Ojava Coach (the AI assistant)

How it works

Ojava Coach reads your accepted records (with consent) to give context-aware, educational answers about your own data. It is a feature that helps you understand your health, not the product’s identity, and not a chatbot you should treat as a clinician.

It stays focused on your health and wellness, and that focus is the point. Because it can see your own labs, wearables, sleep, recovery, training, nutrition, and medications over time, it connects the dots across your data in a way a generic chatbot you paste a single file into cannot. If you ask it something unrelated (writing code, trivia, legal or financial questions), it will gently point you back to how it can help with your health instead.

Capability groups

The Coach opens as a document rather than a chat window. A bar across the top names it and holds its actions: New conversation, the conversations you have saved this session, and a copy of the one on screen once there is something to copy. Below that the page stays quiet, so the greeting, the starter prompts, the composer, and your conversation get the attention, and any conversations you saved earlier are listed on the page instead of hidden behind a menu. A column on the right says what the Coach has to work from: how many results you have accepted, how many files you have imported, and how many are still waiting for your review. Deeper records, labs, messages, reports, trackers, and recommendations live in the navigation, in their own focused areas instead of competing with the conversation.

Records work is part of the same capability map. PDF upload, provider import, reports, and exports route through the records surface, while Coach command prompts use only accepted data and independent-care boundaries when they prepare a packet.

The starter chips use that same map. Choose a Coach capability such as labs, symptoms, care gaps, hidden insights, exercise, diet, or longevity to stage the matching prompt in the composer. It never sends a prompt until you review and submit it. Nutrition Coach uses the same Coach handoff and patient snapshot as the food diary review panel: reviewed food rows, repeated accepted diary foods, nutrition goals, workouts, body context, saved foods, planned meals, labs, symptoms, and medications only appear when those deterministic values are present. If you ask for broader Nutrition Coach food ideas, Coach can receive request-scoped public food-label candidates from the public food database as idea candidates only. Your own foods, meals, recipes, and plans stay primary; public label rows can be incomplete or stale, must show source and missing fields, and require review before saving or logging. They are not medical nutrition therapy, a diagnosis, a treatment plan, a calorie prescription, or a safety guarantee.

Ojava Coach is tuned for structured chat output. When an answer has several comparable pieces, such as labs, trends, medication context, care gaps, tradeoffs, recommendations, or next steps, it leads with a short takeaway, then uses a compact table, and closes with one next step. Short one or two item answers can stay as bullets so the chat does not feel forced. When it cites your data, it keeps the value, unit, date, trend, and plain read close together so you can inspect the evidence quickly. Longer answers use short section headings so the takeaway, record context, and next step stay easy to scan. For symptoms or plans that need more context, it asks only the few missing facts that would change the next step instead of turning the answer into a long intake form. Even a brief greeting or check-in now stays useful: it answers warmly, surfaces one record-grounded next step or one missing record to add, and ends with one health-focused question. With import consent, the same lead context can name broad missing or older record categories to discuss with a clinician, while keeping them as visit-prep questions rather than lab orders.

For care-plan follow-up or clinician-prep questions, Coach treats missing information as local checklist support. It can name the accepted record behind a gap, the detail you may want to add, and a user-reviewed question to bring to the visit. It does not send messages, contact clinicians, change clinician instructions, route care, monitor health, diagnose, prescribe, order labs, create a treatment plan, or prove that a clinician reviewed the packet.

For medication-burden, polypharmacy, anticholinergic-review, deprescribing, Beers Criteria, STOPP/START, or pharmacist-review questions, Coach can also build a source-backed prep packet from current medication and supplement rows. It names current item counts, missing dose, schedule, reason, and reviewer details, possible duplicate-name groups, and clinician or pharmacist questions, with visible source links. It does not score anticholinergic burden, classify a medication as safe or unsafe, apply Beers or STOPP/START to you, predict falls or cognition risk, or suggest medication changes.

If your accepted records include an after-visit summary or discharge instructions, Coach can also organize a post-visit readiness packet. It can name follow-up tasks, medication-change mentions, pending tests or referrals, warning-language excerpts, source evidence, and missing details to review with your clinician. That packet is not medical advice, triage, emergency monitoring, a message to a clinician, a referral route, a lab order, or a change to clinician instructions.

For between-visit or follow-up questions, Coach can also place accepted clinical-history anchors beside your own symptom, mood, food, medication or supplement routine, and wearable tracker days. It names which tracker lanes are present, which source details are missing, and which questions may be useful to bring to a clinician. It does not diagnose, infer causation, predict flares, prove treatment response, prove adherence, monitor you clinically, order labs, route referrals, or prove clinician review.

When your accepted FHIR imports include document references, Coach can also organize the document metadata it can safely see: title, type, category, date, source trail, and whether an attachment pointer exists. That is a review checklist, not document interpretation. Coach does not read embedded attachments, interpret images or scans, send records, decide coverage, or prove that a clinician reviewed the document.

Accepted FHIR specimen and provenance rows can also become a source-lineage review packet. Coach can organize collection and received dates, collection method, body-site labels, provenance target counts, agent labels, entity counts, and source trails. It does not judge sample adequacy, validate lab quality, prove authenticity, prove consent, prove chain of custody, diagnose, order labs, or complete a care action.

For an outside portal message or visit note, Coach can assemble accepted records, wearable context, and active medications into a local draft when clinical-context consent is active. You review and copy it yourself. It is wording and context support only: no message is sent, no provider inbox is created, no clinician response is promised, and Coach does not diagnose, prescribe, order labs, or route emergencies.

For Saved Coach Threads handoff questions, Coach can also summarize saved message threads into a local review packet: user-authored topics, AI reply counts, safety replies, unresolved user turns, and questions to verify with a clinician. It does not send the packet, create a provider inbox, promise a clinician response, triage emergencies, or prove a human clinician reviewed the thread.

The chat also preserves your place while you read. If you scroll up to compare an older answer, a live reply will not drag you back to the bottom. When you are already at the newest turn, Coach can follow the streamed answer as it arrives. If you step away from the live edge, a jump-to-latest control lets you return when you are ready, and reopened threads restore near the last useful turn instead of blindly snapping to the bottom.

For integration questions, Coach now separates source setup from source-account activation. Source setup can name the current import path for a wearable, app, CSV export, lab file, imaging report, or provider-record path. Source-acquisition planning can also rank the next source to add from current gaps and approved provider states. For provider-record questions, that planner reads the same server-side record-network activation rails as the connection route, so Coach can say whether the selected aggregator credentials, webhook ingest, service role, and identity-proofing rails are present without exposing secret values. Source-account activation can name the rails that must exist before Oura, WHOOP, Eight Sleep, Apple Health, Health Connect, or Terra move beyond planning: provider contracts or native entitlements, user consent scope, privacy and security review, source-quality checks, and webhook or import audit trails. Terra sits differently from the rest of that list. It is a possible future paid cloud adapter, not a current path or a live dependency, so Coach must not present it as something waiting on the same activation queue as the native health hubs. It still cannot open OAuth, store source credentials, claim live sync, write back to sources, control devices, monitor in real time, send alerts, diagnose, treat, or make care decisions.

For CGM and glucose-source questions, Coach also receives a request-scoped source-path packet. It keeps the current imported glucose rows, missing timestamped meals, file-ready export paths, and gated native or partner connector paths visible even when other high-frequency wearable rows dominate the recent wearable window.

For home glucose log questions, Coach can also receive a citation-backed readiness packet from source-provided glucose rows. It counts reading days, morning and later-day timestamp coverage, source labels, missing clinician-provided testing schedule or personal range context, and questions to bring to a clinician. It does not diagnose diabetes or prediabetes, grade glucose control, define personal ranges, recommend insulin or medication adjustments, watch for emergencies, or decide care.

For source-quality or records questions, Coach may also see aggregate source plausibility counts: checked rows, unchecked rows, review issues, and unit-mismatch rows. Those lines help Coach say what needs manual source review without exposing raw filenames, hashes, source ids, exact flagged values, or clinical-validation claims.

For fitness source questions, Coach can also see a source reconciliation packet across recent workout rows, reviewed MyFitnessPal exercise imports, and wearable activity rows. The packet chooses one source for comparable active-minute and step totals, keeps reviewed workout calories separate from wearable active calories, and names same-day overlaps so Coach does not add the same movement twice. It cannot certify source accuracy, prove live sync, clear workouts, create a training plan, monitor activity clinically, or prescribe calories.

For personalization questions, Coach can also use a life-fit context packet from visible notes you asked Ojava to remember, active goals, and saved nutrition preferences. Sensitive notes stay withheld, and the packet is only for fitting General-Wellness habits or clinician-prep questions around your schedule, constraints, preferences, and support style. It does not infer hidden sensitive details, diagnose, prescribe, create a treatment plan, or turn preferences into medical nutrition therapy.

