader.Query(`SELECT status FROM ai_runs WHERE skill = 'archive_title' ORDER BY created_at, rowid`)
if err != nil {
t.Fatal(err)
}
defer rows.Close()
var out []string
for rows.Next() {
var s string
if err := rows.Scan(&s); err != nil {
t.Fatal(err)
}
out = append(out, s)
}
if err := rows.Err(); err != nil {
t.Fatal(err)
}
return out
}
func wantTitle(t *testing.T, what string, got *string, want string) {
t.Helper()
if got == nil || *got != want {
t.Errorf("%s = %v, want %q", what, deref(got), want)
}
}
func wantUntitled(t *testing.T, what string, got *string) {
t.Helper()
if got != nil {
t.Errorf("%s = %q, want none", what, *got)
}
}
func deref(s *string) any {
if s == nil {
return nil
}
return *s
}
// TestGenerateTitlesNewList: a new list is titled by archive_title from
// the plan's dinners in day order, a dinner with no group included, and
// the title is stored and returned.
func TestGenerateTitlesNewList(t *testing.T) {
e := titleWeek(t)
e.model.QueueGenerate(titled("Noodles, Tacos and Chili"))
g := e.generate(t)
wantTitle(t, "returned title", g.List.Title, "Noodles, Tacos and Chili")
wantTitle(t, "stored title", e.storedTitle(t), "Noodles, Tacos and Chili")
got, err := e.svc.Get(t.Context(), week)
if err != nil {
t.Fatal(err)
}
wantTitle(t, "title read back", got.Title, "Noodles, Tacos and Chili")
reqs := e.model.Requests()
if len(reqs) != 1 {
t.Fatalf("%d model calls, want 1", len(reqs))
}
want := "Title the week whose dinners are, in day order from Monday:\n- Pad Thai\n- Fish Tacos\n- Beef Chili"
if user := reqs[0].Messages[0].Content; user != want {
t.Errorf("user message = %q, want %q", user, want)
}
if s := e.runStatuses(t); len(s) != 1 || s[0] != string(harness.StatusOK) {
t.Errorf("archive_title runs = %q, want one ok", s)
}
}
// TestGenerateTitleTooLongRepaired: a title over 48 characters is repaired
// through the repair turn, and the repaired title is stored, not a
// truncation of the first.
func TestGenerateTitleTooLongRepaired(t *testing.T) {
e := titleWeek(t)
long := "Pad Thai Monday, Fish Taco Tuesday and Chili Wednesday"
e.model.QueueGenerate(titled(long), titled("Noodles, Tacos and Chili"))
g := e.generate(t)
wantTitle(t, "returned title", g.List.Title, "Noodles, Tacos and Chili")
wantTitle(t, "stored title", e.storedTitle(t), "Noodles, Tacos and Chili")
reqs := e.model.Requests()
if len(reqs) != 2 || !strings.Contains(reqs[1].Messages[2].Content, "has 54 characters; shorten it to at most 48") {
t.Fatalf("model calls = %+v, want a repair turn asking for a shorter title", reqs)
}
if s := e.runStatuses(t); len(s) != 1 || s[0] != string(harness.StatusRepaired) {
t.Errorf("archive_title runs = %q, want one repaired", s)
}
}
// TestGenerateTitleFailureKeepsList: archive_title giving an emoji twice
// fails, and the list is still made, whole and untitled, with the failure
// logged and recorded in ai_runs.
func TestGenerateTitleFailureKeepsList(t *testing.T) {
e := titleWeek(t)
e.model.QueueGenerate(titled("Taco Week 🌮"), titled("Taco Week 🌮"))
g, err := e.svc.Generate(t.Context(), week)
if err != nil {
t.Fatalf("Generate = %v, want the list made though its title failed", err)
}
want := []string{
"0 2026-10-12 Pad Thai: 0 rice noodles [8 oz] needed",
"1 2026-10-14 Beef Chili: 0 ground beef [1 lb] needed",
}
wantLines(t, "list", lines(g.List), want)
wantLines(t, "stored list", e.stored(t), want)
if len(g.Skipped) != 1 || g.Skipped[0].DinnerName != "Fish Tacos" {
t.Errorf("skipped = %+v, want Fish Tacos", g.Skipped)
}
wantUntitled(t, "returned title", g.List.Title)
wantUntitled(t, "stored title", e.storedTitle(t))
if g.List.Status != shopping.StatusCurrent {
t.Errorf("status = %q, want current", g.List.Status)
}
log := e.log.String()
for _, part := range []string{"level=WARN", "Supply run left untitled", "week=2026-10-12", "has an emoji (U+1F32E)"} {
if !strings.Contains(log, part) {
t.Errorf("log lacks %q:\n%s", part, log)
}
}
if s := e.runStatuses(t); len(s) != 1 || s[0] != string(harness.StatusBadOutput) {
t.Errorf("archive_title runs = %q, want one bad_output", s)
}
}
// TestGenerateRetriesTitle: a list left untitled is titled by the week's
// next generation, Generate or GeneratePushed; once titled, a remake keeps
// its title and asks for none.
func TestGenerateRetriesTitle(t *testing.T) {
e := titleWeek(t)
e.model.QueueGenerate(llm.GenerateResult{Err: fmt.Errorf("%w: model down", llm.ErrProvider)})
first := e.generate(t)
wantUntitled(t, "first title", first.List.Title)
if !strings.Contains(e.log.String(), "model down") {
t.Errorf("log lacks the failure:\n%s", e.log.String())
}
e.model.QueueGenerate(titled("Noodles, Tacos and Chili"))
plan, err := db.New(e.st.Reader).GetWeekPlan(t.Context(), week.String())
if err != nil {
t.Fatal(err)
}
second, err := e.svc.GeneratePushed(t.Context(), week, plan.Revision)
if err != nil {
t.Fatal(err)
}
if second.List.ID != first.List.ID {
t.Fatalf("second generation made list %s, want list %s again", second.List.ID, first.List.ID)
}
wantTitle(t, "retried title", second.List.Title, "Noodles, Tacos and Chili")
wantTitle(t, "stored title", e.storedTitle(t), "Noodles, Tacos and Chili")
e.swap(t, "2026-10-12", "Beef Chili")
third := e.generate(t)
wantTitle(t, "title after a remake", third.List.Title, "Noodles, Tacos and Chili")
if n := len(e.model.Requests()); n != 2 {
t.Errorf("%d model calls, want 2: a titled list is not titled again", n)
}
}
// TestGenerateTitleNotStoredOnFrozenList: a list frozen into the Supply
// archive while its title is written keeps no title, and the failure is
// logged.
func TestGenerateTitleNotStoredOnFrozenList(t *testing.T) {
e := titleWeek(t)
e.model.QueueGenerate(titled("Noodles, Tacos and Chili"))
e.beforeTitle = func() {
if _, err := e.st.Writer.Exec(`UPDATE shopping_lists SET status = 'archived', frozen_at = '2026-10-19T04:05:00Z'`); err != nil {
t.Error(err)
}
}
g := e.generate(t)
wantUntitled(t, "returned title", g.List.Title)
wantUntitled(t, "stored title", e.storedTitle(t))
if !strings.Contains(e.log.String(), "the Supply run was frozen or replaced before its title was stored") {
t.Errorf("log lacks the failure:\n%s", e.log.String())
}
}