Armbrain terms and their developer mappings.
Core Concepts
| Term | Meaning | Implementation |
|---|
| Client Mind | A partitioned memory space for one client | One row in client_minds table |
| Memory | Any stored knowledge about a client | One row in memories table |
| Mind ID | The foreign key that partitions all data by client | mind_id FK on memories, ingest_sources, ingest_log, raw_documents, brand_dna_history, meeting_briefs |
| Brand DNA | The 5 structured fields that define a client's marketing identity | tune, brand_positioning, approved_messaging (jsonb), stakeholders (jsonb), content_themes (text[]) on client_minds |
| Active client mind | The currently selected client context for all tool calls | Set via switch_client, stored in server session state |
| Customer ID | The owning user account in hosted mode | customer_id on client_minds, enforced by application-layer query scoping |
| Internal client mind | A non-client mind (e.g., "CMO Preferences") excluded from mind | is_internal = true on client_minds |
Memory Types
| Type | What It Captures | Example |
|---|
decision | A choice made by or for the client | "Pausing LinkedIn ads through Q2" |
preference | A stated or observed preference | "Prefers casual, direct brand voice" |
campaign_outcome | Results and learnings from a campaign | "Email open rate 34%, curiosity gap subject lines won" |
commitment | An action item someone committed to | "Ben will send revised messaging by Friday" |
stakeholder | Information about a key person | "Sarah Chen, VP Marketing, owns rebrand brief" |
fact | A general fact about the client | "Series B, 45 employees, developer-first API platform" |
question | An open question to investigate or revisit | "Do we need FTC compliance review for the testimonial ads?" |
Source Types
How memories are created. Stored as memories.source_type.
| Value | Description |
|---|
manual | Stored via store_memory tool |
memory | Stored via memory tool (no classification) |
meeting | Extracted from meeting-oriented ingestion |
session_extract | Extracted from watched AI conversation logs |
folder_ingest | Extracted from documents ingested via CLI |
folder_watch | Extracted from documents detected by the folder watcher |
email | Extracted from email via Gmail polling |
slack | Extracted from Slack (future) |
Architecture Terms
| Term | Meaning |
|---|
| Product layer | CMO-specific tools: brand DNA, Business OS rocks, meeting prep, daily loop, campaign history |
| Memory layer | Generic storage engine: store, search, classify, embed, dedup, ingest |
| Gateway | Cloudflare Worker at api.armbrain.io that proxies calls to Anthropic and Voyage AI |
| MCP | Model Context Protocol -- the interface between your AI and the Armbrain server |
| Reconsolidation | Automatic dedup: when a similar memory exists above a similarity threshold, the existing row is updated instead of creating a duplicate |
| Extraction pipeline | Turns a meeting transcript into memories: chunk -> dedup -> LLM extract -> store |
| Embedding | A 512-dimensional vector representation of memory text, used for semantic search |
| pgvector | PostgreSQL extension for vector similarity search |
| RLS | Row-Level Security -- PostgreSQL policies that filter rows at the database level. In Armbrain these are defined in the schema as defense-in-depth; database-level enforcement is being activated in stages, and application-layer query scoping is the enforced isolation boundary until then |
Business OS Terms
| Term | Meaning | Implementation |
|---|
| Rock | A quarterly goal (Business OS methodology) | Row in rocks table |
| Sprint | A time-boxed work period within a quarter | Tracked via sprint plan + deliverables |
| L10 | Level 10 meeting (Business OS weekly meeting format) | Generated by business_os tool |
| IDS | Identify, Discuss, Solve -- Business OS issue resolution workflow | Managed by business_os / get_issues |
| Scorecard | Weekly metrics tracked in Business OS | Included in L10 prep output |
| Cadence | The business rhythm for a client (Business OS quarterly, monthly, etc.) | cadence jsonb field on client_minds |
User Story Vocabulary
The original Armbrain personal memory system uses different terms. Here is how they map to the CMO product:
| User Story Term | CMO Product Term | Notes |
|---|
| Stone | Memory (high-confidence, decision/fact type) | Permanent knowledge that rarely changes |
| Ember | Memory (preference/insight type) | Living knowledge that may evolve |
| "Stone it" | store_memory with appropriate type |
| "Ember it" | store_memory with appropriate type |
| "Task it" | store_memory with type "commitment" | No separate task system; commitments are memories |
- Memory Storage and Search -- working with memories
- Database Schema Reference -- table and column details
- Architecture Overview -- how the layers fit together