Configuration

Brain configuration lives in ~/.brain/index.json under the config key. All settings have sensible defaults set during /brain:init.

Full Configuration

{
  "config": {
    "max_depth": 6,
    "consolidation_threshold": 0.3,
    "decay_check_interval_days": 7,
    "strength_boost_on_recall": 0.05,
    "auto_consolidate": true,
    "propagation_window_days": 7,
    "association_config": {
      "co_retrieval_boost": 0.10,
      "link_decay_rate": 0.998,
      "link_prune_threshold": 0.05,
      "spreading_activation_depth": 2,
      "spreading_activation_decay": 0.5
    }
  }
}

Core Settings

Setting	Type	Default	Description
`max_depth`	integer	6	Maximum directory nesting depth for memory placement. Higher values allow more specific categorization but deeper paths.
`consolidation_threshold`	float	0.3	Strength below which memories become candidates for consolidation. Lower values wait longer before consolidating.
`decay_check_interval_days`	integer	7	How often (in days) to suggest decay maintenance. The agent will remind you to run `/brain:sleep` after this many days.
`strength_boost_on_recall`	float	0.05	Base strength increase per recall event, before spacing and diminishing adjustments are applied.
`auto_consolidate`	boolean	true	Whether to suggest consolidation when candidates are found during recall. Set to false to only consolidate via explicit `/brain:consolidate` or `/brain:sleep`.
`propagation_window_days`	integer	7	How far back (in days) to look for recent memories during Knowledge Propagation in sleep. Recent memories within this window are evaluated against the existing hierarchy.

Association Config

Settings for the Associative Network:

Setting	Type	Default	Description
`co_retrieval_boost`	float	0.10	Hebbian reinforcement increment for co-retrieved memories. When two memories are recalled together, their edge weight increases by this factor times the remaining distance to 1.0.
`link_decay_rate`	float	0.998	Daily decay factor for association edge weights. An unused link retains 99.8% of its weight each day.
`link_prune_threshold`	float	0.05	Minimum weight before an association link is pruned during sleep. Links below this weight are removed.
`spreading_activation_depth`	integer	2	Maximum number of hops for spreading activation during recall. Higher values surface more distantly related memories.
`spreading_activation_decay`	float	0.5	Activation decay factor per hop. At depth 2 with decay 0.5, activation at the second hop is 25% of the original.

Tuning Guide

Making Memories Last Longer

To slow down decay and keep memories active longer:

Increase consolidation_threshold (e.g., 0.4) — memories are consolidated sooner, preserving knowledge before it fades too much
Decrease decay_check_interval_days (e.g., 3) — more frequent reminders to run sleep

More Aggressive Association Discovery

To surface more related memories during recall:

Increase spreading_activation_depth to 3 — allows 3-hop traversal (more distant connections)
Increase co_retrieval_boost (e.g., 0.15) — faster link strengthening through Hebbian learning
Decrease link_prune_threshold (e.g., 0.03) — keeps weaker links alive longer

More Conservative Association

To reduce noise and only surface strongly related memories:

Decrease spreading_activation_depth to 1 — only direct neighbors
Increase spreading_activation_decay (e.g., 0.7) — faster activation falloff per hop
Increase link_prune_threshold (e.g., 0.10) — more aggressively prune weak links

Deeper Hierarchy

To allow more specific categorization:

Increase max_depth (e.g., 8) — allows deeper nesting for very specific memories
Note: deeper hierarchies can make manual browsing harder

Editing Configuration

You can edit the configuration directly in ~/.brain/index.json, or ask the agent to update it:

You: Change the consolidation threshold to 0.4

Agent: Updated consolidation_threshold from 0.3 to 0.4 in ~/.brain/index.json.
Memories with strength between 0.3 and 0.4 will no longer be consolidation candidates.

Warning

Be careful when editing index.json directly — it contains the full memory inventory alongside the config. Make a backup before making manual edits.

Tip

The defaults work well for most users. Consider tuning only after you have been using Brain Memory for a few weeks and have a feel for how your memories behave.