{% extends "_layout.html" %} {% set active = "memory" %} {% block title %}Active memory{% endblock %} {% block main %}
Compiled macros {{ kpi.compiled }} avg conf {{ "%.2f"|format(kpi.avg_macro_conf) }} · coverage {{ kpi.coverage }}
Hebbian-tuned {{ kpi.hebbian }} trigger embeddings drifted on success
Counterfactuals {{ kpi.counterfactual }} alternatives generated by REM
Schemas {{ kpi.schemas }} meta-skills above clusters

Compiled macros distilled by the dreamer

These skills bypass the LLM at wake time when the task strongly matches their trigger. Compiled from accumulated successful trajectories.

{% if compiled_skills %} {% for s in compiled_skills %} {% endfor %}
idname steps macro conf fitness s/t
{{ s.id[:8] }} {{ s.name }} {{ s.compiled_steps }} {{ "%.2f"|format(s.compiled_conf) }} {{ "%.2f"|format(s.fitness_mean) }} {{ s.successes }}/{{ s.trials }}
{% else %}

No compiled macros yet — let the library accumulate ≥5 successes per skill, then run sleep.

{% endif %}

Counterfactual lineage REM-generated alternatives

Alternative strategies the dreamer proposed for skills that kept failing. They compete against their parent for retrieval.

{% if counterfactual_skills %} {% for s in counterfactual_skills %} {% endfor %}
cf skillname parent fitness
{{ s.id[:8] }} {{ s.name }} {% if s.parent_id %} {{ s.parent_id[:8] }} {% else %} {% endif %} {{ "%.2f"|format(s.fitness_mean) }} ({{ s.successes }}/{{ s.trials }})
{% else %}

No counterfactuals yet — they only appear when a skill has ≥3 trials and fitness ≤ 0.5.

{% endif %}

Schema hierarchy meta-skills above clusters

Skills sharing a domain are clustered into a schema, a meta-skill that picks among its children.

{% if schemas %} {% for sch in schemas %}
{{ sch.name }}
{{ sch.trigger }}
{% endfor %} {% else %}

No schemas yet — they emerge when ≥3 skills share a domain.

{% endif %}

Hebbian tuning drift on success

Each successful application drifts the skill's trigger embedding toward the task that just succeeded. Skills become magnets for the kind of work they keep solving — without retraining anything.

Total tuned skills: {{ kpi.hebbian }} of {{ kpi.skills_total }}.

Forward replay pure retrieval

Before the wake loop runs, the agent projects the action sequence from past successful episodes that used the top-retrieved skill, and injects it as a ## PREDICTED PATH block in the user prompt.

Watch the live event stream (/events) for forward_replay events.

{% endblock %}