chore: scaffold project files

IDEA.md, SESSION.md, DECISIONS.md, CLAUDE.md, .gitignore.
Next session: invoke writing-plans skill against design spec.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

.gitignore

@@ -0,0 +1,9 @@
+__pycache__/
+*.pyc
+.pytest_cache/
+*.egg-info/
+venv/
+.env
+assets/
+samples/
+.backup/

CLAUDE.md

@@ -0,0 +1,25 @@
+# AI Hell — Project Instructions
+
+## What This Is
+
+Passive horror webapp — AI-generated hellscape with infinite escalation. No avatar, no interaction. Server generates assets + controls escalation, frontend composites at 60fps via WebGL. XTTS voices cloned from random non-voice samples for emergent creepiness.
+
+## Stack
+
+- **Container:** LXC on pve197 (V100 32GB, ~6.5GB VRAM)
+- **Backend:** Python 3.11+, FastAPI, uvicorn
+- **Image gen:** SDXL Turbo (~5GB VRAM)
+- **Voice gen:** XTTS v2 (~1.5GB VRAM)
+- **Frontend:** Fullscreen WebGL shader compositor, Web Audio API
+- **Reverse proxy:** Caddy at hell.sethpc.xyz
+
+## Key Files
+
+- `docs/superpowers/specs/2026-04-10-ai-hell-design.md` — full design spec (START HERE)
+- `IDEA.md` — project brief
+- `DECISIONS.md` — decision log
+- `SESSION.md` — session memory
+
+## Next Step
+
+Run the `writing-plans` skill against the design spec to create an implementation plan, then execute.

DECISIONS.md

@@ -0,0 +1,27 @@
+# DECISIONS.md — AI Hell Decision Log
+
+Project-specific decisions. For global/cross-cutting decisions, see `~/bin/DECISIONS.md`.
+
+Format: `YYYY-MM-DD: <decision> — <why>`
+
+## Architecture
+
+2026-04-10: Hybrid architecture (server brain + client rendering) — server generates assets + controls escalation, frontend composites at 60fps via WebGL. Best of both: smooth visuals, server-controlled pacing, GPU gen hidden behind asset pool
+2026-04-10: No avatar/3D model — pure environmental horror. Dropped Three.js 3D scene, GLB models, morph targets from claude-avatar
+2026-04-10: WebGL 2D shader compositor, not Three.js 3D scene — no 3D environment needed, shader effects on AI images is cheaper and more dynamic
+2026-04-10: Based on claude-avatar infra — reuse FastAPI skeleton, WebSocket, SDXL/XTTS loading, config pattern, systemd/Caddy patterns
+
+## Implementation
+
+2026-04-10: Logarithmic escalation curve (rate=0.05) with randomized delivery timing — predictable timing kills horror, stochastic expression prevents habituation
+2026-04-10: Three-layer audio (ambient drone + whisper pool + direct address) — builds dread foundation, mid-layer unease, and genuine scare moments
+2026-04-10: XTTS fed non-voice audio samples as cloning source — emergent creepiness from the model trying to "speak" as machinery, animals, static. The AI is creepy without being told to be
+2026-04-10: Pre-generated asset pool with continuous background generation — bridges GPU latency gap (SDXL 1-2s per image), no dead air in the horror experience
+2026-04-10: Liminal imagery (coherent <-> abstract) — things emerge from and dissolve back into chaos, more unsettling than purely recognizable or purely abstract
+
+## Deferred / Rejected
+
+2026-04-10: Deferred webcam integration — detect viewer's face, use it in horror. V2 feature
+2026-04-10: Deferred viewer voice cloning — mic input -> XTTS clones them back at them. V2 feature
+2026-04-10: Rejected interactive elements — passive viewing is scarier, you can't control it
+2026-04-10: Rejected server-streamed frames — GPU-bound 1-2fps too slow, client-side compositing at 60fps is better

GITEA_API.md

@@ -0,0 +1 @@
+../GITEA_API.md

IDEA.md

@@ -0,0 +1,17 @@
+# IDEA.md — AI Hell
+
+## What is this?
+
+Passive horror webapp. You open it, it slowly destroys your comfort. AI-generated imagery shifts between abstract hellscape and recognizable-but-wrong forms. Audio: ambient dread soundscape with AI-voiced whispers cloned from random non-voice samples. Escalation is infinite — never peaks, always worse, with randomized timing. No avatar, no interaction, no UI. You just watch.
+
+## Problem it solves
+
+Proof of concept for AI-driven horror experience. Explores emergent creepiness from feeding XTTS non-voice samples, SDXL horror imagery, and WebGL shader distortion. Based on claude-avatar infrastructure.
+
+## Constraints / preferences
+
+- Runs on pve197 V100 32GB (~6.5GB VRAM: SDXL 5GB + XTTS 1.5GB)
+- LXC container on pve197 (reuse CT 167 or new CT)
+- Python + FastAPI, single process
+- LAN accessible, reverse proxied via Caddy at hell.sethpc.xyz
+- Hybrid architecture: server generates assets + controls escalation, frontend composites at 60fps via WebGL

SESSION.default.md

@@ -0,0 +1 @@
+../SESSION.default.md

SESSION.md

@@ -0,0 +1,32 @@
+# SESSION.md — AI Hell
+
+## Project Summary
+
+Passive horror webapp. AI-generated hellscape with infinite escalation, randomized timing, WebGL shader compositor, XTTS voices cloned from non-voice samples. Based on claude-avatar infra.
+
+## Context Files
+
+- `~/bin/CLAUDE.md` — global homelab instructions
+- `./IDEA.md` — project brief
+- `./DECISIONS.md` — decision log
+- `./docs/superpowers/specs/2026-04-10-ai-hell-design.md` — design spec (START HERE)
+
+## Infrastructure
+
+- **Target container:** CT on pve197, V100 32GB GPU passthrough
+- **Caddy:** hell.sethpc.xyz -> CT_IP:8400
+- **Repo:** git.sethpc.xyz/Seth/ai-hell
+
+## Session Notes
+
+### Design (2026-04-10)
+
+- Brainstormed from claude-avatar baseline
+- Key decisions: no avatar (pure environmental horror), hybrid arch (server brain + client rendering), infinite asymptotic escalation with randomized timing, XTTS fed non-voice samples for emergent creepiness
+- Stack: SDXL Turbo (images) + XTTS v2 (voices) + WebGL shaders (distortion) + Web Audio (3-layer mixer)
+- Design spec written and approved
+
+## Next Steps
+
+- [ ] Write implementation plan (invoke writing-plans skill against design spec)
+- [ ] Execute plan