v1-data-model-spec · supabase_json
Limitless V1 Lean Thought-to-Action Data Model
v1-data-model-spec artifact · for Limitless OS · status draft
Agent Handoff
Start Here
v1-data-model-spec artifact · for Limitless OS · status draft
Completion Evidence
No explicit evidence field yet. Require tests, screenshots, linked PRs, or reviewed outputs before marking complete.
Artifact Shape
- title: Limitless V1 Lean Thought-to-Action Data Model
- source: atlas-codex
- purpose: Design the V1 data model for turning captured thoughts into memory, decisions, tasks, projects, drafts, approvals, workflow rules, questions, outcomes, and learning updates. This is design-only and does not create database objects yet.
- entities: 13 items
- client id: limitless
- project id: 07dee66f-8143-4f4d-9af9-b1ed413d2767
- updated at: 2026-05-10T02:23:19.448Z
- hcs mapping: principle: string, hcs_payload_fields: object, forbidden_in_hcs_payloads: object
- project name: Limitless OS
- artifact type: v1-data-model-spec
- status models: task_status: object, draft_status: object, project_status: object, approval_status: object, workflow_rule_status: object
- schema version: 3
Structured Payload
Machine-readable source fields
title
Limitless V1 Lean Thought-to-Action Data Model
source
atlas-codex
purpose
Design the V1 data model for turning captured thoughts into memory, decisions, tasks, projects, drafts, approvals, workflow rules, questions, outcomes, and learning updates. This is design-only and does not create database objects yet.
entities
| name | notes | purpose | key fields | relationships | source of truth |
|---|---|---|---|---|---|
| thought | Existing Limitless MCP stores source=limitless-mcp and tag=limitless-os. | Raw captured input from Seth/ChatGPT/Claude/voice/email. The source of future memory and work. | - id - content - source - captured_by - captured_at - tags - metadata - embedding - project_id | - has many thought_classifications - has many thought_outputs - may link to projects/tasks/drafts/decisions | Projects Supabase public.brain_memory in current MCP implementation; future app model may wrap this with Limitless-specific tables/views. |
| thought_classification | One thought can have multiple classifications and lanes. | Agent-generated interpretation of what a thought might become. | - id - thought_id - classifier_agent_id - output_types - lanes - topics - sensitivity - confidence - reasoning_summary - created_at | - belongs to thought - feeds thought_outputs | Limitless V1 app/read model |
| thought_output | This is the join record that explains how one thought became multiple things. | Concrete routed result proposed or created from a thought. | - id - thought_id - classification_id - output_type - entity_type - entity_id - status - route_reason - created_by_agent_id - created_at | - belongs to thought - points to memory_item/decision/task/project/draft/workflow_rule_candidate/question | Limitless V1 app/read model |
| memory_item | Canonical operating principles require Seth approval. | Structured durable context derived from a thought or outcome. | - id - source_thought_id - memory_type - content - topics - canonical - sensitivity - created_by_agent_id - approved_by - created_at | - may reference thoughts - may inform agents - may link to decisions/projects/workflow rules | Projects brain_memory and/or future Limitless memory read model |
| decision | Strategic/business decisions are Seth-owned. | A recorded direction, priority, rule, or tradeoff with rationale. | - id - source_thought_ids - title - decision - rationale - owner - decision_date - status - supersedes_decision_id - related_project_id | - may belong to project - may create tasks/workflow rules - may update memory | Studio artifact or Limitless decision read model |
| task | Task = do this. It should be small enough to complete or route in one work cycle. | A concrete action for an agent or human. Tasks are the execution unit. | - id - hcs_topic - hcs_event_id - source_thought_ids - project_id - lane - task_type - summary - assigned_agent_id - human_owner - priority - due_timing - sensitivity - approval_required - status | - may belong to project - may produce draft/approval/outcome - may originate from thought_output | HCS queue for claim/audit; Supabase read model for current state |
| project | Project = build or operate this system. Seth approves promotion from thought/task to project. | A larger initiative made of tasks, decisions, drafts, workflow rules, milestones, risks, and approvals. | - id - title - goal - why_it_matters - owner - status - source_thought_ids - success_criteria - risks - approval_gates - created_at | - has many tasks - has many decisions - has many drafts - has many workflow_rule_candidates - has many outcomes | Studio project/artifact when canonical; Limitless app read model for operational project cards |
