Ompractice Find Yoga/Meditation Class

Purpose

Given a user's interests (style, audience), availability window (date/time range and class length), and experience level, return a ranked list of upcoming live online yoga / meditation / movement classes on Ompractice that match. Each result includes class title, start time, duration, teacher, recommended experience level, intensity, tags, enrolment count, membership requirement, and the canonical join URL (https://app.ompractice.com/c/{slug}?classId={id}). Read-only — never books or pays. Booking requires an authenticated Ompractice membership and is a separate skill.

When to Use

• "Find me a gentle beginner meditation class this week before 9pm Pacific."
• "Suggest a 30-minute vinyasa flow tomorrow morning suitable for someone with some experience."
• "What chair-yoga or seated classes are on tap in the next 3 days for an arthritis-friendly intensity?"
• Any flow that needs to surface live, scheduled Ompractice class instances matched against user preferences. On-demand video recommendation is a separate concern — this skill is for the live class schedule only.

Workflow

The Ompractice public REST API (Django REST Framework on Heroku at api.ompractice.com) exposes the same data the schedule page consumes — no auth, no cookies, no anti-bot, no residential proxy required. Lead with the API. Drive the browser only if you also need to book (a different skill) or visualise filter state.

1. Fetch upcoming classes for a date window

GET https://api.ompractice.com/api/v1/classes/
    ?start_date=<ISO-8601 with Z>
    &end_date=<ISO-8601 with Z>
    &ended=false
    &format=json

• start_date / end_date are millisecond-precision ISO-8601 UTC timestamps (e.g. 2026-05-19T07:00:00.000Z). The official client uses the local midnight of "today" through "today + 7d" as default. The start_date / end_date pair is the only date-filter shape the backend honors — Django ORM-style start__gte=, generic from=, start_after= etc. are silently accepted and ignored (the query returns the full 22,438-record historical dump). See gotcha.
• ended=false excludes classes whose end is in the past — required, otherwise stale records show up.
• format=json is mandatory — without it, DRF content-negotiates to the browsable HTML API (text/html, ~5× larger and unparseable as JSON).

Response is a flat JSON array of class instances (not {count, results} like the unfiltered endpoint). Each item:

{
  "id": 31970,                            // canonical class instance ID — used in the URL
  "slug": "yoga-nidra-allison-jeraci-tuesday",
  "title": "Yoga Nidra",
  "short_desc": "...",
  "description": "...",                   // HTML allowed in body
  "type": 0,                              // 0=regular live class, 3=members-only / specialty
  "exp_lvl": "none",                      // "none" | "some" | "lots"
  "intensity": 1,                         // 0..20 numeric — see mapping below
  "duration": 30,                         // minutes
  "start": "2026-05-20T00:30:00Z",        // ISO UTC
  "end": "2026-05-20T01:00:00Z",
  "timezone": "America/New_York",         // teacher's tz; the start/end above are UTC
  "is_cancelled": false,
  "has_substitute": false,
  "requires_membership": true,
  "max_capacity": 100,
  "enrolled": 7,
  "teacher": { "id": 195, "slug": "allison-jeraci", "user": {"first_name":"Allison","last_name":"Jeraci",...}, "bio_short": "...", "photo_headshot": "https://...s3.amazonaws.com/...", ... },
  "tags": [ {"id":40, "name":"Meditation", "category":"MOVE", "display_to_public":true}, ... ],
  "recommended_props": "yoga mat, two blocks, blanket, bolster or pillow",
  "youtube_video": "",
  "main_photo": "...",
  "group_id": 4421
}

2. Filter client-side

The backend has no other filter params for the date-window endpoint — all other filtering happens client-side on the returned array.

3. Rank

A simple recommended scoring is fine — there's no popularity / rating signal in the response. Suggested ordering:

1. Filter to matches as above.
2. Sort ascending by start (soonest first).
3. Tie-break by ascending enrolled / max_capacity (less crowded preferred) or by intensity proximity to the requested level.