What it sees: your computed numbers, not guesses

Ojava Coach reads the same computed wellness estimates your app surfaces, rather than re-deriving them from raw rows: a whole-record trend recap that compares latest versus earlier accepted labs, wearable rows, food logs, symptoms, mood check-ins, goals, and reviewed record activity when enough history exists, condition-aware tracker guidance for chronic-condition questions, flare recovery-window context that compares onset, peak day, latest symptom severity, post-peak energy, stress, sleep, and relief notes, your readiness and recovery score, fitness age, training load, sleep debt, duration consistency, timing regularity, Daily Sleep Review context with focus areas, strengths, missing sleep inputs, and the next review prompt, Sleep routine readiness context when source-provided or saved routine inputs exist, annual sleep recap context with tracked nights, average sleep, total source-backed sleep, longest tracked night, short-night count, available source sleep score, HRV, and resting-heart-rate context, annual activity recap context with logged workouts, active days, total logged active time and distance, longest active streak, busiest tracked month, most logged activity, and total logged strength volume, source-backed sleep environment context when exports include room, bed, skin, breathing, oxygen, snoring, or breathing-variation rows, Sleep blueprint readiness when age, profile sex, sleep-stage, REM, preferred sleep-temperature, and room or bed-temperature inputs are present, request-scoped sleep-device automation readiness for smart-bed, bed-temperature, temperature-schedule, smart-alarm, or sleep-automation questions, request-scoped sleep-system attribution readiness for shared-bed, bed-partner, side-of-bed, dual-zone, or whose-sleep-row questions, daylight and sound exposure context when exports include time in daylight, environmental audio, or headphone audio rows, activity streak, resting heart rate, HRV, and temperature trends against your own baseline, source-provided running power, cycling power, and cycling FTP when exports include them, source-provided recovery-time context when exports include it, source-provided arterial stiffness and cardiovascular-age context when exports include PWV or source age rows, strength volume and estimated 1RM from logged sets in Timeline, your recent workout bests from loaded workout and strength history, such as longest session, longest distance, fastest logged pace, recent strength estimate, account-backed user-authored workout routine templates when present, request-scoped workout-routine follow-through from recent workout and strength rows, user-authored planned-workout calendar review that compares planned sessions with completed workout or strength rows, repeated workout patterns that can become reviewable routine starters, and saved exercise-goal progress for numeric weekly targets, medication and supplement routine-completeness summary, reminder/refill prep, source-backed medication-burden review prep for polypharmacy, anticholinergic-review, deprescribing, Beers Criteria, STOPP/START, or pharmacist-review questions, self-reported dose history from Profile, GLP-1 dose-log fidelity for latest status, timing, injection or location notes, mg amount, next schedule check, and missing context when a current GLP-1 medication is present, diary-native GLP-1 context that lines routine days up with food entries, calories, protein grams, water, appetite or digestive notes, body trend or same-day weight rows, and workout entries, and GLP-1 support plus weight-care prep context, your week-in-review digest, a weekly healthspan habit report that summarizes biomarker, sleep, movement, nutrition, recovery, and self-tracked habit coverage without requiring an active goal, request-scoped body-clock readiness that organizes source sleep timing, daylight exposure, food timing, later food or drink clues, fasting windows, and recovery context when you ask about circadian rhythm or eating-window timing, user-logged fasting windows with target-reference and active-timer context, fasting readiness recap context that combines fasting rhythm, food-window consistency, reviewed water rows, same-day food timing, and recovery or body-clock evidence for fasting, eating-window, hydration, body-clock, or nutrition-rhythm questions, habit signals from your own logs (how workouts, fasting, and wellness sessions line up with your sleep and recovery, associations only, never causation), mental-recovery patterns from your own mood, stress, mindfulness, and breathwork logs when they have enough matching recovery days, weekly mental-wellness recap context over recent mood, sleep, mindfulness-style practice, and recovery metric coverage, request-scoped cognitive check-in readiness for brain fog, focus, memory, attention, or mental-clarity questions using recent symptom notes beside sleep, mood, practice, and recovery rows without cognitive testing or screening claims, meditation-practice streak context with recent practice days, last-practice gap, consistency cue, and missing inputs, structured mental-wellness tool readiness that selects from Ojava’s own breathing, grounding, journaling, sleep, and support-plan tools without inventing a therapy plan, a mental companion readiness packet that separates reflective chat and tool handoff from voice companion entry, therapy services, risk scoring, diagnosis, and crisis counseling, a request-scoped wellbeing navigator packet that separates local wellness tools, record-grounded reflection, crisis-resource signposts, outside-provider conversation prep, and therapy cost or benefits questions when you ask where to start, including a sleep wind-down readiness packet when mood or stress logs, sleep rows, and mindfulness or holistic practice rows are all present, a stress and energy status packet from recent mood check-ins plus sleep, HRV, resting heart rate, and source-provided stress or energy rows when present, a deterministic recovery habit ranking that chooses one to three next wellness actions from finished readiness, sleep debt, training load, HRV, mood and stress, hydration, and goal context, with missing inputs named instead of guessed, cycle and recovery context that compares logged cycle or hormone-context timing with recent wearable recovery, sleep, HRV, symptoms, energy, and stress when enough rows are present, symptom-factor correlation context that compares repeated food, mood, workout, fasting, wellness, routine, cycle, or wearable factor days with symptom entry count and severity after minimum day guards, metabolic source readiness for imported glucose rows plus timestamped meals, post-meal glucose context when timestamped food logs line up with source-provided glucose readings, multi-day nutrition app versus reviewed diary agreement when overlapping source and diary days exist, local food-name writeback readiness from reviewed diary rows with external partner writeback off, Nutrition Progress Overview wording that compares logged-day calories, protein, carbs, and fat across the current 7 days and prior 7 days, private nutrition challenge readiness for food-logging streaks, protein anchors, water consistency, fasting rhythm, movement support, and meal-plan follow-through, Missed-day recovery plan context that detects recent missed tracking days across food, water, workouts, and fasting before naming low-pressure reset actions, and hydration digest readiness from reviewed Water log rows with recent row count, total and average mL per logged day, target or reference context, same-day status, reminder context, and missing inputs, account-backed custom foods, saved meals, recipes, favorite food shortcuts, user-owned recipe-discovery starters, and nutrition library maintenance review items for missing serving, macro, label-source, freshness, or duplicate-conflict detail as bounded reusable-template context, meal-plan prep checklists from your own planned-meal prep notes, account-backed food reminder checkpoints as diary habit context with device-permissioned local reminder delivery kept outside medical alerts or adherence claims, saved per-meal calorie-goal progress when a logged meal has reviewed calories, your saved weekday-specific nutrition goal schedule with today’s target, target-date estimate when saved, and upcoming override days, the saved 0 to 100 percent exercise-calorie add-back preference and same-day wearable active calories that control same-day energy context, nutrient-depth completeness for recent logged-food label detail, with saved nutrient targets when present, reviewed MyFitnessPal progress-import measurements such as neck, chest, hips, thigh, arm, calf, forearm, and shoulders when the CSV includes them, a consolidated body progress report over accepted measurements, reviewed progress imports, compact trend lines, and bounded body progress photo metadata such as check-in counts, recent dates, and safe short captions without image bytes, paths, links, image interpretation, body judgment, or appearance scoring, your biological-age estimate, allostatic-load context with source links and variability caveats when enough reviewed systems are present, pre-lab context windows around accepted lab dates, lipid draw fasting or nonfasting context for accepted lipid rows plus nearby meal and fasting logs, lab-visit prep readiness for blood-draw or testing questions from upcoming appointment context, lab-like sources, accepted and pending lab rows, and medication or supplement names, how your markers sit against general typical ranges, and a longevity evidence-readiness summary that combines biological-age marker completeness, recovery and sleep coverage, body-trend context, and appearance prep coverage, plus aging-clock result prep for biological-age, epigenetic-clock, pace-of-aging, and source-provided organ-system age questions when the record has real evidence. So its guidance matches what you see on screen, and when a number is missing it says so plainly instead of inventing one.

The whole-record trend recap is a factual packet, not an interpretation layer. It can show the latest value, earlier value, date, direction, source count, and missing-history note for each available signal. It cannot infer causation, diagnose, prescribe, order labs, validate whether a lab was collected correctly, create alerts, or treat a trend as a medical decision.

The weekly healthspan habit report is also deterministic. It summarizes source coverage across accepted biomarkers, sleep, movement, nutrition, recovery, and self-tracked habits, then names one next habit and missing inputs. It does not diagnose, prescribe, order labs, score disease risk, monitor emergencies, create medical nutrition therapy, or prescribe a training plan.

For food-logging education questions, Coach can also receive the same cited guide that appears in Nutrition docs. It names the reviewed logging days, input modes, serving context, nutrition label context, source references, and review-only boundary before answering. It does not turn a diary into a prescribed diet, medical nutrition therapy, diagnosis, treatment plan, or instruction about what to eat for a condition.

For macro target questions, Coach can receive a deterministic adaptive review when there are enough recent food logs and body-weight entries. The packet names the logged-day count, food-entry count, body-weight count, calorie gap, protein-reference percent, and weekly weight trend before drafting any small target change. The draft is review-only and is never saved by Coach. It does not appear for thin data, protein-only focus, medical nutrition therapy, calorie prescription, or weight-management treatment.

Nutrient-depth completeness is also deterministic. Coach can name how many recent food rows have reviewed label nutrient detail even before you save nutrient targets. When saved nutrient targets exist, it can also name which supported nutrients have enough row and day coverage to discuss, and which saved nutrient targets need more logged detail. It receives that as bounded answer context only, so micronutrient questions stay grounded without crowding out broader record context. Coach can also see a same-day nutrient evidence quality line for saved targets, separating supported nutrients from estimated source-mode rows and missing target detail. If no saved nutrient targets exist, Coach gets only the compact summary. It cannot diagnose a deficiency, recommend supplements, certify nutrient accuracy, score foods, prescribe a diet, create medical nutrition therapy, diagnose, or treat.

When a user comes from the public food-log quality checker, Coach treats the handoff as a prep packet rather than a private diary import. The packet can name the checker result, cited food-logging references, entry methods reviewed, serving or label context, missing inputs, and the user’s question. Coach can then explain how to tighten the log or what to ask next, while keeping the answer in general wellness education and visit prep.

For acute digestive or foodborne-illness prep questions, Coach can receive a request-scoped acute GI nutrition packet from reviewed food rows, acute digestive symptom entries, water logs, medication names, and supplement names. The packet names ready inputs, missing context, timing questions, disabled actions, and the safety-preflight prompt to use if symptoms sound severe or worsening. It is clinician-prep context only: no food-poisoning diagnosis, food-causation proof, dehydration assessment, fluid or electrolyte prescription, antibiotic or anti-diarrheal recommendation, stool test or lab ordering, outbreak detection, public-health reporting, urgent-care booking, emergency triage, or replacement for medical care.

For chronic-condition tracking questions, Coach can use deterministic quick-answer and tracker packets instead of inventing a plan. If your current question or symptom notes mention a supported chronic pattern, it maps your own symptoms, mood or energy check-ins, food timing, routine timing, sleep, wearable context, and accepted record rows into a prep-only quick answer, clinician questions, record gaps, next data to collect, and tracker lanes such as standing tolerance, flare notes, food timing, relief notes, recovery window, function impact, or medication timing. It names which lanes are ready and which inputs are missing. It does not diagnose, predict flares, monitor you clinically, adjust medication, create a treatment plan, or replace clinician guidance. When you ask for a chronic illness report, flare worksheet, or visit brief, Coach can also receive the same Reports readiness rollup: ready section count, missing inputs, symptom and trigger summary, clinician questions, and the report boundary before it answers.

When you have logged symptoms across enough months and seasons, Coach can also surface symptom seasonality context: which symptoms lean toward a particular season, the months they show up most, and how much of their history falls in that season. This is pattern observation to help you prepare for predictable seasonal patterns with a clinician. It is never an allergy diagnosis, a weather or environmental-trigger cause, a flare prediction, or treatment advice.

Cycle and recovery context stays in the General-Wellness lane. Coach can use it to say that recovery signals look easier, strained, or mixed around a logged cycle or hormone-context window, and to name what to track or ask about. It does not predict fertility or periods, diagnose a hormone condition, stage menopause, give pregnancy or contraception guidance, recommend hormone therapy, clear workouts, monitor you clinically, or create a treatment plan.

For menopause, perimenopause, postmenopause, hot flashes, night sweats, midlife sleep changes, genitourinary symptoms, FSH, estradiol, or hormone-therapy questions, Coach can receive a menopause and midlife-health prep packet. It organizes accepted cycle timing, symptom context, hormone labs, thyroid context, cardiometabolic context, bone-health context, medication and supplement context, body context, wearable sleep or recovery context, missing inputs, clinician questions, and linked STRAW+10, Menopause Society 2022, ACOG, and Endocrine Society sources. It does not diagnose menopause or perimenopause, stage reproductive aging, interpret hormone labs, recommend hormone therapy, recommend nonhormonal treatment, clear hormone-therapy safety, calculate breast, cardiovascular, clot, dementia, osteoporosis, or cancer risk, order labs, refer to a specialist, or claim clinician review.

For reproductive, fertility, pregnancy, or postpartum prep questions, Coach can also read a deterministic prep packet from the same consent-gated record path: cycle logs, symptoms, goals, procedures, appointments, medications, supplements, allergies, accepted record rows, and related lab or vital context when present. It names ready sections, missing context, and questions for a licensed-care conversation. It does not predict fertility, manage pregnancy, prescribe contraception, give obstetric advice, order tests, interpret fetal or pregnancy risk, diagnose, treat, or replace a clinician.

For physical function, functional reserve, frailty, sarcopenia, grip strength, gait speed, chair stand, SPPB, SARC-F, stair climbing, balance, mobility, rehab, or physical therapy questions, Coach can receive a physical-function reserve prep packet. It organizes accepted grip-strength, gait-speed, chair-stand, balance, SPPB-style, SARC-F-style, wearable mobility, symptom, body, rehab, and goal context with linked SPPB, PURE grip-strength, SARC-F, and EWGSOP2 sources. It does not diagnose frailty or sarcopenia, calculate SPPB or SARC-F, calculate disability, fall, mortality, or sarcopenia risk, predict disability, clear activity, prescribe exercise, interpret a formal physical-performance test, order tests, refer to a specialist, or claim clinician review.