| draft | All V1 drafts require approval before external send/publish. | Language generated for review: message, script, post, newsletter, education snippet, SOP text, or brief. | - id - source_thought_ids - task_id - project_id - draft_type - audience - content - tone_notes - risk_flags - created_by_agent_id - status | - belongs to task/project optionally - requires approval_request before external use - may create memory_update after edits | Limitless V1 app/read model; final canonical docs may become Studio artifacts |
| approval_request | Approval owners include Seth, Betsy/front office, clinician/performance team, and privacy/guardrail review. | Human review request for a draft, task, project promotion, workflow rule, or sensitive memory update. | - id - target_entity_type - target_entity_id - approval_owner - approval_reason - allowed_actions - status - reviewed_by - reviewed_at - edits - decision_note | - belongs to draft/task/project/workflow_rule/memory item - creates outcome | HCS review queue plus Supabase read model |
| workflow_rule_candidate | Automation_mode values: manual_only, draft_only, human_approved, eligible_for_future_automation. | Proposed repeatable rule, SOP, checklist, cadence, or automation boundary discovered from thoughts/outcomes. | - id - source_thought_ids - title - rule_statement - trigger - action - owner - automation_mode - approval_required - status - risk_flags | - may belong to project - may create tasks - may become SOP/artifact after approval | Limitless app/read model until approved; approved rules may become Studio artifacts/SOPs |
| clarification_question | Questions prevent weak assumptions from becoming bad tasks. | A question generated when thought intent, owner, sensitivity, or next action is unclear. | - id - source_thought_id - question - asked_to - reason - status - answer - answered_at - followup_output_id | - belongs to thought - may produce classification/output after answer | Limitless app/read model / HCS human queue |
| outcome | Outcomes close the learning loop. | What happened after approval/action: accepted, edited, rejected, sent, published, scheduled, converted, deferred, escalated, or became SOP. | - id - entity_type - entity_id - outcome_type - outcome_summary - human_actor - agent_id - hcs_event_id - created_at - metrics | - belongs to any output entity - may create memory_update - may update project/task status | Limitless app/read model with optional HCS completion event refs |
| memory_update | Examples: Seth prefers shorter CTAs, Betsy wants two follow-up attempts before escalation, this content claim needs clinician review. | Structured learning created from an outcome, edit, rejection, or repeated pattern. | - id - source_outcome_id - update_type - content - topics - applies_to_agents - approved_by - created_at | - belongs to outcome - informs future classifications/drafts/routing | Projects brain_memory / Limitless memory model |
client id
limitless
project id
07dee66f-8143-4f4d-9af9-b1ed413d2767
updated at
2026-05-10T02:23:19.448Z
hcs mapping
principle
HCS carries queue/audit pointers, not sensitive bodies. Full content lives in authorized stores/read models.
hcs payload fields
- task_id
- role_topic_slug
- entity_type
- entity_id
- project_id
- sensitivity
- opaque_source_refs
- created_at
forbidden in hcs payloads
- PHI body text
- payer details
- patient financial details
- full draft text when sensitive
- staff-sensitive content
project name
Limitless OS
artifact type
v1-data-model-spec
status models
task status
- queued
- claimed
- drafted
- awaiting_approval
- approved
- completed
- deferred
- rejected
- escalated
- blocked
draft status
- drafted
- awaiting_approval
- approved
- edited
- rejected
- published_or_sent
- archived
project status
- proposed
- approved
- active
- paused
- completed
- cancelled
approval status
- requested
- approved
- approved_with_edits
- rejected
- deferred
- escalated
workflow rule status
- candidate
- under_review
- approved_manual
- approved_draft_only
- approved_human_gated
- rejected
- superseded
schema version
3
output type map
| approval | use when | output type | primary entity |
|---|---|---|---|
| none unless sensitive/canonical | The thought is useful context but not actionable yet. | memory | memory_item |
| Seth for strategic/business decisions | The thought records a direction, tradeoff, or priority. | decision | decision |
| depends on action sensitivity | There is one clear next action someone can do. | task | task |
| Seth approves project promotion | The thought implies a multi-step system, campaign, workflow, or initiative. | project | project |
| required before send/publish/use externally | The next useful output is language or content. | draft | draft |
| Seth/Betsy or clinician depending on domain | The thought describes repeatable behavior or an automation candidate. | sop_workflow_rule | workflow_rule_candidate |
| answered by assigned human owner | Intent, owner, risk, or next action is unclear. | question | clarification_question |
refinement note
Version 2 collapses conceptual entities that do not need independent V1 lifecycles. The goal is fewer tables/read models while preserving all behavior from the original design.
