Ingest this research and triage any work to be done here #946

Closed
opened 2026-03-22 16:58:18 +00:00 by Rockachopa · 2 comments
Collaborator

Triage Review (perplexity)

PDF Contents

The attached PDF — "Timmy Time Integration Architecture: Eight Deep Dives into Real Deployment" — covers 8 architectural topics. It was written during the Veloren exploration phase but many patterns are game-agnostic and directly apply to Morrowind.

Eight Deep Dives

  1. Veloren client sidecar — WebSocket JSON-line pattern (game-specific, adapt for OpenMW Lua bridge)
  2. Agno Ollama tool calling is broken — use format param with Pydantic schemas instead (applies to all Ollama usage)
  3. MCP is the right abstraction — FastMCP at 26ms overhead per tool call (applies directly)
  4. Human + AI see each other as normal players — co-op architecture (game-specific but pattern applies)
  5. Real latency numbers — VLM bottleneck analysis, pipeline timing (superseded by Perception Bottleneck #963 which uses API-first approach)
  6. Content moderation — Llama Guard + game-context system prompts (applies directly to Morrowind mature themes)
  7. Qwen3-8B vs Hermes 3 — model selection for perception/decision/narration (partially superseded by #966 three-tier router)
  8. Split Hetzner + Mac architecture — deployment topology (applies directly)

What Still Applies Post-Morrowind Pivot

Topic Status Action
Pydantic structured output for Ollama Directly applicable Already covered by #966 (three-tier router)
FastMCP tool server Directly applicable Covered by #984 (MCP restore)
Content moderation pipeline Directly applicable NEW — needs its own issue
Split Hetzner + Mac deployment Directly applicable Referenced in #982 sprint plan
Co-op architecture ⚠️ Adapt for Morrowind Defer until after tutorial completion
Veloren sidecar pattern Superseded Replaced by OpenMW Lua bridge (#964)
VLM latency analysis Superseded Replaced by API-first perception (#963)

Recommendation

  • Do NOT close — the PDF has value as reference material
  • One net-new issue needed: content moderation pipeline (Llama Guard + game-context prompts for Morrowind)
  • Add label needs-extraction for the moderation work
  • The body "We need latency" is answered by #963 (Perception Bottleneck) which eliminates the VLM bottleneck entirely
## Triage Review (perplexity) ### PDF Contents The attached PDF — **"Timmy Time Integration Architecture: Eight Deep Dives into Real Deployment"** — covers 8 architectural topics. It was written during the Veloren exploration phase but many patterns are **game-agnostic** and directly apply to Morrowind. ### Eight Deep Dives 1. **Veloren client sidecar** — WebSocket JSON-line pattern (game-specific, adapt for OpenMW Lua bridge) 2. **Agno Ollama tool calling is broken** — use `format` param with Pydantic schemas instead (applies to all Ollama usage) 3. **MCP is the right abstraction** — FastMCP at 26ms overhead per tool call (applies directly) 4. **Human + AI see each other as normal players** — co-op architecture (game-specific but pattern applies) 5. **Real latency numbers** — VLM bottleneck analysis, pipeline timing (superseded by Perception Bottleneck #963 which uses API-first approach) 6. **Content moderation** — Llama Guard + game-context system prompts (applies directly to Morrowind mature themes) 7. **Qwen3-8B vs Hermes 3** — model selection for perception/decision/narration (partially superseded by #966 three-tier router) 8. **Split Hetzner + Mac architecture** — deployment topology (applies directly) ### What Still Applies Post-Morrowind Pivot | Topic | Status | Action | |-------|--------|--------| | Pydantic structured output for Ollama | ✅ Directly applicable | Already covered by #966 (three-tier router) | | FastMCP tool server | ✅ Directly applicable | Covered by #984 (MCP restore) | | Content moderation pipeline | ✅ Directly applicable | **NEW — needs its own issue** | | Split Hetzner + Mac deployment | ✅ Directly applicable | Referenced in #982 sprint plan | | Co-op architecture | ⚠️ Adapt for Morrowind | Defer until after tutorial completion | | Veloren sidecar pattern | ❌ Superseded | Replaced by OpenMW Lua bridge (#964) | | VLM latency analysis | ❌ Superseded | Replaced by API-first perception (#963) | ### Recommendation - **Do NOT close** — the PDF has value as reference material - One net-new issue needed: content moderation pipeline (Llama Guard + game-context prompts for Morrowind) - Add label `needs-extraction` for the moderation work - The body "We need latency" is answered by #963 (Perception Bottleneck) which eliminates the VLM bottleneck entirely
claude was assigned by Rockachopa 2026-03-22 23:31:30 +00:00
Collaborator

PR #1057 created.

What was done:

  • Ingested the 11-page research PDF into docs/research/integration-architecture-deep-dives.md — structured summary of all 8 deep dives with current issue cross-references
  • Created issue #1056 for the one net-new work item: content moderation pipeline (Llama Guard + game-context prompts for Morrowind)
  • All other research topics already covered by existing issues (#963, #964, #966, #982, #984) or deferred

The "we need latency" concern in the issue body is addressed by #963 (Perception Bottleneck) which eliminates the VLM bottleneck via API-first approach.

PR #1057 created. **What was done:** - Ingested the 11-page research PDF into `docs/research/integration-architecture-deep-dives.md` — structured summary of all 8 deep dives with current issue cross-references - Created issue #1056 for the one net-new work item: **content moderation pipeline** (Llama Guard + game-context prompts for Morrowind) - All other research topics already covered by existing issues (#963, #964, #966, #982, #984) or deferred The "we need latency" concern in the issue body is addressed by #963 (Perception Bottleneck) which eliminates the VLM bottleneck via API-first approach.
Sign in to join this conversation.
No Label
3 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Rockachopa/Timmy-time-dashboard#946