Files
obelisk/PRODUCT.md
T

34 lines
1.9 KiB
Markdown
Raw Normal View History

# Product
## Register
product
## Users
Developers and coding-agent users who need to revisit past work, understand what an agent did, verify why a decision was made, and manage durable memory without leaving the evidence behind.
## Product Purpose
Obelisk is explicit memory infrastructure for coding agents. Its core indexes local agent history into a queryable evidence layer; the desktop app is the human observability and memory-control surface for inspecting sessions, auditing evidence, managing the index, and reviewing or revoking durable memories. Success means a person can move from a remembered question to the exact historical evidence with low friction and high confidence.
## Brand Personality
Calm, exact, auditable. Obelisk should feel trustworthy around private local history, dense without being hostile, and confident without hiding uncertainty or provenance.
## Anti-references
Obelisk should not resemble an undifferentiated chat-log browser, a terminal log wall, an opaque AI-summary dashboard, or a decorative analytics product that separates conclusions from their source evidence.
## Design Principles
1. Evidence before assertion: every interpretation keeps a clear path back to the raw session record.
2. Human-readable by default: expose protocol and storage detail on demand, not as the primary reading experience.
3. Preserve uncertainty: never present inferred structure as observed execution fact.
4. Progressive density: make long sessions scannable without discarding the depth experts need.
5. Local trust is visible: controls, provenance, and failure states should reinforce that the user remains in charge of their history and memories.
## Accessibility & Inclusion
Target WCAG AA contrast for essential text and controls. Core session navigation and disclosure controls should work from the keyboard, focus must remain visible, state cannot rely on color alone, and motion should respect reduced-motion preferences.