Skip to content

configuration.md

Latest commit

 

History

History
116 lines (81 loc) · 4.31 KB

configuration.md

File metadata and controls

116 lines (81 loc) · 4.31 KB
Raw

Configuration

All user-specific configuration lives in ~/.agents-neuron/ (outside the repo). This keeps the skill reusable and shareable.

Config files

File Purpose
~/.agents-neuron/config.yaml Wiki vault path, user vaults to scan, and settings
~/.agents-neuron/filter-identity.md Your identity prompt: who this wiki is for and what matters
~/.agents-neuron/query-log.md Query history, used by neuron filter evolve
~/.agents-neuron/last-scan Timestamp of last neuron scan run
~/.agents-neuron/last-ingest Timestamp of last neuron ingest run (drives the daily auto-ingest TTL)
~/.agents-neuron/last-evolve Timestamp of last neuron filter evolve pass (drives the weekly auto-evolve TTL)

make install copies example templates from schema/*.example.* into ~/.agents-neuron/ on first run. Existing files are never overwritten.

config.yaml

# The vault where wiki pages and sources are written
vault_path: "~/path/to/your/wiki-vault"

# Your personal note vaults that neuron scan reads from
user_vaults:
  - "~/path/to/your/main-notes-vault"
  - "~/path/to/another-vault"   # add as many as needed

# Customizable folder names (defaults shown)
wiki_folder: "Agents-Neuron"
sources_folder: "Neuron-Sources"

# Page types and scoring
page_types:
  - entity
  - concept
  - topic
  - recipe
  - comparison

tag_prefix: "agents-neuron"
min_relevance_score: 0.4

# Scheduled maintenance (runs automatically after `neuron add`, when due)
auto_ingest_ttl_hours: 24   # ingest at most once per day
auto_evolve_ttl_days: 7     # filter evolve at most once per week

Scheduled maintenance after neuron add

neuron add keeps the wiki current without you remembering to run things. After each import it makes one cheap TTL check (scripts/auto-maintenance.sh check) and then:

  • Runs neuron ingest if the last ingest was more than auto_ingest_ttl_hours ago (default 24h), so pending sources get turned into wiki pages at least once a day, not on every add.
  • Runs a neuron filter evolve pass if the last one was more than auto_evolve_ttl_days ago (default 7d), so the filter is tuned about once a week.

If both are due, ingest runs first, then evolve. The timestamps live in ~/.agents-neuron/last-ingest and ~/.agents-neuron/last-evolve; running either command manually also resets its timer. Set the TTLs higher to run these less often, or lower to run them more aggressively.

Identity filter

Generating your filter-identity.md

The fastest way to seed a personalized filter-identity.md is to ask an LLM what it already knows about you from your conversation history. Paste this prompt into ChatGPT (or any LLM you've been talking to):

Based on everything you know about me from our conversations (my job, projects,
interests, goals, and the topics I regularly ask about), please generate a
filter-identity.md file for my personal Agents Neuron wiki.

Be specific and honest. The more precise the identity, the sharper the filtering.

Structure:

# Identity Filter

## Who is this wiki for?
[2-3 sentences: my role, primary domains, key interests]

## What matters (scoring dimensions)

| Dimension | Weight | Description |
|-----------|--------|-------------|
[6-9 rows tailored to me, weights summing to 1.0]

## Minimum relevance threshold
Score: **0.4** out of 1.0

## Scoring instructions
[Brief guidance with a concrete example using my domains]

## Evolution log
*No changes yet.*

Make the dimensions specific to what I actually care about. Weights should reflect
how central each domain is to my life and work.

Save the output to ~/.agents-neuron/filter-identity.md. Alternatively, neuron bootstrap will draft it automatically from your vault structure.

How the filter works

The filter defines scoring dimensions and weights. Each source is scored 0-1 against each dimension; the weighted sum is compared against a minimum threshold (default: 0.4). Sources below the threshold are marked ingested but don't get wiki pages. They stay in Neuron-Sources/ for reference without cluttering the wiki.

Run neuron filter evolve periodically to tune weights based on what you actually query and link to. All filter changes require your approval and apply to future ingests only. Existing pages are never retroactively removed.