Section 3.2 of the test plan, which keeps the app overview, the feature status, the known limitations and the index of areas. Since TFP-253 each area’s cases are in a file of their own; a ticket names the area it touches.
| ID | Test | Expected Result | Status |
|---|---|---|---|
| GEN-01 | Suite-creation UI renders (automated smoke: “renders critical suite creation UI”) | Header shows “Mr Onion”; the “Test Suite Creation” panel and the “Generated Suite” section are both visible before any suite is generated. | ✅ |
| GEN-02 | Generate a suite from a plain-English prompt (automated smoke: “generates a Playwright suite from prompt input”) | Submitting Base URL + description produces a non-empty .spec.ts suite in the Generated Suite panel. |
✅ |
| GEN-03 | Creation inputs, generation mode, and best-practice inference (automated: “covers the creation inputs, generation mode, and inferred best-practice state”) | The creation form correctly reflects which generation mode (template vs. AI) is active and which best-practice coverage areas are inferred from the description. | ✅ |
| GEN-04 | SEO coverage is generated when SEO-relevant input is given (automated: “covers SEO metadata and crawler-facing files”) | Generated suite includes SEO checks (meta description, crawler-facing files) when the description or selected practice areas call for them. | ✅ |
| GEN-05 | SEO coverage is withheld when Functional is explicitly selected (automated: “does not generate SEO coverage when Functional is explicitly selected”) | Selecting Functional explicitly, without SEO, produces a suite with no SEO test block. | ✅ |
| GEN-06 | Content-profile SEO checks withheld under the same condition (automated: “does not generate content-profile SEO checks when Functional is explicitly selected”) | No content-profile-derived SEO checks appear in the generated suite when Functional was explicitly and exclusively selected. | ✅ |
| GEN-07 | Regeneration required after changing selected best-practice areas (automated: “requires regeneration before running when selected best-practice areas change”) | Changing the selected best-practice areas after a suite exists disables Run until the suite is regenerated. | ✅ |
| GEN-08 | Manual: generate with no OPENAI_API_KEY set |
Template-based generator produces a suite; no request to OpenAI is made. | ⬜ |
| GEN-09 | Manual: generate with OPENAI_API_KEY set |
OpenAI path produces a suite with descriptive test names, labelled generatedBy: "openai" with the model picked from src/lib/generate/models.ts. |
⬜ |
| GEN-10 | Generate executes with no external backend configured (automated: tests/regression/mr-onion.sections.spec.ts — “executes generate, run, crawl, introspect, and save-suite without an external backend configured”) |
POST /api/generate succeeds and returns real Playwright code, with no environment variable set beyond .env.development. |
✅ |
| GEN-11 | An element hidden at introspection time does not get an unconditional visibility claim | A synthetic element with isVisible: false generates toBeAttached() plus a “Hidden at introspection time” comment, not toBeVisible(). Evidence. |
✅ |
| GEN-12 | A detected toggle group generates an ARIA-state assertion | A synthetic 3-member groupId-sharing group generates a presence test (every member exposes an ARIA state attribute) and, since one member’s ariaChecked is "true", a toggle-behavior test. Evidence. |
✅ |
| GEN-13 | Accessibility suite asserts a real accessible name and role per discovered element | A visible element gets a toHaveAccessibleName(/.+/) and a role check; a hidden one gets toBeAttached() instead. Evidence. |
✅ |
| GEN-14 | Responsive suite checks real discovered elements per viewport and adds a touch smoke test | Up to 5 visible elements are checked at mobile, tablet and desktop; a touch smoke test claims only that a tap doesn’t error. Evidence. | ✅ |
| GEN-15 | Usability suite asserts content-specific claims instead of an interactive-element count | Labelled elements aren’t a generic phrase (submit, click here); up to 5 form or action elements are above the 900px fold. Evidence. |
✅ |
| GEN-16 | A detected action pair (e.g. Undo/Redo) gets a negative-assertion regression guard, and only a real pair does | A matched opposite pair (undo/redo, save/discard) gets one test that an unrelated interaction changes neither; an unpaired element gets none. Evidence. | ✅ |
| GEN-17 | A detected drop target gets valid- and invalid-content drag-and-drop tests | A valid-content drop (a 1x1 PNG) must change the DOM without a page error; an invalid one (a text file) must not crash. Evidence. | ✅ |
