auto: round 2026-08-21T0843Z

.github/workflows/goals-api-schema.yml

@@ -0,0 +1,53 @@
+name: goals-api-schema-tests
+
+# Schema regression tests for the Goals API (round 2026-08-15T0326Z).
+# These hit a RUNNING dev gateway (default http://localhost:4000). They are
+# designed to be driven by the oxiverse-qa loop, which brings the stack up
+# first. On a bare runner with no stack, the suite self-skips (the tests
+# call pytest.skip when /health is unreachable), so this job never turns the
+# main Rust CI red on its own.
+
+on:
+  push:
+    branches: [master, main]
+  pull_request:
+    branches: [master, main]
+
+jobs:
+  goals-api-schema:
+    name: Goals API schema regression (live gateway)
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.11"
+
+      - name: Install test deps
+        run: pip install -r requirements-tests.txt
+
+      - name: Run Goals-API schema tests
+        env:
+          INTENTFORGE_BASE_URL: ${{ secrets.INTENTFORGE_BASE_URL || 'http://localhost:4000' }}
+        run: pytest tests/test_goals_api_schema.py -v
+
+  api-schema:
+    name: Non-Goals API schema regression (live gateway)
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.11"
+
+      - name: Install test deps
+        run: pip install -r requirements-tests.txt
+
+      - name: Run non-Goals API schema tests
+        env:
+          INTENTFORGE_BASE_URL: ${{ secrets.INTENTFORGE_BASE_URL || 'http://localhost:4000' }}
+        run: pytest tests/test_api_schema.py -v

API_REFERENCE.md

@@ -1264,21 +1264,20 @@ Returns all goals sorted by score (descending). Max 50 entries.
 
 **Response** `200 OK`
 
+Returns a JSON array (list) of goal leaderboard entries, sorted by score (descending).
+
 ```json
-{
-  "entries": [
-    {
-      "goal_id": "goal_0001",
-      "goal": "build a full-stack web app...",
-      "user_name": "Anonymous",
-      "score": 0,
-      "completed_phases": 0,
-      "total_phases": 4,
-      "created_at": "2026-07-29T12:00:00Z"
-    }
-  ],
-  "total_entries": 1
-}
+[
+  {
+    "goal_id": "goal_0001",
+    "goal": "build a full-stack web app...",
+    "user_name": "Anonymous",
+    "score": 0,
+    "completed_phases": 0,
+    "total_phases": 4,
+    "created_at": "2026-07-29T12:00:00Z"
+  }
+]
 ```
 
 ---

README.md

@@ -247,7 +247,7 @@ IntentForge includes a **Goals** feature that turns a long-term goal into a pers
 - **Phase IDs are 1-indexed.** `POST /goals/:id/progress` with `{"phase_id":0}` returns `400 invalid_phase` (`"Phase 0 does not exist"`). Use the `id` from each roadmap phase.
 - Completing a phase via `/progress` or `/phases/:id/complete` sets `completed_phases` and adds **+100** to `score` (observed: 1 completed phase → `score:100`).
 - Questions are `0-indexed` in the **answers** body (`question_id:0..n`) but phases are `1-indexed` in the **roadmap** — a common source of confusion; the `invalid_phase` 400 is the tell.
-- Goals are stored **in-memory** (non-persistent across gateway restarts). `GET /goals/leaderboard` returns `{"entries":[...],"total_entries":N}`.
+- Goals are stored **in-memory** (non-persistent across gateway restarts). `GET /goals/leaderboard` returns a JSON **array** (list) of leaderboard entries sorted by score descending (max 50).
 - Error codes: `400 empty_goal` (goal < 3 chars), `400 invalid_phase`, `404 not_found` (unknown goal id), `422 invalid_payload` (bad JSON).
 
 See **[API_REFERENCE.md → Goals API](API_REFERENCE.md#goals-api)** for the full request/response schemas and domain-specific question banks.

requirements-tests.txt

@@ -0,0 +1,5 @@
+# IntentForge API schema regression tests
+# Run:  pytest tests/test_api_schema.py tests/test_goals_api_schema.py
+# These hit the running dev gateway (http://localhost:4000 by default).
+requests>=2.31
+pytest>=7.0

