/dm:save-knowledge

Purpose

Save brand knowledge to the persistent memory layer (Pinecone or Qdrant vector database) for semantic retrieval in future sessions. Stores campaign learnings, competitive intelligence, brand guidelines, and performance insights with proper metadata tagging so that valuable knowledge is never lost between sessions. Every stored item is content-hashed for deduplication, tagged with brand context, and indexed for natural language search — turning ad-hoc learnings into durable institutional memory that every agent can draw from. Designed for targeted, intentional knowledge capture — for bulk session syncing, use /dm:sync-memory instead.

Input Required

The user must provide (or will be prompted for):

• Content to store: The knowledge to save — can be plain text typed directly, a reference to content in the current conversation (e.g., "save that email analysis we just did"), structured data from a campaign report or audit, or a URL to external research. Content is stored as-is with optional summarization for the index entry
• Content type: One of: guideline (brand rules, voice standards, style restrictions), campaign-learning (what worked or failed in a campaign with supporting evidence), competitive-intel (competitor findings, positioning, pricing, strategy moves), performance-insight (metrics, benchmarks, trends, statistical patterns), or brand-asset (approved copy, templates, creative references, messaging frameworks)
• Tags: Descriptive tags for filtered retrieval — e.g., "email", "q4-2025", "subject-lines", "audience-millennials", "paid-social", "black-friday". If not provided, auto-suggested based on content analysis using brand context, industry taxonomy, and channel detection. Multiple tags encouraged for richer retrieval
• Source context: Where this knowledge originated — current session analysis, imported report, campaign retrospective, external research, competitor monitoring, or team input. Used for provenance tracking, credibility weighting during retrieval, and audit trail compliance
• Priority (optional): high (surface this knowledge proactively in relevant contexts), normal (standard retrieval weight), or low (archive-grade, retrieve only on direct queries). Default is normal
• Expiration (optional): Date after which this knowledge should be flagged as potentially stale — useful for time-sensitive competitive intel, seasonal campaign data, or pricing information that changes quarterly. No default (knowledge persists indefinitely unless expired)
• Related entries (optional): References to existing stored knowledge this entry connects to — enables knowledge graph linking and richer cross-reference retrieval

Process