| GEN-18 | Candidate-journey review screen: ranked journeys, nothing pre-selected | “Generate selected” sends exactly the selection as selectedJourneyIds. Evidence. |
✅ |
| GEN-19 | End-to-end journey tests, one per selected journey | Each walks its real pages and elements, asserts each step and a named flow’s exit; skipped unless interactive tests are on. Evidence. | ✅ |
| GEN-20 | Named flows can be entered in the UI and behave exactly as if sent to the API | A “Named flows” textarea in the Content Rules section takes one flow per line as label: /entry -> /exit (=> and the arrow character also accepted; a missing leading slash is added). Evidence. |
✅ |
| GEN-21 | Local model generates a suite | With MRONION_OLLAMA_MODEL set, /api/generate returns 200, generatedBy: "ollama", the model’s name and unfenced code, even with OPENAI_API_KEY set. Evidence. |
✅ |
| GEN-22 | Local path: Ollama not running (TFP-105; automated: “Ollama not running falls back as ollama_unreachable”) | Template suite, generatedBy: "fallback", fallbackReason: "ollama_unreachable", returned promptly — a refused connection does not wait out the 180s budget. |
✅ |
| GEN-23 | Local path: model never created (TFP-105; automated: “the local model never created falls back as ollama_model_missing”) | Ollama answers but lists only the raw qwen2.5-coder:7b: template suite, fallbackReason: "ollama_model_missing", and no completion request is sent — the route never falls through to the raw model. |
✅ |
| GEN-24 | Local path: timeout (TFP-105; automated: “a hung local model times out within its own budget and falls back”) | A hung model is abandoned at MRONION_OLLAMA_TIMEOUT_MS (default 180s, independent of the OpenAI path’s 30s) with fallbackReason: "timeout". |
✅ |
| GEN-25 | Local path: Ollama returns an error or an empty answer | HTTP 500 from Ollama → fallbackReason: "ollama_error", not ollama_unreachable. Evidence. |
✅ |
| GEN-26 | Local path: a suite that doesn’t type-check is never labelled model output | The local model’s suite is type-checked in-process against the installed @playwright/test (TypeScript compiler API, strict: false, only errors in the generated file count). Evidence. |
✅ |
| GEN-27 | Local path: accessibility requests generate real checks, not the template | Only on the local path, src/lib/generate/local-rules.ts is appended to the system message. Evidence. |
✅ |
| GEN-28 | Local path: per-category baseline | One real generation per best-practice area, all 10 in BEST_PRACTICE_OPTIONS (the Phase 31 record listed 9 and missed reliability), against testforge-qwen7b through a production build with the TFP-106 rules. Evidence. |
✅ |
| GEN-29 | Local path: every assertion is awaited | The shared block of LOCAL_MODEL_RULES, not a category section, says to write await before every expect(...), because an unawaited assertion on page or a locator never runs. Evidence. |
✅ |
| GEN-30 | Local path: the rules example’s URL does not leak into suites | ACCESSIBILITY_EXAMPLE navigates only to https://example.invalid/EXAMPLE-ONLY, and the rules text contains no localhost. Evidence. |
✅ |
| GEN-31 | Local path: a suite with a never-run assertion is rejected | After the type check passes, generateWithOllama runs findUnawaitedAssertions (GEN-29). Evidence. |
✅ |
| GEN-32 | Local path: reliability tests simulate a real failure | Reliability here means Mr Onion’s own definition: graceful degradation when a dependency or storage fails. Evidence. | ⚠️ |
| GEN-33 | Local path: performance tests measure time | The guide requires a timing measurement against a budget: Date.now() around page.goto, or Navigation Timing through page.evaluate. Evidence. |
✅ |
| GEN-34 | Local path: SEO tests check what search engines read | The guide asks for the title (toHaveTitle), the meta description (meta[name="description"] content) and a single level-1 heading, with the real text used if the request or profile gives it. Evidence. |
✅ |
| GEN-35 | Local path: database tests check persistence through the page | The guide asks the suite to save a unique value (Date.now()), reload, and check the value is still shown. Evidence. |
✅ |
| GEN-36 | Local path: security tests don’t assume HTTPS | HTTPS is asserted only behind a positive test.skip for http pages; invalid input must not be accepted; script-like input must not open a dialog. Evidence. |
✅ |
| GEN-37 | Generator comparison, template vs. local model, against real targets | One command generates both suites from the same request and introspection data through the real /api/generate. Evidence. |
✅ |