lean v1 entities
| name | notes | purpose | lifecycle | key fields | route types |
|---|---|---|---|---|---|
| thoughts | Current implementation source is Projects public.brain_memory through the Limitless MCP. | Raw captured input from Seth/ChatGPT/Claude/voice/email. Source material for routing, memory, tasks, projects, and drafts. | - captured - classified/routed - referenced - archived if needed | - id - content - source - captured_by - captured_at - tags - metadata - embedding - project_id | |
| thought_routes | Replaces thought_classification and thought_output. | Combined classification and output-routing record. Represents what a thought may become and what entity was created if accepted. | - proposed - accepted - rejected - superseded - created_entity | - id - thought_id - route_type - lane - topics - sensitivity - confidence - reasoning_summary - agent_id - status - entity_type - entity_id - human_owner - created_at | - memory - decision - task - project - draft - sop_workflow_rule - question |
| tasks | Replaces standalone clarification_question. Task = do this. | Concrete execution unit for agents or humans. Also represents clarification questions. | - queued - claimed - drafted - awaiting_approval - approved - completed - deferred - rejected - escalated - blocked | - id - hcs_topic - hcs_event_id - source_thought_ids - project_id - lane - task_type - summary - assigned_agent_id - human_owner - priority - due_timing - sensitivity - approval_required - status | |
| projects | Project = build or operate this system. Seth approves promotion to project. | Multi-step initiative/container for systems, campaigns, workflows, templates, or operating improvements. | - proposed - approved - active - paused - completed - cancelled | - id - title - goal - why_it_matters - owner - status - source_thought_ids - success_criteria - risks - approval_gates - created_at | |
| drafts | Workflow/SOP candidates can be drafts when they are proposed text rather than multi-step projects. | Generated language or structured artifact body for review: message, script, social post, newsletter, education snippet, SOP text, brief, or project note. | - drafted - awaiting_approval - approved - edited - rejected - published_or_sent - archived | - id - source_thought_ids - task_id - project_id - draft_type - audience - content - tone_notes - risk_flags - created_by_agent_id - status - version | |
| approval_requests | Approval owners: Seth, Betsy/front office, clinician/performance team, privacy/guardrail review. | Human review request for a draft, task, project promotion, sensitive memory item, or automation boundary decision. | - requested - approved - approved_with_edits - rejected - deferred - escalated | - id - target_entity_type - target_entity_id - approval_owner - approval_reason - allowed_actions - status - reviewed_by - reviewed_at - edits - decision_note | |
| outcomes | Outcomes close the loop and feed memory_items. | What happened after action/review: accepted, edited, rejected, sent, published, scheduled, converted, deferred, escalated, became SOP, or became project. | - recorded - learned_from - archived | - id - entity_type - entity_id - outcome_type - outcome_summary - human_actor - agent_id - hcs_event_id - created_at - metrics | |
| memory | Replaces memory_items naming. Raw MCP capture remains public.brain_memory / limitless.thoughts. RedKey platform memory remains public.agent_memory. Product-scoped reusable memory is limitless.memory. | Cleaned, structured, reusable product memory for Limitless OS agents. This is distinct from raw MCP thought capture and RedKey platform agent_memory. | - proposed - active - canonical - superseded - archived | - id - source_type - source_id - memory_type - content - topics - applies_to_agents - canonical - status - sensitivity - approved_by - created_at |
design principles
- Separate entities only when they have materially different lifecycle, ownership, approval, or query needs.
- Thoughts remain the raw source; routes explain what the system thinks a thought should become.
- Tasks are the execution unit; projects are the multi-step initiative/container.
