Memory Model (What You Should Expect To See)

Memory is not a repo dump. It’s a curated stream of high-signal items (decisions, conventions, patterns) that your assistant stores and later retrieves semantically.

What Memory Is Not

veriova does not automatically ingest your entire codebase, README, or git history. You only see what you (or your assistant) explicitly store.

What Memory Is

Each item is a small, structured unit with a type, title, and content. Examples:

examples

decision: "Use PostgreSQL with pgvector"
convention: "All API routes are /api/v1/*"
pattern: "Semantic search → keyword fallback"
lesson: "Cert provisioning needs DNS apex A record"
snippet: "nginx proxy headers"
reference: "OpenAPI schema URL"

Store Something (API)

terminal

curl -X POST \
  -H "Authorization: Bearer cv_..." \
  -H "Content-Type: application/json" \
  -d '{
    "project_slug": "my-app",
    "items": [{
      "type": "decision",
      "title": "Use PostgreSQL",
      "content": "Chose PostgreSQL for pgvector support.",
      "importance": 8,
      "tags": ["database"]
    }]
  }' \
  https://api.veriova.com/api/v1/memory/candidates

List + Search

terminal

curl -H "Authorization: Bearer cv_..." \
  "https://api.veriova.com/api/v1/memory?project=my-app"

curl -X POST \
  -H "Authorization: Bearer cv_..." \
  -H "Content-Type: application/json" \
  -d '{"project_slug":"my-app","query":"database choice","k":5}' \
  https://api.veriova.com/api/v1/memory/search

MCP Usage

When MCP is configured, your assistant stores and retrieves memories via tool calls (not by scraping your repository).

terminal

npx veriova-mcp@latest setup