Family and pediatric prep uses that same deterministic path. When your accepted records or consented tracker rows include family, caregiver, pediatric, or dependent context, Coach can organize a structured packet across immunizations, allergies, medications and supplements, symptoms and visit questions, growth, development, school, feeding, sleep, documents, prior history, and follow-up items. It names which sections are present and which details are missing before you talk with a clinician. It does not give pediatric medical advice, act for a dependent, route urgent care, prescribe, order labs, diagnose, or change clinician instructions.

If the question is specifically about family benefits, Coach can compose those family and reproductive packets with payer prep into a single family-benefits readiness packet. It separates fertility and family-building, maternity and postpartum, newborn and pediatric, caregiver and dependent, EAP or behavioral-health support, and plan-call readiness. The packet names evidence, missing plan details, consent scopes, audit placeholders, and questions to ask before the user calls a plan, employer, benefits administrator, pharmacy, clinician office, or pediatric office. It does not decide coverage, submit claims, submit prior authorization, choose a provider, book care, act for a dependent, guarantee benefits, diagnose, prescribe, or replace a plan, employer, clinician, or benefits administrator.

The longevity readiness summary is a coverage read only. Coach can use it to say which evidence is present, which input would help next, and why a biological-age number is or is not available. It can also read the deterministic biological-age marker-set packet to name which reviewed lab markers are present, which requirement set is closest, and which markers are still missing. It cannot turn that coverage score or a source-provided organ-system age result into a diagnosis, organ risk score, clinical risk score, treatment plan, lab order, prescription, facial-age read, biological-age guarantee, lifespan estimate, questionnaire-derived biological age, external-clock or organ-model validation, diagnosis, treatment, or prescribing.

That metabolic source readiness and post-meal glucose context is a pattern summary only: imported glucose rows, timestamped meals, missing source inputs, matched meals, post-meal peaks, and pre-meal deltas when the source data supports them. Ojava does not connect directly to glucose devices or turn CGM or glucose rows into a diagnosis, a glucose score, alerts, diabetes management, insulin or medication guidance, medical nutrition therapy, or a treatment plan.

The CGM source-path packet reuses that same glucose summary and post-meal matcher, then adds source setup state from the wearable provider catalog. It can say whether the best next path is reviewing the current glucose import, adding timestamped meals, uploading a glucose export, or waiting on native or partner activation.

When the source-backed meal and glucose rows support it, Coach can also read a metabolic meal experiment comparison. That packet compares repeated meals or large macro contrasts from the finished pre-meal and post-meal glucose context, then names the source count, matched meals, peak deltas, macro context, and missing inputs. It is general wellness pattern context only, not glucose control, diabetes management, alerts, medication advice, diagnosis, treatment, a prescribed diet, or medical nutrition therapy.

For self-experiment, baseline, follow-up, protocol, glucose-response, training-recovery, sleep-experiment, or supplement-routine questions, Coach can also read a self-experiment readiness packet. It organizes the hypothesis frame, baseline window, follow-up window, metrics, source-backed evidence, missing inputs, and review questions for a user-owned tracking plan. It is ready-to-review pattern context only. It cannot say the experiment worked, claim causation, recommend an intervention, change medication or supplements, prescribe diet or dosing, run a clinical trial, diagnose, treat, order labs, or monitor in real time.

If those source-backed meal-glucose signals line up with nutrition-recovery associations and active nutrition, fitness, body, recovery, or longevity goals, Coach can also see a reactive metabolic coaching readiness line. It summarizes present signals, missing inputs, the next review cadence, and a first review focus before Coach explains what to inspect next. It is not real-time monitoring, clinician messaging, glucose control, medication advice, a prescribed diet, diabetes management, alerts, diagnosis, treatment, or medical nutrition therapy.

For cardiometabolic visit-prep questions, Coach can also read a bounded prep packet that combines accepted glucose or A1c, lipid or ApoB context, blood pressure, body, food, activity, sleep, and recovery rows. It names ready and missing domains for a focused clinician conversation. It does not diagnose, calculate cardiovascular or diabetes risk, order labs, guide medications, create alerts, treat, or provide medical nutrition therapy.

Daily Sleep Review is a synthesis packet over the same finished wearable numbers. Coach can summarize sleep score or debt, efficiency, latency, timing regularity, stages, environment rows, readiness, and HRV trend into focus areas, strengths, missing inputs, and one next review prompt. It cannot diagnose sleep conditions, replace a sleep study, create alerts, control devices, prescribe care, or change a treatment plan.

Sleep routine readiness for Coach uses the same deterministic path. When source data or saved settings support it, Coach can receive bedtime target, wake target, alarm time, sleep goal, timing regularity, sleep debt, latency, efficiency, sleep score, and environmental signals as finished routine context with missing inputs named. It cannot set alarms, schedule notifications, wake you, automate temperature, control a bed, recommend exact bed or wake times, diagnose insomnia or circadian disorders, replace a sleep study, send alerts, prescribe, or change treatment.

For explicit sleep routine follow-through questions, Coach can compare source routine targets with recent source sleep rows before answering. The packet names matched targets, drifting targets, missing clock-level sleep rows, disabled actions, and the review boundary. It does not score adherence, schedule alarms, control devices, diagnose sleep problems, or prescribe a sleep plan.

Coach also receives stress resilience context when the record has enough stress-load and recovery inputs. The packet combines recent stress and energy check-ins, source stress or energy rows, readiness, HRV trend, sleep debt, and training load into a transparent Ojava level and score. It cannot reproduce a proprietary wearable formula, diagnose stress, clear activity, monitor you clinically, provide therapy, or create a treatment plan.

Coach also receives recovery signal consistency when the record has enough finished recovery summaries. That packet compares readiness, HRV, sleep debt, training load, resting-heart-rate recovery, temperature, strain-recovery balance, and stress-energy context before Coach turns them into prose, so supportive and strained signals stay visible together. It is source-review context, not diagnosis, real-time monitoring, alerts, activity clearance, recovery prescribing, or care decisions.

Annual sleep recap context is also deterministic. Coach can summarize the current year from accepted sleep rows, name missing score, HRV, resting-heart-rate, or month-comparison inputs, and provide user-reviewed share text. It cannot post publicly, control a sleep device, automate temperature, diagnose sleep quality, replace a sleep study, send alerts, recommend treatment, or turn missing data into a health conclusion.

Sleep environment context is also source-only. Coach can say which rows are present or missing, such as room temperature, bed temperature, skin temperature, respiratory rate, oxygen saturation, snoring minutes, or breathing variation. It cannot infer apnea, illness, room-control needs, device automation, alerts, or care changes from those rows.

Sleep blueprint readiness is an input-readiness line, not a control system. Coach can say whether the record includes age, profile sex when provided, sleep stages, REM score or REM minutes, preferred sleep temperature, and room or bed temperature context, then name the next missing input to add. It cannot automate temperature, control a sleep device, diagnose sleep quality, replace a sleep study, send alerts, recommend care changes, or turn missing data into a personalized medical plan.

Sleep-device automation readiness is also review-only. When you ask about a smart bed, temperature schedule, smart alarm, or sleep automation, Coach can inventory source bed temperature, room temperature, preferred sleep temperature, routine target, and sleep outcome rows, then name activation gaps such as consent, rollback, and audit logging. It cannot control a bed, schedule alarms, automate temperature, recommend exact temperature settings, diagnose sleep conditions, replace a sleep study, send alerts, or change care.

Sleep-system attribution readiness is request-scoped too. When you ask about a shared bed, bed partner, side of the bed, dual-zone sleep system, or whose sleep rows are represented, Coach can inspect source labels for side, user, and shared-system attribution before using those rows. If shared sleep context is present without attribution labels, Coach names that gap instead of treating every row as yours. It cannot infer partner health, merge partner data into your record, share data, control a bed, automate temperature, diagnose sleep conditions, replace a sleep study, send alerts, or change care.

Daylight and sound exposure context is source-only too. Coach can summarize 7 and 14 day coverage for time in daylight, environmental sound, and headphone audio rows, plus missing exposure days. It cannot infer cause, hearing status, sleep quality, care needs, hearing screening, clinical monitoring, alerts, or care advice from those rows.

Body-clock readiness uses those same source-only lanes when the question asks about body clock, circadian rhythm, sleep timing, daylight, shift work, food timing, late caffeine, late meals, or fasting windows. Coach can receive ready domains, missing inputs, review questions, sleep timing context, daylight coverage, food timing, later food or drink clues, fasting windows, and recovery context. It cannot recommend exact bed or wake times, set alarms, schedule notifications, diagnose circadian or sleep disorders, prescribe fasting or diet changes, treat sleep conditions, monitor in real time, send alerts, or automate devices.

Pre-lab context windows are also deterministic. When accepted lab dates are available, Coach can summarize the sleep debt, activity, meal or fasting timing, symptoms, and medication or supplement context already logged in the 24 to 72 hours or week before the result. This is prep context only: it does not judge whether a lab is valid, diagnose, prescribe, order labs, or decide what caused a result.

Personal lab-baseline shifts use the same deterministic boundary. With clinical-context consent, Coach can compare your latest accepted lab result with your prior personal median for that same marker when at least three dated readings across at least 30 days in one unit exist. If units conflict, dates are missing, or there is not enough accepted history, it says the baseline is not ready instead of guessing. The wording stays neutral and history-only: no abnormality claim, diagnosis, target, lab order, treatment, prescription, or medical decision.

When you ask why lab results that look fine might still deserve a prevention discussion, Coach can build a typical-range lab debrief from the same accepted rows. It names recognized markers inside general typical-range bands, highlights any typical-range marker that moved against your prior personal median, lists missing accepted context such as ApoB, Lp(a), or hs-CRP when lipid rows are present, and carries linked MedlinePlus plus cardiology-marker sources. It does not clear safety, diagnose, calculate ASCVD, PREVENT, MESA, cancer, plaque, event, or prevention risk, order labs, set targets, prescribe, recommend imaging, or decide care.

When the question is about cumulative LDL exposure, non-HDL exposure, or cholesterol history over time, Coach can receive a longitudinal lipid exposure packet from accepted dated lipid rows. It reports reading count, date span, time-weighted average, and exposure area in mg/dL-years, with source links to longitudinal LDL exposure research. It stays source-history only: no ASCVD, PREVENT, MESA, plaque, event, or lifetime risk, no target, no statin or treatment advice, no lab order, and no safety clearance.

For home blood-pressure questions, Coach can organize source-provided paired systolic and diastolic readings into a log-readiness packet. It counts paired days, morning and evening timestamp coverage, source labels, missing cuff or setup notes, clinician questions, and linked AHA, USPSTF, ACC/AHA, and MedlinePlus source context. It does not diagnose hypertension, assign a blood-pressure category, set a target, validate the cuff, prescribe, change medications, monitor, alert, or decide care.

For pulse-wave-velocity or source cardiovascular-age questions, Coach can also receive a source-provided arterial-stiffness summary. It names logged PWV, source cardiovascular age, recent reading coverage, and linked source context, while keeping ASCVD, PREVENT, MESA, plaque, event-risk, diagnosis, treatment, monitoring, alerting, and care decisions out of the answer.