4. Emit the join URL

Canonical user-facing URL for a class:

https://app.ompractice.com/c/{slug}?classId={id}

Both {slug} (from the slug field) and ?classId={id} (from the id field) are required — visiting /c/{slug} alone may resolve to a different recurring instance.

5. (Optional) Enrich with public catalog data

If you also need the full set of tags or teacher bios:

• GET https://api.ompractice.com/api/v1/tags/?categories=AUDI,MOVE&format=json — 38 tags total across MOVE / AUDI / OUTC categories (the schedule UI only renders MOVE+AUDI, but OUTC tags exist e.g. "Digestive Health", "Veterans-Only").
• GET https://api.ompractice.com/api/v1/teachers/?format=json (paginated, ?limit=N&offset=M) — 56 teachers total with bios, photos, social handles. /teachers/all/ returns the same data without pagination.

Browser fallback

Only needed if the API is down or you have to demonstrate the user-visible filter UI. Stealth + proxy NOT required for app.ompractice.com (no anti-bot observed).

1. browse open https://app.ompractice.com/schedule --remote — server-side renders the next 7 days of classes as /c/{slug}?classId={id} anchors.
2. Click "Filter Classes" button (top of the schedule heading). A modal opens with sections: TEACHERS (dropdown), EXPERIENCE LEVEL (3 checkboxes), CLASS LENGTH (0–180 min dual-thumb slider), CLASS TYPE (~20 checkboxes including hierarchical Meditation→{iRest, Sound Bowls, Yoga Nidra} and Yoga→{Chair, Decompress & Rest, Gentle Flow, Power Yoga, Restorative, Yin}), STUDENTS / audience (~14 checkboxes), INTENSITY (Gentle / Moderate / Vigorous / Sweaty with sub-options).
3. Tick desired filters → click "Show Classes" at the bottom of the modal.
4. The filtered set re-renders in the main page; each class card has an href to /c/{slug}?classId={id} — harvest those, plus the visible title / teacher name / start time string per card.

Use the API path unless you're specifically asked to use the UI. The browser path costs ~25–30 actions and ~10 seconds of wait time for the same data the API returns in one HTTP request.

