CognitiveMiddleware

An invisible, file-native cognitive layer for AI-assisted long-form drafting and character runtime.

Body-first psychology, dual-aspect psyche (wound ↔ gift), realm-aware somatics — running off-page so manuscripts and RP stay clean of system jargon.

---

What it is

CognitiveMiddleware is a psychological / physical runtime for characters, plus thin application layers:

Layer	File	Job
Cognitive Pipeline	Framework/CognitivePipeline.md	Mind-body simulation → Feels / Thinks / Says / Does
Book Writing Layer	Framework/Main.md	Manuscript prose, style locks, ledgers
Roleplay Engine	Simulator/CharacterRuntime.md	Interactive chat host, OOC commands
Rules & somatics	Rules_Index.md, Psychology/realm_data.yaml	Hard bans + body catalogs
State	Characters/[slug]_log.yaml + live snapshot schema	Durable vs per-tick state

It targets common AI fiction defects: therapy-speak, perfect recall, symmetric dialogue, filler, and framework jargon leakage.

---

Core pillars

Pillar	Why it matters
Body before insight	Somatic tells precede labeled emotion and dialogue
Dual-aspect psyche	Wound and gift modes — not trauma-only engines
4-channel output	Structured intent vector for any downstream app
Durable + live state	Logs for continuity; psychosomatic snapshot per tick
100% off-page matrix	Rules + linter keep system terms out of prose

---

System architecture

                 Cognitive Pipeline
          (psych / physical character runtime)
             Feels · Thinks · Says · Does
            + psychosomatic_state schema
                        │
          ┌─────────────┴─────────────┐
          ▼                           ▼
   Book Writing Layer           Roleplay Engine
      (Main.md)               (CharacterRuntime.md)
   manuscript prose              live chat RP

State:

• Card (Characters/[slug].md) — build identity (immutable defaults)
• Log (Characters/[slug]_log.yaml) — durable runtime evolution
• Live snapshot — matches Framework/Schemas/psychosomatic_state.json each tick

Module system: Downstream applications register injectors in Framework/Modules.md to enter the cognitive loop at defined hooks (pre_somatic, affect_filter, pre_arbitration, post_vector, app_render, on_commit). Core pipeline always runs; modules are subordinate and cannot override Rules_Index, prism law, or age invariants. Core ships with an empty active module registry.

---

Quick start (drafting)

1. Load Framework/Main.md + CognitivePipeline.md + Rules_Index.md + Psychology/realm_data.yaml
2. Load on-scene character cards and their _log.yaml overlays
3. Run ledger integrity pass; execute movement brief through the pipeline
4. Render clean prose; on approval, commit durable fields to the log

Quick start (RP)

1. Paste Simulator/CharacterRuntime.md into a chat host
2. Load a character pack (card + log)
3. Play; use /state, /save, /adult (host depiction ceiling toggle) as needed

Deploy to other book folders

From this repository:

python scripts/run.py deploy              # interactive
python scripts/run.py deploy MyBookName   # named target under parent dir

Deploy ships the pipeline, schema, Main, Rules, realm data, simulator, scripts, and mechanics craft files.

---

License & disclaimer

• Software utilities (.py): MIT
• Framework specs / markdown: CC BY-SA 4.0 — see LICENSE.md
• Compliance and 18+ terms: DISCLAIMER.md

Author-local cast files and private simulator sessions stay private (see LICENSE §3 and .gitignore).