It also reads a visit-prep summary built from those same records. That summary can say which source details are present for a primary-care appointment, medication questions, lab-result questions, a weight-care conversation, or a mental-wellness conversation. It can also organize urinary symptom visit prep from bladder or urination concerns, allergy context, medication or supplement context, urinalysis, urine culture, kidney, or related lab context, and timing or visit questions. It can also organize asthma and respiratory follow-up context from logged wheeze or breathing concerns, inhaler records, trigger notes, allergies, respiratory vitals, and appointment questions. It only lists which source details are already on file and which ones are missing. It does not calculate clinical risk or screening scores, diagnose a UTI, choose antibiotics, order cultures, triage kidney infection or emergency symptoms, diagnose asthma, prescribe or adjust inhalers, monitor breathing, triage emergencies, or replace a clinician.

Accepted FHIR and C-CDA clinical-history rows feed that same prep layer. Encounters, procedure history, service requests, medication dispense or administration history, immunization history, family-history context, care plans, goals, bounded Device source context, bounded Provenance source-lineage context, questionnaire-response input context, imaging-study metadata, and document references can help Coach organize visit-prep, vaccine-history questions, family-history questions, referral questions, second-opinion context, imaging or procedure follow-up, and post-visit checklists after you accept them from the review queue. Care-plan follow-up readiness can also turn those accepted rows into a local summary of what changed, whether explicit follow-up timing or encounter context is present, and which source details are still missing before a follow-up visit. It reaches ready only with an accepted anchor, follow-up timing or encounter context, a dated timeline, and a source file or import. Post-visit instruction readiness separately organizes accepted after-visit summaries or discharge instructions into review-only follow-up tasks, medication-change mentions, pending tests or referrals, warning-language excerpts, and source gaps. Between-visit tracker packets can place accepted condition, encounter, document, or procedure anchors beside dated symptom, mood, food, routine, and wearable tracker lanes, but only as context and clinician questions. Coach can also use accepted C-CDA allergy and adverse-reaction rows as allergy context after clinical-context consent, while nested reaction or severity labels stay out unless they are part of a reviewed row. Exporting those rows in the health-data package does not expand Coach’s boundary: allergies stay medication-safety context only, medication dispense and administration rows stay medication-history context only, immunizations stay vaccine-history context only, and family-history resources stay document-style family-history context only, not allergy diagnosis, severity verification, vaccine scheduling, family diagnosis validation, genetic-risk scoring, screening decisions, genetic-test decisions, recommendation, prescribing, pharmacy fulfillment, administered-care proof, adherence proof, clinician review, or clinical validation. Adding package-local FHIR Provenance resources to the health-data package also does not expand Coach’s boundary: Coach may explain export traceability when asked, but it does not treat package Provenance as clinical evidence, provider authorship, source correctness, identity proofing, clinician review, or legal attestation. Imported Provenance rows can provide source-lineage labels and counts after review, but they do not prove authenticity, consent, chain of custody, source correctness, clinician review, clinical validation, or authorization. Accepted resource-family metadata can also be summarized for Coach as bounded RelatedPerson contact counts, Device source labels, Specimen provenance labels, FamilyMemberHistory relationship and condition labels, QuestionnaireResponse item labels, and ImagingStudy modality, series, instance, and endpoint-reference counts. It does not expose raw contact values, serial numbers, UDI values, questionnaire answers, endpoint URLs, image bytes, or attachment contents. Device rows can provide source-device labels and counts after review, but they do not pair devices, activate live sync, monitor, alert, control hardware, validate a medical device, advise on recalls or warranties, prove source correctness, prove clinician review, diagnose, or recommend care. These rows remain context only: Coach does not diagnose, order imaging, decide a vaccine, clear a procedure, create a referral, decide coverage, complete a care plan or goal, change clinician instructions, create a treatment plan, monitor you remotely, message a clinician, or treat a document reference as proof of clinician review.

When you ask for second-opinion or record-review help, Coach can also receive a deterministic second-opinion record packet. It names accepted records and results, treatment or procedure timeline context, medication, supplement, allergy, and background-history context, missing inputs, and questions to bring to the reviewing clinician. Coach uses it as packet prep only. It does not render the second medical opinion, recommend treatment, override a treating clinician, order tests, send records, or claim clinician review.

When you ask what to bring to a visit, Coach can also receive a deterministic multimodal visit-attachment readiness packet. It names whether outside records, clinician notes, photo or image metadata, and video or voice-note labels are ready, thin, or missing from accepted Ojava context. Coach uses those finished readiness lines for organization only. It does not interpret images, transcribe audio or video, send messages, create a provider inbox, claim clinician review, diagnose, prescribe, order labs, or promise a clinician response.

Coach also receives a source-quality and ingestion-confidence summary for imported records when your clinical-context consent is active. It can see aggregate readiness signals such as accepted rows, pending rows, source breadth, parser-loss counts, manual-review files, failed extraction jobs, dated-event coverage, repeated-concept counts, source-trail counts, and next repair steps. If rows are still pending review, Coach can also receive an import-review decision packet that names date, value or source-text, unit, source-trail, and coding checks before you accept or reject the row. It is review organization only, not auto-acceptance, source-correctness certification, clinical validation, or a care decision. It also checks the latest question against broad record areas such as labs, medications, wearables, nutrition, symptoms, and records, so it can say when the answer has direct accepted rows, when support is partial, or when a missing record area should be named before general education. It can also include a source-cohort transparency line for the current question: candidate rows, included rows, excluded rows, exclusion reasons, included and candidate source-trail counts, time window, and coded-row coverage. For source-quality questions, Coach can also receive a replayable source cohort definition manifest with the stable manifest id, inclusion rules, exclusion reasons, date window, source-trail counts, coding coverage, privacy protections, and disabled research or clinical claims. Coach can also receive a source cohort replay audit that compares the saved manifest with a fresh aggregate rebuild, then names whether the checked fields reproduced or changed without exposing row ids, filenames, source names, raw values, or free text. For source-quality questions, Coach can also receive a question evidence workbook: a stable review id, requested record areas, candidate and included counts, bounded generic evidence-row labels, generic source-trail labels, excluded-row reasons, date window, and coding coverage. Records-search coverage also names accepted searchable categories, source-trail counts, imaging-report coverage, doctor-note coverage, report-document coverage, and missing context labels for visit prep. Source contract receipts add only aggregate source coverage, evidence levels, row coverage, hash-presence counts, and limitations. Coach does receive source identity consistency as aggregate counts for accepted source trails, subject-signature groups, dependent-context hints, generic source rows, same-day agreement review, and reconciliation findings, so it can name source blending gaps without seeing raw subject ids or source identifiers. Coach does receive unit-normalization readiness for numeric source rows, including recognized unit rows, missing units, unrecognized unit text, and mixed-unit groups, so it can ask for source cleanup without converting values or judging clinical meaning. Coach does not receive raw filenames, source-file hashes, storage paths, source resource ids, raw subject ids, or a claim that identity is proven, records were merged, a source is clinically correct, or the data is clinically validated, certified, research-ready, or regulatory anonymized. Unstructured extraction governance adds only aggregate readiness context for scan, photo, report, and note extraction: parsed-file counts, manual-review files, failed files, completed and failed extraction jobs, skipped parser items, pending rows, and the requirement that model output carry exact source phrases, context qualifiers, confidence labels, and schema or model accounting before rows are trusted for review. When the current question is about OCR, scanned PDFs, photos, or medication-list extraction, Coach can also include document-extraction readiness: whether the local photo and common scanned-PDF path is active, which optional higher-fidelity OCR provider is selected, and whether the extraction model, fail-closed validator, route wiring, and pending review queue are ready. That readiness line never includes secret values or implies that extracted rows bypass review. When you ask about a gut-health or microbiome test report, Coach can also name whether Ojava has an uploaded report or only result-access guidance. Test purchase, sample processing, result unlock, partner-account access, microbiome scoring, probiotic prescribing, diet prescribing, diagnosis, treatment, and medical nutrition therapy stay outside Ojava.

That Coach context is the Coach-facing part of the source-quality workflow. Reports and Ingestion can also carry a source workflow manifest, source identity consistency packet, reconciliation-conflict profile, source-priority policy, whole-health coverage profile, and repair loop so the record set has a clearer provenance trail before Coach leans on it. Coach can name the source-priority policy when a question depends on conflicting or uneven source trails, but only as a human review order for accepted records. It must not choose the clinically correct source, hide conflicting rows, or change the record. Feed-drift review needs a real prior source-quality snapshot to compare with the current one. Reports and Ingestion can save minimized source-quality snapshots for that comparison, and Coach can name the saved comparison date when one exists. A first upload is treated as missing comparison data, not as evidence that a source feed drifted.

For lab-depth questions, Coach can also see a deterministic biomarker panel readiness packet built from accepted lab results. It groups represented areas such as metabolic, lipid, thyroid, liver and kidney, blood counts, inflammation, iron, vitamins and minerals, electrolytes, hormone context, urine and kidney detail, immune context, exposure context, coagulation and cardiac-stress context, and specialty screening markers, then names which areas are still not represented. The biomarker panel catalog belongs to this same readiness lane, so Coach can compare accepted results against Ojava’s 74-marker Standard target and 149-marker Concierge target (external reference-panel depth names, unrelated to Ojava’s own Free, Starter, and Pro plans), explain represented markers, missing education, and core versus add-on style panel coverage without creating a separate lab flow. For advanced lipid and cardiology-marker questions, Coach can also receive a request-scoped advanced cardiovascular marker packet for ApoB, lipoprotein fractionation, Lp(a), hs-CRP, Lp-PLA2, myeloperoxidase, oxidized LDL, F2-isoprostanes, ADMA, SDMA, and TMAO context. Within that packet, ApoB, Lp(a), and hs-CRP can be summarized as cardiology risk-enhancer discussion context using source-backed thresholds, but Coach does not calculate an ASCVD, PREVENT, MESA, plaque, event, diagnosis, treatment, medication, lab-order, or treatment-risk score. The same biomarker education cards shown in Insights render linked marker-specific sources for matched labs, so the user-facing card and Coach packet share the same citation registry. This is coverage, education, and visit-prep context only: no lab ordering, checkout, diagnosis, treatment plan, clinical score, external score calculation, or care decision. The same lab lane now adds annual plus mid-year panel profile readiness. It checks accepted uploaded results for a broad annual profile, a mid-year recheck window, missing marker families, upload-only Function Health and Superpower report profiles, visit-discussion prep flags, and protocol review prompts. It can prepare questions for discussion, but it never claims completed clinician review, orders labs, sells testing, diagnoses, treats, prescribes, or creates a treatment plan.

For kidney, renal, eGFR, creatinine, cystatin C, BUN, uACR, urine albumin, urine protein, albuminuria, proteinuria, or CKD questions, Coach can also receive a kidney function and albuminuria prep packet. It organizes source rows for filtration markers, urine albumin or protein context, blood pressure, metabolic context, medications, missing inputs, clinician questions, and linked KDIGO, National Kidney Foundation, and NIDDK source references. When an adult profile has age, sex, and a serum creatinine row in mg/dL but no imported eGFR row, Ojava can add a 2021 CKD-EPI creatinine eGFR estimate as source-backed visit-prep context. When serum cystatin C in mg/L is also available, it prefers the 2021 CKD-EPI creatinine-cystatin C estimate. The same packet also includes kidney source reconciliation: direct accepted eGFR remains the source row when present, creatinine in mg/dL or umol/L can support a method-check line, and same-date urine albumin plus urine creatinine can produce uACR source-readiness only when compatible units are present. It does not replace a lab result, prove source correctness, diagnose CKD, stage CKD, diagnose albuminuria, calculate kidney-failure risk, recommend treatment, order labs, refer to nephrology, or claim clinician review.

