Demo

August 23, 2026 ยท View on GitHub

This folder holds the reusable terminal demo source for Agent Memory Bridge.

The current terminal demo shows:

  • a small durable memory bundle shaped like belief -> concept-note -> procedure
  • one coordination signal
  • claim_signal
  • extend_signal_lease
  • ack_signal
  • a later recall_first(...) moment where the bridge surfaces useful task memory
  • a small benchmark snapshot

There is also a lightweight before/after story source for the "viral moment": without Agent Memory Bridge, the user has to re-teach a repo gotcha; with Agent Memory Bridge, the next session recalls the gotcha before touching code.

The checked-in demo is meant to show both coordination state and a later "the agent remembered something useful" retrieval moment, while still matching the current benchmark snapshot. Historical v0.5 demo assets are kept separately and are not the primary public demo.

The v0.30 Project Knowledge Activation adoption demo is separate at v030-project-knowledge/. It is a reproducible four-stage local proof of a commit-bound repository WHAT fact plus a durable explicit project-decision WHY written and read through fresh MCP stdio processes. Code tells the derived, rebuildable WHAT; conversations teach the WHY; inspect reports only the supported repository/provenance boundary. It does not replace or regenerate the historical terminal demos.

Files:

  • before-after-gotcha.cast.md as a text-only source transcript for the before/after gotcha story
  • scripts/demo_terminal.py for the terminal flow itself
  • scripts/build_demo_cast.py to regenerate the asciicast
  • terminal-demo.cast as the canonical recorded source
  • terminal-demo.tape as an optional VHS source
  • v030-project-knowledge/README.md and v030-project-knowledge/run_demo.py for the v0.30 project-knowledge activation proof

To use the before/after gotcha story:

python -c "from pathlib import Path; print(Path('examples/demo/before-after-gotcha.cast.md').read_text())"

It is source material only. If you want to turn it into a shareable recording, save the generated asciicast as examples/demo/before-after-gotcha.cast; if you render a GIF from that cast, use:

agg ./examples/demo/before-after-gotcha.cast ./examples/demo/before-after-gotcha.gif

The checked-in text source is enough for readers; generated binaries are optional release assets.

To regenerate the cast:

python ./scripts/build_demo_cast.py

The checked-in cast uses a scrubbed placeholder prompt path so the public demo does not expose a maintainer workstation path.

To render the cast to GIF after installing agg:

agg ./examples/demo/terminal-demo.cast ./examples/demo/terminal-demo.gif