| GEN-38 | Naming performance in the request produces a performance test on every target | The shared template builder used to skip its performance test whenever the description said “performance”, “load time”, “speed” or “response time”. Evidence. | ✅ |
| GEN-39 | The template writes a database persistence round trip | For database, which previously had only a manual follow-up comment, the template finds forms worth a round trip in the inspected pages. Evidence. | ✅ |
| GEN-40 | The template’s performance checks measure real load behaviour and name the culprit | A performance suite has four read-only checks with Phase 34 Decision 2’s loose budgets. Evidence. | ✅ |
| GEN-41 | The template checks security on every target, not only HTTPS ones | The template’s only security check was “the URL is HTTPS”, written only for https targets; an http target got a note and nothing else. Evidence. | ✅ |
| GEN-42 | Selectors built from page values survive quotes and backslashes | quoteSelector escapes the whole selector for its JS string, so values holding a quote or backslash compile and select the right element. Evidence. |
✅ |
| GEN-43 | Option B: the local model fills only the areas the template left empty | With MRONION_OLLAMA_MODEL unset, the response is exactly the template-only one. Evidence. |
✅ |
| GEN-44 | Gap-filling skips areas the template found nothing to test in, and the local block runs only with interactive tests on | An area the template found nothing to test in is notApplicable, not a gap; the local block runs only with interactive tests on. Evidence. |
✅ |
| GEN-45 | Option B measured: template vs. template + gap-fill, against the fixture and a real target | The comparison now pits the template alone against the template with the local model’s gap-fill, both through the real /api/generate. Evidence. |
✅ |
| GEN-46 | Every form field on the inspected pages has an accessible name | One test per inspected page with fields: a field with no accessible name fails; one named only by its placeholder is a finding. Evidence. | ✅ |
| GEN-47 | A generate request with no page data is inspected automatically | When /api/generate gets no contentProfile, it runs the same introspection the app’s own flow runs, by calling inspectPages directly (not over HTTP). Evidence. |
✅ |
| GEN-48 | The local-model benchmark measures a configuration repeatably | Each case’s messages are built by the route’s own buildGapFillMessages and sent through generateLocalSuite, its call and gates. Evidence. |
✅ |
| GEN-49 | The local model configuration is chosen by measurement | Eight configurations, 21 suites each, on the reference machine (RTX 2050, 4 GB); C8 qualifies and is fastest: median 39.5 s against C1’s 68.2 s. Evidence. | ⚠️ |
| GEN-50 | A healthy real site doesn’t turn red on the template’s own mistakes | Four failures on the-internet, each a template error, fixed: the sign-in check visits a page that has one, and link locators match what renders the link. Evidence. | ✅ |
| GEN-51 | The cookie/session test can fail | The always-true check is gone; a session cookie valid over 30 days or set for a parent domain fails; no session cookie is Not run. Evidence. | ✅ |
| GEN-52 | Optional SEO items are findings; baseline SEO items fail, naming what is missing and where | Missing structured data or web-app manifest is a finding; missing baseline items (meta description, canonical link, robots.txt) fail, named per page. Evidence. | ✅ |
| GEN-53 | No two tests in one generated file share an ID | IDs are TF<area code>-<nnn>. Evidence. |
✅ |
| GEN-54 | Phase 33’s category bar on C8, the adopted local model | Three generations per category on testforge-qwen7b, read by hand: performance, SEO, security, accessibility 3/3; reliability and database 1/3; functional 0/3 clean. Evidence. |
⚠️ |
| GEN-55 | The gap-fill request is a plain request for the gap areas | buildGapFillMessages returns exactly buildGenerationMessages for the gap areas; the sentence that cut reliability and database to 1/3 is gone. Evidence. |
✅ |
| GEN-56 | Gate: a Promise handed to a plain matcher is rejected | findPromisesToPlainMatchers rejects an expect of a promise followed by a plain matcher, and the template is used instead. Evidence. |
✅ |
| GEN-57 | Gate: tests that repeat one body are rejected | Two or more tests with the same body and setup reject the suite as duplicate_tests; titles play no part. Evidence. |
✅ |
| GEN-58 | The template’s output is proven unchanged between two refs by a committed check | Each ref’s own mockGenerate runs on 172 fixed inputs (scripts/template-unchanged/inputs.mjs), compared byte for byte. Evidence. |
✅ |
