wonderbot-v1

A clean consolidation base for the archived LLM/agent experiments.

This repo is deliberately not another fragile wrapper around a standard tokenizer pretending to be tokenizerless. Instead it separates the system into four clear layers:

1. Event codec — resonant segmentation + lossless byte encoding + feature signatures.
2. Memory — append-only, priority-ranked, searchable, non-destructive.
3. Ganglion — a clocked CA bus that gives the agent a continuously evolving internal substrate.
4. LLM backend — swappable. The default backend works with no external dependencies; optional HuggingFace support can be enabled later.

That split is the point: the old projects drifted because the “new tokenizer” was treated as if it could be dropped into a pretrained LM without retraining the representational contract. This base stops doing that.

What was salvaged conceptually

• resonant-llm → event segmentation, local-agent shell, continuous memory emphasis.
• riemann-resonance-llm → resonance framing and the wish for cognition to be organized by resonance rather than only next-token prediction.
• woflchess / claude's-neural-chess → ganglion / CA bus / continuously ticking substrate.
• wofl-brain → coordination / “brain hub” framing.

What this repo does now

• Runs immediately with no third-party dependencies.
• Supports an interactive, always-on-ish CLI agent that forms and searches memory continuously.
• Uses a replacement tokenizer architecture where it is actually sound: segmentation, salience, memory, and internal event coding.
• Keeps the LLM backend abstract so you can:
  • stay fully local and lightweight now,
  • plug in a HuggingFace backend later,
  • or replace the backend entirely with a future native event-stream model.

What this repo does not pretend to do

• It does not claim that a pretrained LM has become tokenizerless.
• It does not require camera/mic/Whisper/BLIP/TTS just to boot.
• It does not destroy memory entries when consolidating.

Quick start

python -m wonderbot.cli

Or after install:

pip install -e .
wonderbot

Type text normally. Useful commands:

• /tick 5 — advance internal time and allow spontaneous thoughts
• /state — inspect ganglion / memory stats
• /memory 10 — show top memories
• /search your query — semantic-ish memory lookup using the event codec
• /save — persist state
• /quit — exit cleanly

Optional HuggingFace backend

pip install -e .[hf]
python -m wonderbot.cli --backend hf --hf-model distilgpt2

Note: the HF backend still uses its own tokenizer internally. That is intentional. The agent contract is event-coded text and memory; the backend is only one possible renderer.

Repo layout

wonderbot/
  agent.py
  cli.py
  config.py
  event_codec.py
  ganglion.py
  llm_backends.py
  memory.py
  resonance.py
  sensors/
configs/
  default.toml
docs/
  ARCHITECTURE.md
  CONSOLIDATION_NOTES.md
  LEGACY_MAP.md
scripts/
  seed_from_legacy.py
tests/

Design choice that keeps this on-course

The event codec is the new center of gravity.

• It segments raw text into resonant events.
• It produces lossless byte IDs when exact reconstruction matters.
• It produces feature signatures for memory, salience, novelty, and internal routing.
• It can later become the input contract for a trained native event model.

That is the bridge from the old “tokenizer replacement” idea to something that actually survives contact with reality.