For iron, ferritin, anemia, hemoglobin, hematocrit, transferrin, TIBC, transferrin saturation, low iron, iron panel, MCV, or reticulocyte questions, Coach can also receive an iron and anemia prep packet. It organizes your reviewed iron panel and CBC source rows (ferritin, serum iron, transferrin or TIBC, transferrin saturation, hemoglobin, hematocrit, MCV, RDW, reticulocyte), anemia-associated symptom context, menstrual or other blood-loss context, current iron or multivitamin supplement context, medication context, missing inputs, clinician questions, and linked NHLBI, MedlinePlus, and American Society of Hematology source references. It does not diagnose anemia or iron deficiency, interpret or classify iron status against reference thresholds, calculate iron stores, dose iron supplements or infusions, recommend transfusion, order labs, recommend treatment, or claim clinician review.

For testosterone, low testosterone, hypogonadism, TRT, androgen, low libido, or erectile questions, Coach can also receive a testosterone and low-libido hormone prep packet. It organizes your reviewed hormone panel source rows (total and free testosterone, SHBG, LH, FSH, estradiol, prolactin), hormone-associated symptom context, current supplement and medication context, missing inputs, clinician questions, and linked American Urological Association, Endocrine Society, and MedlinePlus source references. It reminds you that a single low reading is not a diagnosis and that guidelines call for a repeat morning fasting draw, which is a clinician decision. It does not diagnose hypogonadism, low testosterone, or androgen deficiency, interpret or classify hormone levels against reference thresholds, recommend or dose testosterone therapy or supplements, interpret fertility results, order labs, recommend treatment, or claim clinician review.

For vitamin D, vitamin B12, folate, magnesium, homocysteine, vitamin B6, omega-3, zinc, micronutrient, or vitamin- or mineral-deficiency questions, Coach can also receive a vitamin and micronutrient panel prep packet. It organizes your reviewed micronutrient source rows (vitamin D 25-hydroxy, vitamin B12, folate, magnesium, homocysteine, vitamin B6, omega-3 index, zinc), micronutrient-associated symptom context, current supplement and medication context, missing inputs, clinician questions, and linked NIH Office of Dietary Supplements and MedlinePlus source references. It only appears when you ask about a vitamin, mineral, or deficiency. It does not diagnose a vitamin or mineral deficiency, interpret or classify micronutrient levels against reference thresholds, recommend or dose supplements or repletion, prescribe a diet, order labs, recommend treatment, or claim clinician review.

For thyroid, TSH, hypothyroid, hyperthyroid, Hashimoto, Graves, T4, T3, TPO antibody, thyroglobulin, levothyroxine, goiter, or thyroid-nodule questions, Coach can also receive a thyroid panel prep packet. It organizes your reviewed thyroid source rows (TSH, free T4, free T3, total T4 or T3, TPO antibodies, thyroglobulin antibodies, thyroid-receptor antibodies, reverse T3), thyroid imaging or nodule report context, thyroid-associated symptom context, current supplement and medication context, missing inputs, clinician questions, and linked American Thyroid Association, AACE and ATA hypothyroidism guideline, and MedlinePlus source references. It flags that some supplements such as high-dose biotin can affect thyroid immunoassays, so lab timing is a detail to confirm with the ordering clinician and lab. It does not diagnose thyroid disease, hypothyroidism, hyperthyroidism, Hashimoto disease, or Graves disease, interpret or classify TSH or thyroid-hormone levels against reference thresholds, recommend or dose thyroid medication such as levothyroxine, interpret thyroid nodules, ultrasound, or biopsy, order labs, recommend treatment, or claim clinician review.

For uric acid, urate, gout, hyperuricemia, podagra, tophi, allopurinol, febuxostat, or colchicine questions, Coach can also receive a uric acid and gout prep packet. It organizes your reviewed serum uric acid result, gout or flare symptom context, kidney-stone context, current urate-lowering or flare medication context, urate-raising medication context (some diuretics and low-dose aspirin), diet or alcohol context, missing inputs, clinician questions, and linked American College of Rheumatology gout guideline, 2020 ACR gout management, and MedlinePlus source references. It does not diagnose gout or hyperuricemia, interpret or classify uric acid against reference or treat-to-target thresholds, recommend or dose urate-lowering therapy such as allopurinol or febuxostat, recommend or dose flare treatment, order labs, recommend treatment, or claim clinician review.

For MASLD, MASH, fatty liver, steatosis, liver enzyme, ALT, AST, GGT, FibroScan, elastography, hepatic fat, fibrosis, or FIB-4 questions, Coach can also receive a MASLD and liver-metabolic prep packet. It organizes accepted liver enzymes, platelets, bilirubin, albumin, metabolic and body context, alcohol context, medication and supplement context, imaging or report text, missing inputs, clinician questions, FIB-4 input readiness for age, AST, ALT, and platelets, and linked AASLD nomenclature and practice-guidance sources. It does not diagnose MASLD or MASH, stage fibrosis, calculate FIB-4 or NAFLD fibrosis score, interpret elastography or imaging, recommend treatment, prescribe diet, order labs, refer to hepatology, or claim clinician review.

For DEXA, bone density, BMD, T-score, osteoporosis, osteopenia, fracture history, FRAX, fall prevention, gait, balance, walking steadiness, vitamin D, calcium, or PTH questions, Coach can receive a bone health and fall-prevention prep packet. It organizes accepted bone-density report language, fracture or fall history, mobility context, vitamin D, calcium, PTH, thyroid or kidney context, body context, medication and supplement context, FRAX-input details, missing inputs, clinician questions, and linked BHOF, USPSTF, CDC STEADI, and FRAX sources. It does not diagnose osteoporosis or osteopenia, calculate FRAX or fracture risk, score fall risk, interpret DEXA or imaging, recommend treatment, prescribe supplements or exercise, order labs or imaging, refer to a specialist, or claim clinician review.

For result-waiting, test-status, activation, account-access, or provider error-screen questions, Coach can also see a lab result activation readiness packet. It counts lab-like imported sources, accepted lab rows, and pending lab rows, then names provider-owned next steps. It cannot sell tests, activate partner kits, process samples, unlock partner accounts, order labs, interpret unreleased results, diagnose, prescribe, or create a treatment plan.

Coach can also see measurement education coverage from those accepted labs and any source-provided glucose summary. That packet maps represented measurement domains, missing source areas, source-glucose coverage, and marker education categories so Coach can explain what the record can support before answering biomarker questions. It does not create lab orders, connect devices, diagnose, score glucose control, monitor in real time, send alerts, give medication or insulin guidance, create a treatment plan, or provide medical nutrition therapy.

Nutrition questions can also use deterministic food-quality context from reviewed food rows. Coach can read label-detail coverage, protein anchors, fiber clues, plant-variety clues, processing clues, and current 7-day versus prior 7-day movement for those same signals. Nutrition Progress Overview macro lines for protein, carbs, and fat appear only after those weekly averages and directions are computed outside the model. They stay pattern wording, not a diet prescription, macro prescription, diagnosis, treatment, or medical nutrition therapy. When exact HEI-2020 constituent rows are present, Coach can also read a finished Healthy Eating Index 2020 score or a named missing-input list from Ojava’s deterministic calculator; ordinary food logs and macros do not complete that score. Coach also gets a stricter plant-variety context that counts distinct plant-food names only when your reviewed descriptions or notes support them, such as berries, leafy greens, legumes, whole grains, nuts and seeds, herbs, and spices. Broad meal names without ingredients become detail gaps, not inferred foods. The same packet now includes gut-nutrition clues from reviewed rows, such as fiber, plant variety, fermented foods or probiotics, processing clues, total-sugar context, and saturated-fat context when present. When the reviewed diary has enough recent logged days, fiber detail, explicit plant-food clues, and fermented-food context, Coach can anchor a compact gut nutrition readiness summary for gut, microbiome, fiber, fermented-food, plant-diversity, probiotic, ultra-processed, or nutrition-readiness questions. It also includes a carb-quality readiness packet from reviewed total carbs, parsed fiber, sugar, added-sugar, sugar-alcohol, and net-carb support, so Coach can discuss what the diary is ready to review without creating a glycemic score, food ranking, diet prescription, diabetes-management plan, diagnosis, treatment, or medical nutrition therapy. It also includes an ingredient and processing packet from reviewed food descriptions and notes only, with ingredient-group clues and package, prepared, restaurant, takeout, snack, dessert, or sweet-drink context when those words are present. A related processed-food risk-factor readiness packet gives Coach the evidence coverage, cue families, missing label or ingredient inputs, and review questions to use without assigning a score. This is diary review context only: no microbiome test, gut score, food score, plant-variety score, processed-food risk rating, processed-food score, diet prescription, diet advice, microbiome claim, diagnosis, treatment, or medical nutrition therapy. When enough reviewed food-log days and source-backed recovery rows are present, Coach can also see association-only nutrition habit patterns, such as late eating, protein-anchor days, high label-detail days, later caffeine or alcohol clues, and late heavier-meal clues compared with the following sleep, readiness, HRV, or resting-heart-rate day. Those associations are sample-guarded context only, not causation, food scoring, alerts, a diet prescription, treatment, or medical nutrition therapy. When explicit adaptive movement context is present, such as wheelchair pushes, seated workouts, chair workouts, mobility-aid notes, adaptive cardio, or source-provided adaptive mobility rows, Coach can also pair those movement days with reviewed food diary context before preparing access-aware review questions. It does not infer disability from low activity, classify disability, prescribe exercise, clear activity, set calorie targets, diagnose, treat, or create medical nutrition therapy.

For food database, restaurant food, community food, recipe catalog, quick-log, photo scan, label scan, barcode, food-search, food entry review readiness, phone food logging, mobile-sync, HealthKit, Health Connect, metabolic app-detail parity, MyFitnessPal, Fitbit, connected-grocery, shopping-list, and partner-writeback questions, Coach receives the matching readiness boundary too. It can name the live substitutes Ojava supports now: user-owned foods, saved meals, recipes, account-backed saved preference filters, account-backed custom foods, favorites, goals, reminders, shopping-list groceries, and Meal 5 or Meal 6 labels, MyFitnessPal CSV import, Open Food Facts label search, reviewed public restaurant menu paste import, reviewed recipe import prep, editable scan or barcode drafts, local ingredient matching from foods you already reviewed, account-backed reviewed tracker rows, source counts and completeness gaps from recent accepted food entries, phone access through Fitness > Nutrition at /fitness-tracker?tab=nutrition, accepted diary rows plus reviewed drafts for scan-to-grocery readiness, meal-plan grocery lists plus saved recipes for local connected-grocery export readiness without guessing provider state, Levels-style app-detail coverage across nutrition, workout, source-glucose, lab, report, privacy, and Health Connect activation lanes, private account-backed body progress photo storage with browser fallback for transient upload or preview failures, and the visible Nutrition sync status card in Fitness, including source-readiness profiles for MyFitnessPal, Fitbit, and food-name writeback. For MyFitnessPal, Fitbit, or Health Connect field questions, Coach can also read a reference-only partner sync contract that names meal summaries, custom meal mapping to Other, link-date-forward behavior, Premium-gated individual food names, no historical meal backfill, one-way water semantics, and partner weight-field behavior. Coach also receives one consolidated native and platform sync diagnostic packet for those questions, so account-backed Ojava rows, native shells, HealthKit, Health Connect, live MyFitnessPal sync, external writeback, and local grocery export are shown as live or gated from the same readiness sources. The separate server-backed connected-grocery card is the source of truth for actual Kroger and Chicory availability and Kroger account status. Coach must not claim a copied corpus, scraper, exposed provider catalog, automatic recipe nutrition, automatic image logging, food database accuracy certification, food scoring, full mobile tracker-setting sync, live provider sync, external writeback, MyFitnessPal writeback, checkout, cart creation without the activated Kroger flow, delivery scheduling, substitution acceptance, diagnosis, treatment, prescribed diet, or medical nutrition therapy.