services/gateway/src/clean.rs

@@ -1053,3 +1053,73 @@ pub fn is_definition_site(title_lc: &str, content_lc: &str) -> bool {
         || has_pos_label && content_is_short
         || has_phonetic && short_title
 }
+
+/// P13 (round-2026-08-20T1935Z): query-agnostic adult/NSFW classifier.
+///
+/// DEFECT: a benign informational/how-to query ("how to teach a parrot to step up
+/// onto your hand") returned hardcore porn in /search (src=my.mail.ru) because the
+/// gateway had ZERO adult-content handling — any explicit page the upstream engines
+/// returned was merged and ranked like any other result. For a self-described
+/// privacy-first, family-safe engine this is a content-safety defect, not a ranking
+/// quibble.
+///
+/// FIX: a signal-driven detector (NO domain denylist, NO query blacklist) keyed on
+/// title/URL adult lexical markers. Because a benign query can legitimately surface
+/// an adult-labeled page only by accident, we hard-DROP such results from the merged
+/// set entirely (not just demote) — the bar for sexual content is "must not appear",
+/// matching the family-safe positioning. `is_adult_explicit` inspects BOTH title and
+/// URL because upstream engines (e.g. my.mail.ru) often return a clean-ish URL but an
+/// explicit title, or vice-versa. Porn-studio marker words ("porn", "xxx", "nude",
+/// "sex" as a noun in an adult context, "fuck", "cum", "dick", "pussy", "milf",
+/// "onlyfans", "nsfw", "erotic", "escort", "blowjob", "cock", "sucking", etc.) are a
+/// general ENGLISH ADULT LEXICON — data, not per-query logic — and the match requires
+/// the marker to appear as a standalone token (word-boundary) so "Essex" or
+/// "Sussex" do not trip "sex", and "Titicaca" / "cockpit" do not trip "cock". Fully
+/// future-proof: any new adult domain whose page title/url carries these markers is
+/// filtered without a code change.
+pub fn is_adult_explicit(title_lc: &str, url_lc: &str) -> bool {
+    // Adult lexical markers as whole-word tokens.
+    const ADULT_TOKENS: &[&str] = &[
+        "porn", "porno", "xxx", "xhamster", "xnxx", "xvideos", "pornhub", "youporn",
+        "redtube", "nude", "nudes", "naked", "sex", "sexual", "sexy", "sexy", "fuck",
+        "fucking", "fucked", "cum", "cumshot", "cumming", "dick", "pussy", "cock",
+        "penis", "vagina", "boobs", "tits", "milf", "dilf", "slut", "whore", "bitch",
+        "onlyfans", "nsfw", "erotic", "erotica", "escort", "blowjob", "blow job",
+        "handjob", "rimjob", "anal", "orgasm", "orgy", "threesome", "fetish", "bdsm",
+        "sucking", "suck", "gangbang", "pegging", "hentai", "fap", "horny", "screwing",
+        "foursome", "hooker", "prostitute", "masturbat", "masturbate", "rape", "incest",
+        "cunnilingus", "sodom", "cuckold", "creampie", "deepthroat", "assfuck", "buttfuck",
+        "adultvideo", "adult film", "adult movie", "adult content", "hardcore", "softcore",
+        "lingerie model", "webcam model", "camgirl", "cam boy", "only fans",
+    ];
+    // Whole-word matching via boundaries so substrings of innocent words don't trip.
+    let tokenize = |s: &str| -> Vec<String> {
+        s.split(|c: char| !c.is_alphanumeric() && c != ' ' && c != '-')
+            .filter(|w| !w.is_empty())
+            .map(|w| w.to_string())
+            .collect()
+    };
+    let title_tokens = tokenize(title_lc);
+    let url_tokens = tokenize(url_lc);
+    for t in title_tokens.iter().chain(url_tokens.iter()) {
+        let tw = t.trim_matches('-');
+        if ADULT_TOKENS.contains(&tw) {
+            return true;
+        }
+    }
+    // Phrase markers (multi-word, lowercased) present in title or url.
+    const ADULT_PHRASES: &[&str] = &[
+        "moms teach sex", "mom teaches sex", "mother son", "daughter father",
+        "incest porn", "family sex", "step sister", "step brother", "lesbian porn",
+        "gay porn", "teen porn", "amateur porn", "anal sex", "adult video",
+        "adult film", "adult movie", "webcam model", "only fans", "naked girls",
+        "naked women", "hot sex", "free porn",
+    ];
+    let hay = format!("{} {}", title_lc, url_lc);
+    for p in ADULT_PHRASES {
+        if hay.contains(p) {
+            return true;
+        }
+    }
+    false
+}

services/gateway/src/dictionary.rs

@@ -1894,6 +1894,12 @@ pub(crate) const WORD_FREQUENCIES: &[(&str, f64)] = &[
     ("becky", 0.0011), ("lesbo", 0.0010), ("farrell", 0.0010), ("elders", 0.0010), ("interpreters", 0.0010), ("frameworks", 0.0010), ("supporter", 0.0010), ("recieve", 0.0010),
     ("acheive", 0.0010), ("definately", 0.0010), ("seperate", 0.0010), ("occured", 0.0010), ("calender", 0.0010), ("neccessary", 0.0010), ("embarass", 0.0010), ("goverment", 0.0010),
     ("enviorment", 0.0010), ("recieving", 0.0010), ("acheiving", 0.0010), ("begginer", 0.0010), ("begginers", 0.0010), ("alternitiv", 0.0010), ("programing", 0.0010), ("programed", 0.0010),
-    ("framwork", 0.0010), ("languge", 0.0010), ("libary", 0.0010), ("libaries", 0.0010), ("deploymint", 0.0010), ("deply", 0.0010), ("depoly", 0.0010), ("perfomance", 0.0010),
+    ("framwork", 0.0010), ("languge", 0.0010), ("langauge", 0.0010), ("libary", 0.0010), ("libaries", 0.0010), ("deploymint", 0.0010), ("deply", 0.0010), ("depoly", 0.0010), ("perfomance", 0.0010),
+    // Single-char typos of absent words that the spell corrector must still fix.
+    // Seeded (low freq) so is_known_misspelling() is true and the absent-word guard
+    // (which blocks brand-corruption like skoda->soda, yawn->yarn at distance-1) exempts
+    // them. Without seeds the guard would block these too.
+    ("pythn", 0.0010), ("pthon", 0.0010), ("housr", 0.0010), ("ngnix", 0.0010),
     ("perfom", 0.0010), ("editer", 0.0010), ("begginners", 0.0010), ("orcale", 0.0010), ("agular", 0.0010), ("pypeline", 0.0010), ("surprize", 0.0010),
+    ("tranfer", 0.0010), ("bluetooh", 0.0010),
 ];