Site-Specific Gotchas

• No anti-bot. Plain curl / fetch works against api.ompractice.com from any IP — no --verified, no --proxies, no Cloudflare / Akamai. Verified with browse cloud fetch both with and without --proxies.
• ?format=json is mandatory. Without it the DRF API returns the browsable HTML viewer, which is ~5× larger than the JSON response and unparseable. Set Accept: application/json header instead if you prefer header negotiation.
• start_date / end_date are the ONLY date filter shape. Django ORM filters (start__gte, start__lte), generic params (from=, start_after=, date_after=), and the ordering=start param are silently ignored — the unfiltered /classes/ endpoint returns the full 22,438-record historical dump regardless. Use the date-window pair, or you'll be paginating through 5+ years of expired classes.
• ended=false is required. Without it, the date-window query includes classes whose end time has already passed today — useful for "what just ran" recap but not for upcoming-class discovery.
• The unfiltered /classes/ endpoint exposes a sentinel year 0206. When sorted, recurring-template records surface with start: "0206-03-22T17:26:02Z" — these are placeholder dates for recurrence sources, not real instances. The date-window endpoint excludes them automatically; if you ever query without dates, filter r['start'][:4] >= '2020' defensively.
• /api/v1/ root returns 401 with a WWW-Authenticate: JWT realm="api" header, suggesting the API is JWT-gated. It's not — sub-resources /classes/, /teachers/, /tags/ are unauthenticated. The 401 is just on the DRF root index view. Don't be misled into hunting for JWT credentials.
• tags=meditation filter shape → HTTP 500. Don't pass a tag name as the filter; the backend coerces to int and crashes. Filter client-side after fetching the date window.
• type is a sparse enum. Across an upcoming-week sample only type=0 (regular) and type=3 (special / members-only specialty class) appear; type=1 is reserved for one-off events e.g. "SPECIAL CLASS: Spring Equinox Community Gathering". Don't filter by type unless the user explicitly says "special events."
• exp_lvl: "lots" exists but is rare in upcoming weeks. A 14-day window observed only none (132) and some (50). Treat lots as a valid value to allow for, not to assume.
• intensity is numeric, not a label. The UI shows "Gentle / Moderate / Vigorous / Sweaty" but the data is 0..20. Mapping: Gentle→{0,1,2}, Moderate→{3,4}, Vigorous→{10}, Sweaty→{20}. Values are sparse — only 7 distinct numeric levels observed.
• requires_membership: true for 100% of upcoming live classes. This skill returns recommendations; booking still requires the user to subscribe. Surface that requirement in the agent's reply.
• Timezone confusion. start and end are UTC. timezone is the teacher's IANA zone, not the user's — useful for displaying "the teacher is in EDT" but don't subtract it to localize for the user. Always convert UTC start to the user's stated tz before showing.
• Class URL needs both slug AND classId. /c/{slug}?classId={id} — both required because the same class series (slug) recurs weekly with different classIds. Linking to /c/{slug} alone resolves to the next upcoming occurrence which may not be the one matched.
• The /teachers/all/ route returns unpaginated JSON; the default /teachers/ route returns DRF-paginated {count, next, previous, results}. Pick the right one based on whether you need the whole roster or just a page.
• No _next/data/<buildId>/schedule.json static route. The Next.js frontend (build id m6qOAtPhrnC_VqU5RVFrz as of 2026-05-19) doesn't expose a JSON server-component data URL for the schedule page — it server-renders the HTML with an embedded API fetch and hydrates from there. Don't bother probing _next/data/ for the JSON; go to api.ompractice.com directly.
• is_cancelled: true instances still appear in the date-window response. Filter them out client-side.
• has_substitute: true means the listed teacher will be replaced for that occurrence. Surface this to the user — the actual teacher in-class will differ from teacher.user.first_name + last_name.

Expected Output

{
  "success": true,
  "query": {
    "interests": ["Meditation"],
    "experience_level": "none",
    "intensity": ["Gentle"],
    "duration_min": 15,
    "duration_max": 45,
    "window_start": "2026-05-19T14:00:00Z",
    "window_end": "2026-05-26T14:00:00Z",
    "user_timezone": "America/Los_Angeles"
  },
  "total_matching": 5,
  "classes": [
    {
      "id": 32123,
      "title": "Silent Meditation",
      "teacher": "Beth Ciesco",
      "teacher_slug": "beth-ciesco",
      "start_utc": "2026-05-20T17:15:00Z",
      "start_local": "2026-05-20T10:15:00-07:00",
      "duration_minutes": 15,
      "experience_level": "none",
      "intensity": 1,
      "intensity_label": "Gentle",
      "tags": ["Meditation"],
      "tags_audience": [],
      "requires_membership": true,
      "is_cancelled": false,
      "has_substitute": false,
      "enrolled": 3,
      "max_capacity": 100,
      "recommended_props": "comfortable cushion or chair",
      "short_desc": "...",
      "url": "https://app.ompractice.com/c/silent-meditation-beth-ciesco-wednesday?classId=32123"
    }
    // ... up to N matches sorted by start ascending
  ],
  "membership_note": "All Ompractice live classes require an active membership to attend. This skill returns recommendations only and does not book or pay."
}

If nothing matches, return:

{
  "success": true,
  "query": {...},
  "total_matching": 0,
  "classes": [],
  "suggestion": "No classes matched. Try widening the date window, relaxing the intensity/level filter, or removing some interest tags."
}

If the API request fails (rare — no anti-bot, no auth):

{
  "success": false,
  "reason": "api_error",
  "status_code": 502,
  "message": "Upstream api.ompractice.com responded 502; retry in 30s."
}