For data-rights questions, Coach can also prepare a bounded request-readiness packet. It separates selected-data deletion from account deletion, names likely scopes such as third-party platform data, imported records, tracker logs, AI request context, consent and audit trail, or account profile, and points back to Settings privacy choices and retention review. Coach can remind you to export first and identify missing scope details, but it does not delete records, close accounts, delete source-app accounts, purge backups, send legal notices, promise deadlines, or provide legal advice.

For guided workout, meditation library, mindfulness catalog, streamed class, or licensed wellness content questions, Coach receives the licensed-content activation boundary too. It can name the provider contract, rights review, media manifest, program taxonomy, review queue, disabled activation flags, and audit placeholders that must exist before a future library can go live, while pointing to current live substitutes: your own practice logs, Ojava guided wellness tools, Mental wellness mode, self-authored mindfulness goals, and logged workout or holistic sessions. It must not claim hosted third-party media, licensed content access, streamed classes, provider-created programs, workout prescription, meditation therapy, clinician response, or an exposed catalog.

For wearable, health-hub, app, CGM, smart-scale, and CSV setup questions, Coach also receives a source setup readiness packet built from the same provider catalog and recent imported source rows. It can say whether a source already has rows in Ojava, whether file import is ready today, or whether Apple Health, Health Connect, cloud sync, or partner access is still activation-gated. It must not open vendor OAuth, store external credentials, write back to source apps, control devices, monitor in real time, send alerts, diagnose, treat, or promise partner availability. CGM and glucose-source setup questions also get a dedicated blood-glucose source query before the packet is built, so glucose rows are not lost behind denser wearable streams. This same setup packet covers the MyFitnessPal connected-app floor, including Polar Flow, TrainingPeaks, MapMyFitness, MapMyRun, MapMyWalk, MapMyRide, Renpho Health, Strava, Garmin, Fitbit, Withings, Dexcom, Apple Health, Google Fit, and Health Connect.

Coach can also see language-access prep from the same accepted records and communication-readiness builders. It can help organize a plain-language source summary, visit communication needs, and an interpreter handoff checklist. It does not generate certified translations, translate medical instructions, route interpreter services, deliver care, or replace a qualified interpreter or clinician.

For continuity, follow-up, next-visit, or handoff questions, Coach can also receive the Reports continuity timeline. That timeline names accepted record anchors, repeated lab or vital changes, current medication context, and record activity cues, then turns them into patient-controlled carry-forward notes. It is record organization only: local review, no completed care action, and no proof that a clinician reviewed the timeline.

For beauty and appearance questions, Coach also receives a deterministic prep summary when your record includes dermatology, cosmetic, skin, hair, nail, supplement, sleep, stress, or nutrition context. It can name what is ready for a provider conversation, what is missing from the timeline, what questions to ask, and whether the record has enough dated notes, routine or procedure anchors, response details, lifestyle context, and repeated skin or hair follow-up points for continuity tracking. It can also receive a routine-consistency packet with recent logged check-in days, routine anchors, missing routine context, and a safe next question, so it asks for the missing product category or dated check-ins instead of guessing. It does not interpret photos, recommend a procedure, schedule care, prescribe products, diagnose a skin condition, build a treatment plan, or claim clinician review. For skin, hair, progress-photo, before-and-after, or photo-privacy questions, Coach also receives a beauty photo readiness packet. It can explain dated notes, bounded body-photo metadata, capture guidance, and privacy guards, while naming dedicated skin and hair photo intake as activation-gated until private storage, consent, review copy, and audit handling are live. For skincare routine or formula-change questions, Coach receives a routine-change prep packet that organizes goals, current routine context, duration and use frequency, dated reactions, possible triggers, lifestyle context, private photo metadata, and provider questions without advising a formula change. It cannot request image bytes, expose storage paths or signed links, interpret photos, score skin, infer facial age, treat hair loss, prescribe products, recommend procedures, or claim public sharing or clinician review. For appearance-age or face-age questions, Coach can use a saved manual outside result only when you have allowed it, and can say what result or consent context is missing. It cannot request image bytes, interpret a face photo, infer how old you look, score skin, recommend products or procedures, validate an outside result, or claim clinician review.

The snapshot also checks common lab and vital categories for record completeness before the model answers. If broad areas such as lipids, blood sugar, thyroid, vitamin D, kidney function, CBC, liver enzymes, or CRP are missing or older, Coach can surface that as a discussion prompt and record-repair cue. It does not present those prompts as a diagnosis, prescription, lab order, or care plan.

That same prep layer now covers common care pathways too: urgent care, dermatology, UTI and urinary visits, sexual and reproductive health, menopause and midlife health, blood-pressure and heart-health reviews, diabetes and metabolic-care reviews, thyroid conversations, at-home testing discussions, dental and vision visits, joint or muscle rehab, sleep-medicine conversations, nicotine or alcohol discussions, travel-medicine visits, imaging or procedure prep, and screening or genetics conversations. It stays in the same lane for every pathway: what is present, what is missing, and what to bring to a licensed clinician. When you ask about one of those expanded pathways, Coach can promote it into the compact snapshot before unrelated prep items crowd it out.

It can also prepare care-navigation context for conversations you manage outside Ojava: specialist-visit questions, second-opinion record prep, pharmacy and prior-authorization context, insurance and benefits questions, family or pediatric visit prep, fertility and pregnancy questions, and post-visit follow-up questions. Care-plan follow-up readiness stays in that same lane: education, follow-up organization, local report context, and questions to bring to your own clinician. These are record-completeness summaries, not referrals, second opinions, coverage decisions, prescriptions, lab requests, pediatric medical advice, urgent-care decisions, remote monitoring, clinician messages, treatment plans, or care-plan changes.

For outside-provider selection or visit-prep questions, Coach can organize specialty goals, user preferences, candidate profile details to review, source context to bring, and questions to ask before you choose where to book outside Ojava, without choosing a provider, creating a referral, confirming coverage, booking a visit, messaging a clinician, or promising a clinician response.

For broader support-routing questions, such as where to start, which support option fits, what EAP or coverage questions to ask, or where crisis resources fit, Coach keeps local structured wellness tools, record-grounded reflection, crisis-resource signposts, and therapy cost or benefits prep in separate lanes. It does not provide therapy, diagnose, assess crisis risk, match providers, book visits, decide coverage, advise psychiatric medication, escalate crises, or promise a clinician response.

When your accepted imports include payer or benefits records, Coach also gets a separate administrative context section for benefits, network, eligibility, and prior-authorization prep. Those rows are kept out of generic clinical history and lab-result sections. Coach can help you organize questions for a plan, pharmacy, facility, or clinician office, but it does not decide coverage, approve benefits, submit claims, submit prior authorizations, recommend payment, or replace your plan.

The benefits packet now checks payer documents for plan or payer name, member ID, group number, plan phone or portal path, and claim, authorization, invoice, or account references before Coach names missing call details. That lets Coach avoid asking for identifiers already present in accepted payer documents.

For EOB, medical bill, patient statement, billing-office, and itemized-bill questions, Coach can receive a narrower reconciliation packet. It checks whether an EOB or claim response, provider bill, amount-style fields, reference numbers, and service context are present, then turns that into call questions. It still cannot calculate the amount owed, decide whether a charge is covered, submit a claim, file an appeal, send anything to a payer, give legal advice, or recommend payment.

When the question combines preventive care with insurance, such as coverage for a screening, vaccine, annual checkup, genetic-risk review, screening lab, or procedure follow-up, Coach receives a preventive benefits prep packet. It organizes payer evidence, missing plan details, and questions to ask. It still cannot decide coverage, say something is due, order care, schedule care, submit a claim, submit a prior authorization, send anything to a payer, or give payment advice.

The Care panel also exposes Family care, Dental and vision, Nicotine and alcohol, and Travel medicine commands. They stage focused Coach prompts from the same prep layer, organizing caregiver, immunization, allergy, medication, symptom, growth, development, school, feeding, sleep, document, follow-up, oral, eye, substance-use, insurance, pharmacy, family, and appointment context when it is present. They do not diagnose dental or vision conditions, prescribe lenses or dental treatment, diagnose substance-use disorder, manage withdrawal, prescribe cessation medication, create detox plans, prescribe travel vaccines or medications, clear travel, decide which vaccines are needed, or act for a dependent.

Care-plan follow-up readiness adds a narrower prep packet for what changed since the last care-plan context, what still needs a source, and what to ask at a follow-up visit, not a live care plan or proof that a clinician reviewed it.

Reports composes record-completeness readiness into a local visit-prep packet: a visit-brief data room, key changes, attachments, open decisions, and the first questions to ask next. Coach can help turn that packet into plain-language prep text before you review or export it. The packet is still local and review-only. It does not message a clinician, promise a response, prescribe, order labs, create referrals, diagnose, create recipient access, or complete any care action.

The portable emergency card uses the same boundary. Ojava Coach may help explain or polish the card text after deterministic prep gathers accepted identity, allergies, medications, supplements, conditions, emergency contacts when present in source rows, bounded FHIR RelatedPerson emergency-contact metadata, and recent timeline context. It cannot turn the card into emergency triage, decide urgency, create a QR or live sharing link, send to a provider inbox, diagnose, or promise a clinician response. It can also name whether the short local card is ready to print or export for review based on required allergies, medications, emergency contacts, and recommended condition and latest context sections. That readiness line is still local prep only, not QR sharing, live sharing, provider inbox delivery, permissions, external messages, or a care decision.

Those first questions are ranked deterministically before Coach narrates them. The ranker looks at recent symptom entries, user goals, current medication details, older or missing record categories, source-quality gaps, prep-lane gaps, and consented wearable context. Each question carries a reason for its priority, so Coach can explain why it is worth asking without inventing urgency or a care decision. The output stays bounded to visit preparation: no diagnosis, triage, prescription, lab order, referral, treatment plan, or claim that a clinician will act.

Scientific literature when you ask for evidence

If you explicitly ask for studies, citations, guidelines, or evidence on a supported health topic, Ojava Coach can attach a small scientific-literature context block before it answers. The lookup sends only sanitized health-topic terms, such as a marker or wearable concept, to PubMed, ClinicalTrials.gov, DailyMed for supported medication-label context, openFDA for supported medication recall, enforcement, and FAERS adverse-event report context, and, when configured, OpenAlex. It does not send your full message or your patient record to those literature APIs.

Results are bounded, deduplicated, and can be reranked with Cohere when that optional key is configured. Trial registry records are labeled separately, because a listed trial is study landscape context, not proof that an intervention works and not an eligibility decision. DailyMed records are labeled as official medication-label context, not instructions to start, stop, change, or dose a medication. openFDA enforcement records are labeled as official safety context, not instructions to stop, start, replace, or change medication. FAERS aggregate counts are labeled as spontaneous adverse-event report context, not incidence rates, causality proof, diagnosis, or medication-change instructions. The Coach may cite only the exact IDs returned in that block, such as [Evidence 1], and it is told not to invent studies, PMIDs, DOIs, journals, trial results, medication labels, recall records, adverse-event report counts, or findings. If no evidence block is attached, it can still give general education, but it must not pretend it retrieved live papers.

