Ngenaire Docs
Log inSign up free

Project Memory

Project Memory is a durable, per-project note store for the facts, constraints, and decisions that the AI Assistant should carry across conversations. Instead of rebuilding context every session, you record it once here — and the assistant reads it when a task depends on it.

Sidebar location: Track → Project Memory (/projects/memory).

What a note contains

  • Category — what kind of fact this is (see below).
  • Title — a one-line summary.
  • Body — the full fact, including the why when it is not obvious.

Categories

Pick the category that fits how the note will be used later:

  • Decision — a choice the team has already made. ("We chose protocol X over Y for the telemetry link.")
  • Constraint — a hard limit the assistant must respect. ("Power budget is 18 W max in all modes.")
  • Preference — a soft convention. ("This project uses MIL-STD-498 section headings.")
  • Stakeholder — who cares about what. ("The customer reviewer dislikes passive voice in the SOW.")
  • Reference — a pointer to an external resource.
  • Other — anything else worth carrying forward.

Walkthrough

  1. Open Track → Project Memory.
  2. Add a note: pick a category, write a clear title, and put the fact in the body. Lead with the what; add the why when it is not obvious.
  3. Keep notes current — edit a note when the underlying fact changes, and delete notes that are no longer true. Outdated memory is worse than none.

How the AI Assistant uses memory

  • Reading. The assistant reads these notes for context when a task plausibly depends on prior decisions — for example, drafting a new artifact or making a design call.
  • Remembering. When you state a decision, constraint, preference, or stakeholder concern in chat, the assistant can capture it here as a new note. You can also ask it to remember something explicitly.
  • Scope. Notes belong to the active project only — they do not leak across projects.

AI Assistant prompts

Remember that we have decided to use Kerberos for service-to-service auth on this project.

Summarize the current project memory and flag any notes that look stale.

Draft the next requirement using the constraints in project memory.

Tips

  • Lead with the fact, then the why. The first line is the headline the assistant keys on.
  • Capture decisions when they happen. Memory is most useful while the rationale is fresh.
  • Prune. A curated set of high-value notes beats a pile of stale ones.

Related