| GEN-59 | Comprehensive security, passive level: read-only checks on every page, tagged by OWASP Top 10:2025 and ASVS 5.0 | Four tests per page, one page load each, plus one probe per run (src/lib/generate/comprehensive-security.ts). Evidence. |
✅ |
| GEN-60 | Request recording: the passive level sends nothing but page loads and one missing-page probe | The generated suite for three pages runs through the Playwright CLI against a local server that records every request. Evidence. | ✅ |
| GEN-61 | Comprehensive accessibility: an axe-core WCAG 2.2 A/AA scan on every page, split by impact, with undecided items listed for manual review | @axe-core/playwright 4.13.0 and axe-core 4.13.0 (MPL-2.0) are added unmodified. Evidence. |
✅ |
| GEN-62 | The local model’s scenarios in a Comprehensive run: shown, labelled, outside the verdict | A Comprehensive run asks the model for that area under every gate, and adds an accepted suite in its own Local model block. Evidence. |
✅ |
| GEN-63 | The anchor-target check never crashes on href=”#” | href="#" no longer becomes the CSS selector #: the check, which crashed on 5 of 5 rizonetech.com pages, runs on every link. Evidence. |
✅ |
| GEN-64 | The hover-state and HTTPS/SSL link checks bound their own waits and name what they could not check | The hover check no longer waits forever on an off-screen link, and the HTTPS/SSL check bounds its requests; both name what they couldn’t check. Evidence. | ✅ |
| GEN-65 | The Standard and passive-security checks flag the same cookies for missing HttpOnly/Secure | Both checks judge the same cookies (on rizonetech.com they had named 6 against none). Evidence. | ✅ |
| GEN-66 | The Standard accessible-name check tests the link it was generated for | A link with a target is found as introspection recorded it, by its href, text and position, not the first element holding its text. Evidence. | ✅ |
| GEN-67 | Every generated test declares one of three gates, and its skip matches its gate | Each test carries a static { annotation: { type: 'gate', description: 'read-only' / 'interactive' / 'active' } }. Evidence. |
✅ |
| GEN-68 | The accessible-name check follows HTML-AAM 4.1.10 for an <img alt="">, not Playwright’s departure from it |
HTML-AAM 4.1.10 (checked 2026-10-05) gives an image with alt="" no name and uses title only when there is no alt attribute; Playwright 1.59.1 falls back to the title. Evidence. |
✅ |
| GEN-69 | The Standard run’s security checks are the Comprehensive run’s passive checks, on every inspected page, with one missing-page probe per run | Standard security calls Comprehensive’s passive builders per inspected page, then one request to a path that can’t exist. Evidence. | ✅ |
| GEN-70 | The Standard run runs Comprehensive’s axe-core WCAG 2.2 A and AA scan on every inspected page, and lists what axe could not decide for manual review | An axe scan per inspected page, after the existing checks, with its helpers written once. Evidence. | ✅ |
| GEN-71 | Performance graded in Core Web Vitals’ three bands | LCP and CLS pass, give a finding or fail (web.dev’s bands); TBT is recorded only; findings and failures say they were measured once, in a lab. Evidence. | ✅ |
| GEN-72 | A link in a closed menu or behind a collapsed navbar’s toggle is checked as what it is | A hidden link is checked by its address, then its control (an aria-expanded element naming its hidden ancestor) is clicked to reveal it; failing that, a finding. Evidence. |
✅ |
| GEN-73 | The form check finds the <form> itself and the fields introspection recorded in it |
Introspection records each <form>: its action, method, occurrence, and its fields in document order. Evidence. |
✅ |
| GEN-74 | The accessible-name check tests the button, field or form it was generated for, and doesn’t ask an element with no role for a name | A control with no id, label, placeholder or role-and-name is found among its own kind by its recorded text, visibility and position. Evidence. | ✅ |
| GEN-75 | The orphan-page check judges only pages that were loaded | Inspection (inspectPages, /api/introspect) adds requestedPaths (in order) and failedPages ({ path, reason }) to the profile. Evidence. |
✅ |
| GEN-76 | A #/ link is a page on a hash-routed site: the crawl and inspection find each route, named as the address bar shows it |
A # link whose fragment starts with / or !/ is a page; #top, #details and a bare # stay in-page anchors. Evidence. |
✅ |
| GEN-77 | A linked route that the app draws as “not found” fails link health, naming it | Once per run, the app’s view of a route that can’t exist is read, and a linked route showing the same is a failure. Evidence. | ✅ |
