Subconscious Node Setup

The Subconscious Node: Setup & Operations Guide

Architectural Intent

A.I.M. uses a Distributed Cognitive Architecture. To keep frontline dev agents lean, fast, and focused entirely on code generation (The Cortex), the heavy computational lifting of memory refinement and vector chunking is offloaded to a dedicated secondary machine: The Subconscious Node (The Brain PC).

This machine is responsible for:

1. Ingesting raw session transcripts generated by the frontline dev agents.
2. Crunching the 5-Tier Memory Pipeline (aim memory).
3. Chunking and embedding the FTS5/Vector Engram DB (aim index).
4. Mirroring the state to an Obsidian Vault for human review (obsidian_sync.py).
5. Publishing the refined state back to the Git repository via Sovereign Sync for the frontline agents to inherit.

---

Initial Setup (One-Time Configuration)

1. Establish the Baseline

Clone the A.I.M. repository onto the Brain PC and run the initializer.

git clone https://github.com/BrianV1981/aim.git
cd aim
aim init

2. Configure the LLM for Heavy Lifting

The Brain PC performs high-volume context crunching. Ensure it is configured with a provider optimized for long-context reasoning (e.g., Gemini 1.5 Pro, Claude 3.5 Sonnet, or a local Ollama model if hardware permits).

aim tui

Navigate to the AI Provider section and ensure the primary model is set.

3. Wire the Obsidian Vault (The Sovereign Mirror)

To use Obsidian as the visual dashboard for the Brain PC:

1. Create an empty folder on the Brain PC to act as the vault (e.g., ~/Desktop/AIM_Vault).
2. Run aim tui.
3. Set the obsidian_vault_path to the absolute path of that folder.
4. Open the Obsidian app and open that folder as a Vault.

---

Operational Loop (The Subconscious Cycle)

The Brain PC operates in a continuous loop: Pull -> Process -> Review -> Push.

Step 1: Ingest Raw Momentum (Pull)

Pull the latest raw transcripts and git commits generated by the frontline dev agents.

aim update

(This safely pulls from Git, ingests Sovereign Sync data, and prepares the workspace).

Step 2: Trigger the Subconscious (Process)

Run the memory distillation pipeline. This crunches the raw json logs into narrative summaries, generates Delta Ledgers, and proposes new MEMORY.md states.

aim memory

Step 3: Human Review & Ingestion (Review)

1. Open your Obsidian Vault.
2. Review the generated continuity pulses and memory proposals.
3. If you need to surgically correct a hallucination or steer the architecture, edit the Markdown files directly in Obsidian.
4. If you made edits, immediately run the ingest command to pull them back into A.I.M. and re-index the Engram DB:

aim ingest

Step 4: Finalize & Distribute (Push)

If you are satisfied with the memory proposals, commit them to durable memory and push the refined state back to the central repository.

aim commit
aim push "Chore: Automated Subconscious Memory Refinement"

(This triggers obsidian_sync.py to finalize the vault view, translates the Engram DB via Sovereign Sync, and pushes the finalized architecture up to GitHub).

---

Autonomous Operation (Optional)

If you wish for the Brain PC to run entirely headless without manual Obsidian intervention, you can activate the autonomous heartbeat.

aim daemon start

The daemon will run in the background, automatically tracking time, firing the memory tiers, and managing the state. You will only need to periodically run aim push to distribute its findings back to the frontline agents.