When a Coach turn uses retrieved research items, the chat also shows a compact Evidence used block under the answer. That block is not written by the model. Ojava builds it deterministically from the same retrieved items, listing the sanitized topic query, confirming that patient data was not sent, summarizing the evidence mix by source type and limitation, and showing each evidence label, provider, identifier when the source supplied one, URL, title, year, and source label. The evidence mix is retrieval transparency only. It does not grade study quality, prove causality, diagnose, rank treatments, or decide care. The block is the owning surface for this metadata because research evidence only exists for the specific Coach message that invoked it. It is not part of patient source quality or Reports provenance.

Evidence support map

If you ask what supports an answer, which signals matter most, or what to prioritize first, Coach receives a deterministic support map from the same whole-health readiness domains. It sorts current context into direct record support, partial support, missing inputs, and self-tracked wellness support, then names the next source or result worth reviewing before the model writes the answer.

The map is a review aid only. It does not grade studies, calculate clinical certainty, prove causality, diagnose, rank treatments, order labs, prescribe, monitor emergencies, or decide care.

Safety comes first, before the model

Before Ojava Coach answers anything, a deterministic safety check scans your message for signs of an emergency or a mental-health crisis. If it sees unambiguous self-harm language or an emergency symptom, it returns a fixed message pointing you to 911 or the 988 Suicide and Crisis Lifeline and does not run the model on it at all. The chat shows a blocking notice you must acknowledge, plus clear 911 and 988 action rows in the thread, so the next step is not buried inside a paragraph. This check is hard-coded and cannot be talked out of, it is not left to the model’s judgment.

The public free assessment uses the same red-flag-first idea before sign-in. It does not call the model or read your record. It checks fixed urgent warning signs, asks for the basic context a clinician would want, and creates a copyable Ojava Coach prompt for non-emergency symptom prep. It is not a diagnosis, treatment plan, monitoring service, doctor visit, prescription, referral, or emergency service.

Your record is data, never instructions

Ojava Coach treats your records as information to reason about, never as instructions to follow. Text inside a note, lab, or field cannot tell the assistant to change its rules or cross its boundaries, that separation is built in, so a stray line in an imported document can never turn into advice it acts on.

Modes and packets

Modes turn a question into something useful. Visit-prep modes can produce a visit-prep packet, a focused set of questions, or an educational read on your own data that you own and can take to a clinician or other licensed professional, never an action Ojava takes on its own. Coach does not draft a prescription, lab order, referral, or medication change; those licensed actions stay outside Ojava. Mental wellness stays chat-only so it does not open a misleading condition packet.

• Symptom prep: check a symptom for urgent warning signs first, ask the missing questions, then organize the questions and records to bring to a clinician, without triaging or diagnosing.
• Visit prep: assemble a one-page brief, the top questions, and the supporting records for your next appointment, referral packet, second-opinion request, or family-care visit.
• Condition prep: map a common concern to the records it calls for and the questions to ask, without diagnosing.
• Mental wellness: use mood, stress, energy, sleep, recovery, symptoms, goals, medication context, and saved neutral notes to build a supportive wellness plan and questions for a clinician or therapist. It is not therapy, diagnosis, psychiatric-medication advice, or crisis care.
• Wellness plan: review nutrition, training, sleep, recovery, and other habits from your longitudinal trends, with missing context and optional visit questions kept separate.

A drafted packet opens beside the chat so you can read and copy it. If labs or imported results are withheld by consent or are still in review, the draft says that plainly instead of filling in a value.

If you ask about add-on labs, imaging, scans, multi-cancer blood-test questions, screening-test questions, or visit prep, Coach can also receive a narrow discussion-prep packet from uploaded evidence. It uses accepted labs, reviewed reports, procedure rows, visible family history, symptom notes, and goals to name discussion topics, evidence counts, missing information, and questions for a clinician. If the uploaded report mentions coronary CTA, CAD-RADS, plaque, stenosis, Agatston, or coronary calcium context, Coach can carry a CAD-RADS 2.0 source link and report-review questions without interpreting images. It is not test ordering, screening guidance, diagnosis, image interpretation, plaque or event-risk calculation, medication advice, clinician review, Galleri access, MRI access, CT access, CCTA access, or any other test-access path.

If you ask for a cardiology ownership packet, Coach can combine accepted standard lipids, ApoB, Lp(a), hs-CRP, family history, symptoms, CAC, coronary CTA, CAD-RADS, plaque, stenosis, coronary-calcium report context, FAI or perivascular fat attenuation language, and stress-test, nonobstructive-CAD, INOCA, microvascular-angina, coronary-flow-reserve, or vasospasm source language into one review-prep bundle. The packet cites cholesterol and non-HDL source references when standard lipid rows are present, the ACC/AHA lipid and risk-enhancer sources already used for ApoB, Lp(a), and hs-CRP, a European Heart Journal ApoB discordance source when ApoB is reviewed beside LDL-C, non-HDL-C, and triglycerides, the CAD-RADS 2.0 source for coronary CTA report language, CRISP-CT FAI sources when coronary-inflammation language is present, the AHA/ACC chest pain guideline source, and an EAPCI INOCA consensus source. It is not an ASCVD, PREVENT, MESA, plaque, FAI, coronary-inflammation, INOCA, microvascular, or event-risk calculator, and it does not diagnose coronary disease or ischemia, rule out heart disease, interpret images, set lipid targets, recommend medication, order labs or imaging, send referrals, recommend testing, or claim clinician review.

If you ask about an undiagnosed rare disease, diagnostic odyssey, exome or genome reanalysis, HPO phenotype terms, ClinVar context, or genetic-counselor prep, Coach can build a local diagnostic-odyssey prep packet from accepted records. It organizes genetic or genomic source report labels, phenotype and symptom clues, family history, prior workup context, missing inputs, and clinician questions. The packet carries source links for the 2026 NEJM AI rare-disease reanalysis study and the OpenAI study summary. It is not diagnosis, variant interpretation, variant classification, genetic-risk scoring, a genetic-test recommendation, test ordering, referral routing, consumer-AI diagnosis, or clinician review.

If you ask for a caregiver or family summary, Coach can also prepare a local family-caregiver packet from reviewed records, current medication and supplement lists, allergy context, symptoms, goals, upcoming visit metadata, and pending-review counts. Sensitive facts are withheld by default, and imported source rows still require clinical-context consent. The packet is for your own review before you copy or export anything outside Ojava. It does not send records, invite anyone, grant permissions, create caregiver accounts, provide provider access, diagnose, prescribe, triage emergencies, or monitor anyone.

For preventive checkups, screenings, and vaccine questions, Ojava Coach organizes the age, sex, immunization, family-history, prior-screening, lab, vital, medication, allergy, and goal context already in your record. It helps you prepare what to ask a clinician, but it does not decide that a screening, vaccine, genetic test, lab, or procedure is due. It lists present and missing source context for the conversation, not a schedule, recommendation, order, or clinical review.

Recent chats are listed by title and time in the session-history popover in the header. A chat is saved shortly after you send a real message, then the finished reply updates the same thread. Reloads, navigation, and normal browser sessions should preserve the conversation unless you explicitly delete it.

If you ask where to start, what to prioritize this week, or for a 7-day or 30-day plan, Ojava Coach turns the attached record into a short focus plan. It chooses a few priorities, explains why they matter from your own data, and keeps the next steps to wellness habits, tracking, records to import, or questions for a clinician.

The public resources page is the free starting point for lab education, condition-oriented guides, and visit-prep worksheets. It does not interpret your private labs. Record-grounded lab review, trend explanation, and personalized follow-up questions belong inside signed-in Ojava Coach and any plan gates that apply, where Coach can use accepted records and consent choices. Even there, Coach cannot diagnose, prescribe, order labs, promise clinician review, or replace medical care.

If you combine several health questions in one message, such as labs, medication safety, and a plan for the week, Ojava Coach separates the parts into a compact table. Safety and licensed-care boundaries come first, then it gives the next safe step and asks which part you want to explore.

If you ask it to compare health options, such as sleep versus LDL or recovery versus nutrition, it uses a compact table with the option, what your record shows, the upside, the tradeoff, and the next step. The final pick stays inside wellness support: a habit experiment, tracking step, record to import, or clinician question.

If you ask for a deeper explanation, Ojava Coach separates the plain explanation from what your record actually shows. That makes follow-up questions more useful without turning the answer into a diagnosis, treatment plan, lab order, or medication change.

If you ask about off-label prescription medications or research compounds for longevity, Ojava Coach keeps the answer educational. It can organize the proposed mechanism, evidence limits, possible risks, monitoring topics to discuss, and the record context a clinician would need, but it will not recommend, dose, source, prescribe, or explain how to obtain the medication.

If you ask how several signals connect, such as sleep, workouts, nutrition, labs, mood, or symptoms, Ojava Coach uses a synthesis table. It separates what your record shows from possible connections, avoids causation claims, and gives one next useful step to watch or track.

Cross-domain lifestyle discovery also lives in that lane. It looks for sample-guarded associations across user-authored food tags and notes, medication or supplement routine context, mood or symptom notes, wellness practices, workouts, and source-backed sleep, HRV, resting-heart-rate, readiness, recovery, stress, or energy rows. It stays association-only: no causation, diagnosis, treatment, medication dosing, medication start or stop advice, or behavior scoring.

When a response includes clinician or pharmacist questions, Ojava Coach keeps those questions in a labeled copyable block. Draft text uses the same copy control so you can move it into your notes without selecting text by hand.

If you ask for care gaps, hidden insights, missed patterns, or what a clinician might have missed, Ojava Coach frames it as a record review, not a diagnosis. It gives three concise insights and closes with a priority table of what to ask, track, or import next.

If you ask for personalized health insights, whole-health review, or what stands out from your results, Ojava Coach ranks a compact set of insight prompts from reviewed records, labs, medications, wearables, nutrition context, and consent status. It names the record support, the next useful input, and what it cannot do, including diagnosis, lab ordering, prescribing, supplement approval, emergency monitoring, or claiming clinician review.

If you ask for a whole-health check-in, Ojava Coach can privately plan a review across symptoms, mood and stress, sleep and recovery, nutrition, hydration, movement, medications and supplements, body, cycle, and source gaps. It names what is present and what is missing, but it does not send reminders, schedule notifications, diagnose, prescribe, order labs, change treatment, create care tasks, or claim clinician review.

If you ask how confident it is or what supports an answer, Ojava Coach now separates what is grounded in your record from what is uncertain. It shows what would make the answer more reliable without inventing values, studies, or a clinical certainty score.

If you ask what to upload or what data would make the answer better, Ojava Coach reviews the record categories it can see, names what is missing, and gives a short starting list. It focuses on records that improve wellness coaching or clinician-prep questions, not unnecessary sensitive details.

If you ask how to prepare for an appointment, Ojava Coach turns your attached record into a visit-prep table: the visit area, what your record shows, questions to ask, and records to bring. When you ask for a visit handoff, it can combine that visit prep with appointment and language-access context into a portable packet for local review. It is organization support only. It does not diagnose, order labs, book or join care, arrange rides, certify accommodations, write notes, verify coverage, submit claims, provide interpretation or translation, message a clinician, promise a clinician response, or include medication dose or timing details.

If you start with a broad prompt like help me get started or what should we look at first, Ojava Coach turns that into a record-grounded menu instead of a generic feature tour. It shows the areas it can use, how each area can help, and a good first prompt for each one.

