Four Knowledge Bases. Every Clinical Decision Comes From a Rule.
Behind every plan TeraHealth delivers are four proprietary, versioned knowledge bases. They map biomarkers to conditions, conditions to cited protocols, conditions and goals to foods as medicine, and each drug to its coaching protocol. The rules make the clinical decisions. AI only brings their outputs together and puts them into words.
4
proprietary, versioned knowledge bases behind every plan
Hundreds
of biomarkers mapped to possible conditions
1,000+
whole and minimally processed foods mapped to the conditions and health goals they support
How it works
How the four fit together.
The patient's data and prescription enter on the left. Each step is a lookup in a knowledge base, not a guess. What comes out is a draft plan that people, not software, approve.
Patient inputs
Biomarkers
The patient's biomarker results.
Patient health history
Past and current health information the patient shares.
Symptoms
What the patient is experiencing now.
Health goals
What the patient wants to achieve.
Prescription
The drug and where the patient is in its titration schedule.
Four knowledge bases
Biomarkers → Conditions
Possible conditions
Ranked for a practitioner to confirm.
Conditions → Protocols
Cited condition protocols
Each line backed by a peer-reviewed citation.
Conditions & Goals → Foods as Medicine
Specific foods
What to eat, not general advice.
Drug → Coaching Protocol
Drug coaching protocol
Matched to the titration phase and its side effects.
Draft plan
AI brings the rule outputs together
It integrates and summarizes what the knowledge bases decided. It adds no clinical decisions of its own.
Approval
Coach and trainer approve
A health coach and a certified personal trainer review and approve the plan before the patient sees it.
The four knowledge bases.
Each one is maintained by TeraHealth and versioned, so the same inputs always lead to the same rule outputs.
Biomarkers → Conditions
From biomarker results to possible conditions.
- Hundreds of biomarkers, mapped deterministically to the possible conditions they point to
- The Clinical Reasoning Engine ranks the possible conditions and surfaces them for a practitioner to confirm
- No condition protocol is drafted until a practitioner confirms the condition
Built and maintained by our Clinical Lead (Medical Doctor and MS in Nutrition & Longevity)
Conditions → Protocols
From a confirmed condition to an evidence-cited protocol.
- Each condition the Clinical Reasoning Engine identifies maps to protocols across six domains
- Every protocol line carries a peer-reviewed citation
- Practitioners can open any line of the protocol and inspect the citation behind it
Built and maintained by our Clinical Lead (Medical Doctor and MS in Nutrition & Longevity)
Conditions & Goals → Foods as Medicine
From what the patient has and wants to what they eat.
- 1,000+ whole and minimally processed foods, each mapped to the conditions and health goals it supports
- Turns nutrition guidance into specific foods the patient can buy and cook
Built and maintained by our Clinical Lead (Medical Doctor and MS in Nutrition & Longevity)
Drug → Coaching Protocol
From the prescription to coaching for each phase of therapy.
- Drug profiles: injectable and oral GLP-1s, plus the maintenance phase after stopping
- Titration-phase templates: coaching mapped to each step of the titration schedule, updated when a schedule changes
- Side-effect overlays: coaching for the side effects that make patients quit, such as nausea, constipation and muscle loss
- Safety rules and escalation: anything about dosing, labs or risk goes straight to the prescriber
- Switch path: guidance for patients moving from a GLP-1 to a non-GLP-1 weight-loss drug instead of quitting
Built and maintained by our Clinical Lead (Medical Doctor and MS in Nutrition & Longevity)
Example patient journey
One patient, four knowledge bases.
An illustrative example: a patient early in the titration schedule of a GLP-1 who reports nausea. This is what each knowledge base contributes to their plan.
Inputs
What the patient brings
Biomarker results, health history, symptoms such as nausea and fatigue, a goal of losing weight while keeping muscle, and a GLP-1 prescription at an early titration step.
Four rule-based lookups
What each knowledge base contributes
Biomarkers → Conditions
01Possible conditions, confirmed by a practitioner
The patient's biomarkers can point to several possible conditions. The Clinical Reasoning Engine ranks them and surfaces them for a practitioner to confirm.
Conditions → Protocols
02Cited protocols for each confirmed condition
Each confirmed condition maps to its own protocol across nutrition, supplements, fitness, sleep, stress and lifestyle interventions, with a peer-reviewed citation behind each line. Here the AI synthesizes all these protocols into one personalized protocol for the patient.
Conditions & Goals → Foods as Medicine
03Specific foods for the condition and the goal
Whole and minimally processed foods that support the confirmed condition and the goal of keeping muscle while losing weight.
Drug → Coaching Protocol
04Lifestyle coaching for this titration step
The template for the patient's titration step adds the side-effect overlays that match what they report, such as nausea, constipation, fatigue or muscle loss, with eating, hydration and activity coaching for that step. It never touches the dose: dosing questions go to the prescriber.
AI summary
One plan, in plain language
A language model combines the four outputs into a single plan for the patient and the coach. It adds no clinical decisions of its own.
Approval
Coach and trainer approve
A health coach and a certified personal trainer review and approve the plan before it reaches the patient.
Three of the four knowledge bases work the same for any patient. Biomarkers to conditions, conditions to protocols, and foods as medicine don't depend on the drug. Only the drug module changes, starting with hormone therapy next.
Clear responsibilities
Rules decide. AI only summarizes.
Clinical buyers ask what the AI actually does. The answer is deliberately narrow.
Clinical rules
Rules make the clinical decisions
Possible conditions, protocols, foods and drug coaching all come from rules in the four knowledge bases. AI never generates a clinical hypothesis or a protocol.
AI
Language models integrate and summarize
They combine the outputs of the four knowledge bases into one plan and write it in plain language for the patient and the coach.
Human review
People approve every plan
A health coach and a certified personal trainer review and approve every plan before it reaches the patient. Coaches handle lifestyle only. Dosing, labs and risk stay with the partner's prescriber.
The same knowledge bases run under both delivery options. Whether you choose the Managed Solution or Build It Yourself, TeraHealth hosts and maintains the four knowledge bases and every plan goes through the same review workflow.