| GEN-78 | On a hash-routed site the orphan check counts pages by route | For a profile with requestedPaths (TFP-161’s branch), each page and each of its links is counted by its page key, origin + pagePathOf (TFP-165): a hash route keeps its #/..., an in-page anchor is dropped. Evidence. |
✅ |
| GEN-79 | Every page visit in a generated test waits for the page to draw | Every page.goto becomes tfGoto, which loads the page as asked and waits up to 5 s for the network to go quiet; the wait never fails a test. Evidence. |
✅ |
| GEN-80 | A site’s crawled pages are grouped into types by address pattern | Identifier segments become :id, segments varying under one parent :slug; top-level pages stay literal; hash routes group by page name. Evidence. |
✅ |
| GEN-81 | Each page’s outline refines its address type: equal structural signatures merge, different ones split; each type gets a report-only label | Landmarks, forms and their fields, repeated cards, headings and interactive count make each page’s outline. Evidence. | ✅ |
| GEN-82 | The page grouping as revised at Gate A: address types first, split only by form signature, merge only siblings under one immediate parent, never top-level pages or home; no slug before an :id |
A segment directly before an :id names a collection and is never a slug. Evidence. |
✅ |
| GEN-83 | Page types are ranked and Standard covers one page per type up to a hard cap of 8 | Home, every type with a form and the largest type rank first; only home is never cut by the hard cap of 8. Evidence. | ✅ |
| GEN-84 | Comprehensive covers every page type before a second page of any | One page (the representative) of every type in rank order; types beyond the budget are skipped, named. Evidence. | ✅ |
| GEN-85 | The URL field takes a list of pages | One URL behaves as today. Evidence. | ✅ |
| GEN-86 | Hidden elements in the remaining checks | Deep mode takes a hidden link by its own address, not by clicking; reaching it stays the keyboard check’s question. Evidence. | ✅ |
| GEN-87 | The download checks’ pattern keeps its escaped dot | The pattern’s dot stays escaped, so .../getpdf is no longer counted as a download. Evidence. |
✅ |
| GEN-88 | Model and OpenAI page visits go through tfGoto |
Model-written blocks route every page.goto through tfGoto, defined once per file. Evidence. |
✅ |
| GEN-89 | Controls on read-only pages do their job: sort, filter, stepper, link destination, option lists | Per inspected page: sort options order the items, filters send and respect values, steppers change by one, option lists show no error text, nav links land where they say. Evidence. | ✅ |
| GEN-90 | Tab order follows the reading order, and checkboxes and radio buttons that belong together are grouped | Two read-only tests per page, from one builder, in Comprehensive accessibility and Standard’s accessibility suite. Evidence. | ✅ |
| GEN-91 | Comprehensive security probes known paths that leak, once per run, and records guessable sequential ids per page | Once per run, one GET to each of 19 known paths per same-site origin; guessable sequential ids recorded per page. Evidence. | ✅ |
| GEN-92 | Comprehensive security records API fields with internal-sounding names that a page fetches and never shows | A finding for a fetched JSON field with an internal-sounding name whose value the page never shows; Comprehensive only. Evidence. | ✅ |
| GEN-93 | Layout consistency, overlap and images that don’t render | Findings for inconsistent fonts and alignment in forms, lists and cards; failures for overlapping controls and for images that don’t render. Evidence. | ✅ |
| GEN-94 | The spelling of interface text | Words of four or more letters in interface text are checked against the English dictionaries. Evidence. | ✅ |
| GEN-95 | A control inside its label is named as the browser names it | The inspection reads a label’s text without the labelled control’s own content, as Chromium’s accessible name does. Evidence. | ✅ |
| GEN-96 | The library: Phase 41’s checks as pattern entries, applied by the patterns inspection finds | 15 entries in src/lib/library/entries.ts, each with its pattern, binding, technique, gate, where it runs, its builder and the bugs it caught. Evidence. |
✅ |
| GEN-97 | The template checks every link on the pages it selects | One read-only test per distinct link: on its page, visible and named, a valid href; a same-origin target loaded once a run. Evidence. |
✅ |
| GEN-98 | A listing’s repeated cards get one test, by position, not a test per card | A card group is 3 or more links on a page whose targets share an address but a last segment that is an identifier (Phase 40’s isIdSegment: #/product/1, /product/01M50R4WF...). Evidence. |
