Vault

# Vault — Content Intelligence Database **Name:** Vault | **Leader:** V (COO) | **Group:** Content & Media | **Status:** Active **Org Chart:** [Interactive Org Chart](../2026-03-08-ai-org-chart.html) --- ## Identity Vault is the local intelligence database — a 34MB SQLite powerhouse indexing 1,003 episodes, 114 articles, and 1,496 canonical records with full-text search. When any agent needs to find what was discussed, published, or referenced across the entire content library, Vault answers. It's the memory layer that makes content intelligence queryable rather than scattered across Sanity, HubSpot, and the filesystem. **Philosophy:** Intelligence without searchability is just storage. Vault makes every piece of content findable, cross-referenceable, and actionable. **Origin:** Content was scattered: episodes in Sanity, articles in markdown, transcripts on Google Drive, session notes in client folders. Finding "every time we discussed the 12 Traps on a show" required searching 4 different systems. Vault consolidates everything into a single FTS5-indexed database that any agent can query. --- ## Role Type **Not a standup agent. Vault is infrastructure — always available, queried by other agents on demand.** Vault doesn't produce reports or make decisions. It stores, indexes, and serves content intelligence. Other agents (Prism, Herald, Forge, daily-ops) query Vault to answer questions about the content library. **Activated by:** Direct query via `content-vault/query.ts`, agent imports from `content-vault/lib/search`, "What episodes discuss [topic]?", "Find all articles about [subject]" --- ## For Humans | | | |---|---| | **When to engage** | When you need to search across all content assets. "Which episodes covered the Value Path?" "How many articles mention NetSuite?" "What was the last Catalyst episode about?" Also consumed automatically by other agents during their workflows. | | **What you'll get** | Search results with episode metadata, article content, canonical record cross-references. Full-text search (FTS5) across all indexed content. | | **How it works** | SQLite database at `/mnt/d/data/content-vault.db`. Schema v3 with tables for episodes, articles, canonical records, YouTube comments, YouTube live chat. FTS5 indexes enable sub-second full-text search. Sync operations pull from Sanity CMS, filesystem, and YouTube API. | | **Autonomy** | Fully autonomous infrastructure. Sync runs on demand or via background workers. | ### Key Value Indicators | KVI | VP Dimension | What It Measures | Anti-Pattern | |-----|-------------|------------------|----| | Index Coverage | vp_cap_ute_maturity | Every published piece of content is indexed and searchable | Not: records in database | | Query Accuracy | vp_val_platform_leverage | Search results return relevant, current content | Not: queries served | | Freshness | vp_cap_operational_independence | Database reflects the current state of all content platforms | Not: sync frequency | --- ## For AI | | | |---|---| | **Activation** | Import from `content-vault/lib/search`, `content-vault/query.ts` CLI, agent queries | | **Skills** | None — Vault IS the skill (infrastructure layer) | | **Receives from** | Sanity CMS (episodes, articles), YouTube API (comments, live chat), Filesystem (session data), Forge (new episode records) | | **Reports to** | V (leader). Output consumed by: Prism (content multiplication source), Forge (pipeline state), Herald (content context for sessions), `/daily-ops` (content intelligence), `/show-prep` (episode substance) | | **Dependencies** | SQLite, `/mnt/d/data/content-vault.db`, Sanity API (for sync), YouTube API (for comments) | ### Database Schema (v3) | Table | Records | Content | |-------|---------|---------| | `episodes` | 1,003 | All show episodes across 11 shows | | `articles` | 114 | Published articles with full text | | `canonical_records` | 1,496 | Cross-platform content references | | `youtube_comments` | Variable | Community comments for intelligence | | `youtube_live_chat` | Variable | Live stream chat messages | ### Query Interface ```bash # CLI query npx tsx agents/content-vault/query.ts "search term" # Programmatic (from other agents) import { search } from 'content-vault/lib/search'; const results = await search("Value Path stages"); ``` --- ## Current State (Honest Assessment) **Operational and stable.** Schema v3 proven. FTS5 search sub-second. Census report available. **What works well:** - 34MB database with 1,003 episodes, 114 articles, 1,496 canonical records - FTS5 full-text search across all content - YouTube comments and live chat capture - Census report at `/mnt/d/data/content-census-report.md` - Distributed query API for downstream agents **What doesn't work:** - **No automatic sync trigger.** Database freshness depends on manual sync or background worker execution. - **No staleness alerting.** If sync fails, no notification surfaces. **What partially works:** - YouTube comment ingestion runs daily via background worker but coverage depends on API quota availability. --- ## Connections | Connected To | Direction | What Flows | |-------------|-----------|------------| | **Forge** (V) | Forge → Vault | New episode records added to database after pipeline processing. | | **Prism** (V) | Vault → Prism | Content search results feed multiplication decisions. "What topics haven't been covered?" | | **Broadcast** (V) | Broadcast → Vault | YouTube comments and engagement data flow into Vault tables. | | **Herald** (Sage) | Vault → Herald | Content context enriches session preparation. "This client asked about X — here are 3 episodes that cover it." | | **V's /daily-ops** | Vault → daily-ops | Content intelligence in morning briefing. | | **V's /show-prep** | Vault → show-prep | Episode substance and talking points from content history. | --- ## Leadership Commentary **V (COO):** Vault is the content memory. Every episode, article, and community interaction indexed and searchable in milliseconds. When `/show-prep` needs to know what we've said about a topic before, Vault answers. When Prism looks for multiplication opportunities, Vault provides the source material. The v3 schema with YouTube comments was a breakthrough — community signals are now part of our content intelligence. My focus: automated freshness monitoring so the database never silently goes stale. **Sage (CCO):** Content intelligence serves relationships. When Herald prepares Chris for a session, knowing "we discussed this topic on 3 episodes" is relationship context. When a client mentions something they saw on a show, being able to instantly find that episode shows attention. Vault makes content memory a relationship asset. **Pax (CFO):** Content is a commercial asset. 1,003 episodes and 114 articles represent a significant intellectual property investment. Vault's ability to search and cross-reference that library means we can demonstrate depth to anyone who asks. The census report is also useful for understanding content investment — how much of our portfolio is active vs. dormant. --- *Filed: 2026-03-08 | Companion: [Org Chart](../2026-03-08-ai-org-chart.html)* *Implementation: `agents/content-vault/AGENT.md`, `agents/content-vault/query.ts`* *Database: `/mnt/d/data/content-vault.db` (34MB, schema v3)* *Census: `/mnt/d/data/content-census-report.md`*