If you ask what it knows about you, Ojava Coach summarizes only the Ojava record attached to the chat. It separates what it can see from the plain read and the next useful step, can rank the next best record input when your imported context is ready enough, and says when a category is missing.

If you ask for an Ojava Health Score, a health credit score, or what one number says about your health, Coach receives the deterministic health-score packet from the same reviewed records. The headline number only appears when a complete validated primary score is present, currently Life’s Essential 8. Accepted, consented diet-quality and nicotine self-report observations can complete those two lifestyle inputs, including explicit reviewed rows saved from the signed-in Insights check-in; generic notes and food logs cannot. PhenoAge and other published components can appear as companion context, but they do not move the headline number. If the Life’s Essential 8 inputs are incomplete, Coach gets the missing-input list instead of inventing a score. The packet also names the configured freshness windows, the reviewed source-date range used for the current context, and the rule that companion or provenance domains can change confidence and missing-context wording but not the headline score. In Insights, the same cardiovascular wellness card renders linked research-basis sources beside the score rather than leaving the citation trail only in internal metadata.

If Coach receives a lab-based biological-age estimate, the deterministic context includes the PhenoAge formula name, linked research sources, and the same non-diagnostic boundary shown in the result card. Cancer-risk evidence can be discussed only as cohort-level association context, not as a personal cancer-risk score or screening decision.

When accepted fasting glucose, fasting insulin, triglyceride, lipid, waist, height, BMI, or body-metric rows support HOMA-IR, triglyceride-glucose index, cardiometabolic index, lipid accumulation product, visceral adiposity index, METS-IR, or lipid-pattern scorecards, Coach receives the finished values plus compact source references. HOMA-IR links back to the Matthews 1985 HOMA paper, the triglyceride-glucose index links back to the Simental-Mendia 2008 TyG paper, and the cardiometabolic pattern indices link back to their primary formula papers. Those values stay educational metabolic context only, not insulin-resistance diagnosis, diabetes diagnosis, medication guidance, treatment planning, or clinician review.

Lipid-derived scorecard context follows the same rule. When accepted total cholesterol, HDL, LDL, triglyceride, ApoB, or ApoA1 rows support a lipid-pattern index, Coach receives the finished deterministic value plus source URLs for the cholesterol and non-HDL references that also appear on the Insights scorecard. Those links support education and visit prep only, not ASCVD, PREVENT, MESA, plaque, event, diagnosis, treatment, medication, lab-order, or lipid-target advice.

CBC-derived immune-pattern indices follow the same source trail. When accepted absolute neutrophil, lymphocyte, monocyte, and platelet rows support neutrophil-to-lymphocyte ratio, platelet-to-lymphocyte ratio, systemic immune-inflammation index, systemic inflammation response index, or aggregate index of systemic inflammation, Coach receives finished values plus source URLs from the same registry that appears on the Insights card. These values stay educational immune-pattern context only, not infection diagnosis, cancer prognosis, COVID-19 severity scoring, treatment planning, medication guidance, lab ordering, or clinician review.

Kidney, liver, and serum-protein pattern ratios also carry sources when accepted rows support BUN-to-creatinine ratio, AST-to-ALT ratio, or albumin-to-globulin ratio. Coach receives the finished deterministic values plus public source URLs, including the note when globulin was derived from total protein minus albumin. These values stay educational chemistry pattern context only, not kidney disease diagnosis, liver disease diagnosis, dehydration diagnosis, nutrition diagnosis, immune disorder diagnosis, cancer diagnosis, treatment planning, medication guidance, lab ordering, or clinician review.

If you ask what changed lately or what trends stand out, Ojava Coach compares earlier and latest record details across labs, wearables, sleep, training, nutrition, symptoms, mood, and goals when those histories are present. When there is not enough history yet, it says that plainly.

If you ask for a whole-health protocol review or a marketplace-prep packet, Coach receives a deterministic packet built from the same reviewed Ojava context: records, accepted labs, consent, medications and supplements, wearable summaries, nutrition or body context, active goals, and self-tracking changes. The packet lists which sections are ready, what is still missing, and which questions to bring to a clinician or pharmacist. It also includes external-review package readiness: ready data areas, missing attachments, user-controlled sharing actions, disabled outside-service actions, and the boundary that keeps the packet local until you choose what to share. It is prep for your own review or for an outside service you choose, not a protocol, test order, account connection, supplement approval, diagnosis, treatment plan, or claim that a clinician has reviewed anything.

For nutrition progress questions, Coach also receives deterministic daily and weekly nutrition context when enough rows exist: same-day digest, weekly diary digest, Nutrition Progress Overview, Nutrition Progress Report that joins diary provenance with body progress context, prior-week recall, annual nutrition recap over reviewed food-log rows, weekly review-readiness over reviewed diary days, meal detail, body-weight check-ins, and activity context, reviewed-day status, reviewed MyFitnessPal food and mobile nutrition import summaries, reviewed MyFitnessPal exercise-import step context and strength-set counts when present, source-specific MyFitnessPal exercise, water, and progress import provenance packets, workout-entry review readiness for duration, category, distance, source-calorie, step-provenance, and strength-detail gaps, the body-composition readiness packet over accepted measurements, source-provided smart-scale rows, BMI, body-fat range context, body-shape indices, weight trend, and missing inputs, the consolidated body progress report for accepted measurements and bounded photo metadata, food timing review, same-day nutrient focus, weekly calorie and macro budget, saved target-date goal estimates, same-day activity-adjustment state for 0 to 100 percent exercise add-back, including signed source activity adjustments and negative adjustments when reviewed source context supports them, saved per-meal calorie-goal progress, meal macro ratios from reviewed protein, carb, and fat rows, daily macro remaining from saved daily targets, meal-plan coverage, local grocery readiness, consolidated meal-prep sessions, recent meal-plan history, planned-vs-logged review, household meal-prep portions with named split and leftover context, recipe follow-through for saved or starter recipes not yet planned or logged, provider-neutral local connected-grocery export context for reviewed needed items, request-scoped food entry review readiness across recent quick add, barcode, food search, label scan, meal-photo scan, library, recipe, and nutrition-app import rows with calorie-to-macro consistency prompts, review-before-logging food ideas filtered against saved preferences, foods to avoid, and allergy notes, Nutrition Coach history and follow-through changes, logged fasting rhythm and active-timer context, food timestamp consistency against completed and still-open fasting windows, fasting readiness recap context over fasting rhythm, food-window consistency, reviewed water logs, same-day food timing, and recovery or body-clock evidence, gut-nutrition clues, gut nutrition readiness for matching questions, ingredient-processing clues, processed-food education cues over reviewed package, prepared-food, restaurant, snack, dessert, and sweet-drink diary language, association-only meal timing around workouts and movement days, scan-to-grocery readiness with local store-section suggestions from accepted or reviewed rows, imported app agreement, inert food-name writeback readiness, manual water entries, hydration digest readiness with reviewed row counts and recent averages, water reminder status, account-backed food reminder status, private nutrition challenge candidates with current streaks, recent completions, missing inputs, Missed-day recovery plan reset actions, and the private-only boundary, privacy-safe nutrition accountability prep with the same de-identified outside-support prompt and redaction checklist shown in Fitness > Nutrition, lab-enriched metabolic program selection for Heart Health, Weight Management, Glucose Stability, or Metabolic Exploration, the selected metabolic weekly priority, first habit, source lanes, progress fields, and missing follow-through context, selected-day tracker action summary for reviewed or explicitly selected diary days, including food entries, calories and macros, water, movement or steps, latest body context, meal-copy shortcuts, missing inputs, and the next best logging action, Fitness Progress Overview, and adaptive macro review. Nutrition Coach quick actions and custom questions also carry the current deterministic review summary when they open Coach. The snapshot sends finished values, then Coach explains them without recomputing targets or inventing missing numbers. Meal percentages are meal composition ratios, not per-meal macro targets. If net carb view is enabled, Coach also receives that preference and the reviewed-fiber boundary. The preference and allergy-fit packet is a review filter only, not a food-safety guarantee, allergy diagnosis, diet prescription, or medical nutrition therapy. Missed-day recovery plan is the same kind of private self-tracking context: it is not an adherence score, diet plan, calorie prescription, fasting prescription, exercise prescription, diagnosis, treatment, medical monitoring, or medical nutrition therapy. Fasting food-window consistency is timestamp cleanup only, not an adherence score, instruction on when to eat, diagnosis, treatment, or medical nutrition therapy.

Medication and supplement awareness

If you ask what medications or supplements are in your record, Ojava Coach summarizes the list by name and separates what appears in the record from what to confirm. It does not repeat dose, schedule, timing, pharmacy directions, or instructions in the chat. When accepted imported medication-list rows are available under clinical-context consent, Coach can compare them with your current Profile medication list and name which items are Profile-only, imported-only, duplicated by name, or missing matching detail context. That reconciliation stays name-only in chat and is for clinician or pharmacist review prep.

When you are taking more than one medication or supplement, Ojava Coach can raise a well-documented, source-grounded interaction worth verifying with your pharmacist or clinician. It explains the interaction in general educational terms and always defers the decision to a licensed professional. It starts with a short safety assessment, uses a table for the combination, main concern, and what to watch for, then separates urgent-help warning signs and pharmacist or clinician questions. It will not rate severity, set a dose, change timing, or tell you to start or stop anything, that is the line between awareness and advice, and it stays on the awareness side. Reminder/refill prep, medication supply and refill-readiness context, GLP-1 support, GLP-1 dose-log fidelity, diary-native GLP-1 context, weight-care prep from food, water, body, workout, symptom, and routine rows, self-reported dose history, and request-scoped medication response context are treated the same way: useful visit context, not eligibility, prescribing, dosing advice, live alerts, cause claims, side-effect diagnosis, effectiveness judgment, adherence judgment, adherence proof, refill approval, or medication-change guidance.

Unknown pill questions use an even stricter safety shape. Ojava Coach will not identify a pill from appearance, an imprint, a photo, or a damaged bottle label, and it will not list likely matches. It routes the next step to a pharmacist, the prescribing clinician, poison control, or emergency care if the pill may already have been taken or symptoms are present.

Boundaries

Every chat shows a clear disclaimer, and each finished Coach answer includes a small reminder that it is AI support for education and prep, not a licensed clinician. Finished answers also include a clinician-question prep action that stages a follow-up prompt for you to review, not a message to a clinician. That follow-up asks Coach to put exactly three questions in a copyable text block so you can move them into your notes. The assistant explains and prepares; it does not decide. Diagnosis, prescribing, dosing, lab orders, referrals, clinical risk scores, regulated monitoring, and clinician messaging stay outside Ojava. Confident lifestyle coaching (training, recovery, sleep, stress, hydration, and nutrition habits) is encouraged within general wellness, while clinical decisions stay out of the current self-serve chat.

Ojava Coach by text

When texting is available for your account, open Settings and choose Link my phone. Text the six-digit code from the phone you want to use. A verified, linked phone can ask Coach a question, log a workout, send a meal photo, or send a lab photo for the same review queue used by record imports. Imported lab values do not count until you review and accept them in Ojava.

Texts are convenient but are not fully private. Avoid sending anything you would not want in your normal message history. Ojava checks the phone link and current consent again before it saves health information. Replies with several parts are delivered in order, and an uncertain carrier result is held for review instead of being sent twice.

Reply STOP to opt out. STOP cancels replies that have not started sending and prevents an earlier request from saving after the opt-out. Reply START to turn texts back on only for the same previously verified phone. If the phone is not already linked, create a new code in Settings instead. You can also unlink the phone from Settings at any time.