✅ |
| GEN-99 | A failing cards test names each failing card’s product, and every one of them | Every card is checked before the test fails once, with a line per failing card: its name, price and href. Evidence. |
✅ |
| GEN-100 | The sort check waits for the order it asked for, not for a quiet network | After choosing a sort option, the check reads the list until it shows that order, for up to 5 s, and only a list still out of order then fails. Evidence. | ✅ |
| GEN-101 | A page’s read-only checks, every area’s, run on one shared visit of the page | Each read-only test moves into its page’s Page /x group, inside its area’s describe. Evidence. |
✅ |
| GEN-102 | A shared page’s rules: loaded again after a check leaves it and after a failure, typed text cleared, a failure after a change re-run on a fresh page | A page is loaded again after a check leaves it and after a failure. Evidence. | ✅ |
| GEN-103 | A shared check’s assertions wait 2 s, not 5 s | Each page’s group shadows expect with expect.configure({ timeout: 2000 }): every check there only reads a page that has settled. Evidence. |
✅ |
| GEN-104 | A failing shared check keeps its evidence: Playwright’s trace and screenshot, and the shared visit’s video with where the check starts in it | The trace and screenshot as before, and now the shared visit’s video. Evidence. | ✅ |
| GEN-105 | A check that changes its page in place shares the visit too, after the page’s reads and the axe scan | A check that only changes its page in place (clicks, keys, DOM changes put back) shares its page’s visit. Evidence. | ✅ |
| GEN-106 | A link with no text is found by its href, whatever the href holds, and generated selectors and hosts are matched exactly | Textless links whose href holds a quote or backslash are found; DemoQA’s suites match only demoqa.com and hosts under it. Evidence. |
✅ |
| GEN-107 | Every helper the generator can write into a suite is reached by a fixed input of the template check | A helper is any export const *_HELPER(S) or export function build*Helper(s) under src/lib/generate, 13 today; since TFP-269 the comments before its declarations reach an output too. Evidence. |
✅ |
| GEN-108 | A journey whose path visits a page twice survives the generate route’s normalising, so selecting it writes its test | The path keeps its order and repeats, so a form posting back to its own page is no longer dropped; since TFP-271 one dropped for any reason is named under the suite, with why. Evidence. | ✅ |
| GEN-109 | The generated suite says Mr Onion: header, interactive switch and skip reasons, probe routes, test data, fixture names, probe attribute, window global (automated: the template check, mapped; the unit tests of each generator) | check:template-unchanged <before> --rename-map scripts/template-unchanged/rename-map.json: every fixed input identical. Without the map every input differs, and every changed line is exactly the map applied to the old one. The output holds no “testforge” in any case. Phase 48 Decision 3 and 11 (TFP-278). Measured 2026-11-09 against cf501ad: mapped, 172 of 172 identical; raw, 172 different, 307 changed lines, all 307 the map applied; “testforge” in the output 0 times. |
✅ |
| GEN-110 | The mapped check fails on any change that isn’t a name (manual: a mutant) | One non-name character changed in one helper makes the mapped check exit 1, naming the inputs. Measured 2026-11-09: the script payload’s +1 made +2: exit 1, 170 identical, 2 different, the line named. |
✅ |
| GEN-111 | A suite saved by 0.2.0 keeps running as it did (automated: src/lib/run/split.test.ts; manual: local Toolshop) |
A unit whose interactive test checks TESTFORGE_INTERACTIVE is Not run when the run leaves interactive tests off and runs when it turns them on; the UI offers the interactive switch for it. Phase 48 Decision 8 (TFP-278). Measured 2026-11-09 on a 0.2.0 folder’s saved suite (its one interactive unit naming the old setting, reused as it was): without --interactive, TFDB-001 Not run (interactive 1, active 3); with it, TFDB-001 ran (Failed: the contact form’s round trip), interactive 0, active 3. |
✅ |
| GEN-112 | The local model’s tag (automated: src/lib/generate/ollama.test.ts, src/app/api/generate/route.local.test.ts) |
mr-onion-qwen7b asked for: used when built; when only testforge-qwen7b is built, that one, with a “model tag” gate in the generation record naming it and ollama cp as the fix; never a fallback for another name. Phase 48 Decision 3 (TFP-278). |
✅ |