- Drafts, approvals, outcomes, and memory each have separate lifecycles and stay separate.
- Clarifying questions are human tasks, not a separate entity.
- Decisions are memory items in V1 unless they become project-level decisions later.
- Workflow/SOP candidates are drafts or projects in V1; approved rules can later become canonical artifacts.
related artifacts
agent team spec
36c57e41-dbda-434b-ba2b-42e13cff1c29
v1 agent contracts
75daba90-3cc4-4f30-bbc8-046dba537a55
v1 hybrid loop spec
19911d80-32bc-4890-b033-c028aa0c0122
implementation roadmap
3bbfdbc8-0578-4542-8d64-8f64886b60a2
lifecycle examples
| thought | records created |
|---|---|
| Betsy should not have to remember to follow up with unscheduled evals. | - thought - thought_classification: task + project + sop_workflow_rule - project: Unscheduled Eval Follow-Up System - tasks: define cadence, draft templates, decide approval rules, identify source list, test with historical evals - workflow_rule_candidate: unscheduled eval follow-up cadence - approval_request: Seth/Betsy approve workflow rule - outcome and memory_update after review |
| Runners think cadence fixes everything, but it is usually load tolerance. | - thought - thought_classification: memory + draft + patient_education - memory_item: clinical/performance philosophy note - draft: social post/reel script - approval_request: Seth plus clinical review if claim-heavy - outcome: edited/published/deferred - memory_update: approved phrasing and claim boundary |
collapsed entity map
| reason | original | replacement |
|---|---|---|
| Classification and routing are one lifecycle in V1: proposed route, accepted/rejected, optional entity created. | - thought_classification - thought_output | thought_routes |
| Both are durable product memory with different source_type/memory_type values, scoped to Limitless OS. | - memory_item - memory_update | limitless.memory |
| A question is human work with owner, status, and outcome. | - clarification_question | tasks where task_type=clarification |
| A proposed SOP text is a draft; a repeatable system is a project. Approved canonical rules can become artifacts later. | - workflow_rule_candidate | drafts or projects |
| Most V1 decisions are durable Limitless product memory. Split later only if decisions need independent approval/versioning lifecycle. | - decision | limitless.memory where memory_type=decision |
open design questions
- Should V1 projects be Studio artifacts immediately, or lightweight operational project cards that only become Studio artifacts after Seth approval?
- Should content projects and operational projects share the same project entity or use typed project categories?
- Which approval events should write back into brain_memory as durable learning versus stay only in operational logs?
- What is the first external source after MCP: GoHighLevel/CM Hub, PT Everywhere export, Gmail, call logs, or manual queue entry?
task vs project rules
| rule | output | example thought |
|---|---|---|
| Create a task when the next action is concrete and can be done in one work cycle. | task | Draft a text for unscheduled eval follow-up this week. |
| Create a project when the thought implies a repeatable system, campaign, workflow, or multi-step build. | project with tasks | Betsy should not have to remember unscheduled eval follow-ups. |
| Create both when there is a clear immediate action and a larger system hiding behind it. | content draft task plus lead-nurture project candidate | We need a better knee pain post and a follow-up process for people who ask about it. |
| Ask a question when project scope is implied but success criteria or owner is unclear. | clarification_question | We need to get more aggressive with follow-up. |
naming refinement note
Version 3 renames memory_items to product-scoped Limitless Memory. Human/product name is Limitless Memory; database design name is limitless.memory; code/domain name is limitlessMemory.
implementation guidance
- Do not implement the original 13-entity model for V1 unless a lifecycle need emerges.
- Use thought_routes as the key traceability table from raw thought to created work.
- Use tasks for clarification questions, approval work, and project steps.
- Use limitless.memory for decisions and learning until decisions need their own lifecycle.
- Keep raw MCP captures in public.brain_memory / limitless.thoughts until curated into Limitless Memory.
- Keep HCS payloads as opaque pointers to tasks/projects/drafts/approvals, never sensitive bodies.
memory scope distinction
rule
Do not treat raw Seth brainstorming as canonical Limitless Memory until it is cleaned, structured, and optionally approved.
redkey platform memory
public.agent_memory
product scoped reusable memory
limitless.memory