docs(roadmap): OpenSpec 溯源與 §9 NVIDIA Kit／WebRTC 語意對齊 - #22
Conversation
Co-authored-by: Cursor <cursoragent@cursor.com>
📝 WalkthroughWalkthroughThis PR updates the AI-BIM governance SaaS roadmap document with May 8 OpenSpec provenance corrections, NVIDIA/Omniverse terminology standardization, and revised hardware capacity guidance. Key changes include establishing new §9 terminology alignment for kit.exe/Kit Application Instance/Multi-Kit semantics, clarifying OpenSpec spec-to-archive provenance mapping, and updating Phase 4/5 and Tier A capacity tables to reflect the corrected vocabulary. ChangesSaaS Roadmap Documentation Updates
Estimated code review effort🎯 2 (Simple) | ⏱️ ~10 minutes Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 inconclusive)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: e06e15e243
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
|
||
| → 可容納並發 6-9 個 review session、24+ 個 collaboration viewer | ||
| → 「6-9」指的是 **6-9 條可並行的獨立 GPU streaming capacity（每台 Kit 進程一條）**； | ||
| collaboration viewer 數可以 **大於** Kit 進程數（多人連 **同一** Kit／同一 primary stream 時，WebRTC 為 **多 PeerConnection → 同一 signaling endpoint**，見 §9.0-B）。 |
There was a problem hiding this comment.
This capacity note depends on multiple viewers sharing one Kit primary stream, but the referenced bim-streaming-server/SYSTEM_DESIGN.md still defines the sizing model as One Kit process = one stage = one user session and explicitly says there is No multi-user-per-session without a separate session router. Unless that router/support is actually in scope, using viewer counts greater than Kit process counts here will understate GPU/Kit capacity needed for collaborative reviews.
Useful? React with 👍 / 👎.
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@docs/plans/AI-BIM-governance-saas-roadmap-2026-05.md`:
- Line 671: The capacity estimate for "同時獨立 Kit 進程（≈ Kit Application Instance）"
is inconsistent: replace the current "6-9 個（24 GB ÷ 3 GB）" entry with the
correct result of the formula ("8 個（24 GB ÷ 3 GB）") and similarly update the
other occurrences referenced (around the same paragraph at lines ~689-700 and
749) so all mentions of the 24GB estimate match the formula and any reserve
assumptions; locate the table row that contains the header string "同時獨立 Kit 進程（≈
Kit Application Instance）" and the numeric cell "6-9 個（24 GB ÷ 3 GB）" and change
the text to "8 個（24 GB ÷ 3 GB）".
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: 26e72823-519a-42d0-b23f-f6d39283a94e
📒 Files selected for processing (1)
docs/plans/AI-BIM-governance-saas-roadmap-2026-05.md
| |---|---|---|---| | ||
| | **GPU** | RTX 4060 Ti 8 GB | **RTX 4090 / 5090 24-32 GB** | RTX 6000 Ada 48 GB / RTX Pro 6000 96 GB | | ||
| | **同時 Kit instance** | 1（撞牆） | **6-9 個**（24 GB ÷ 3 GB） | 14-30 個（48-96 GB） | | ||
| | **同時獨立 Kit 進程（≈ Kit Application Instance）** | 1（撞牆） | **6-9 個**（24 GB ÷ 3 GB） | 14-30 個（48-96 GB） | |
There was a problem hiding this comment.
請統一 24GB 容量估算，6-9 與文內公式不一致。
這幾處把 24GB 寫成可到 9 個獨立 Kit 進程，和本文件同段公式（含 reserve）不一致，會高估容量規劃。
✏️ 建議修正（示例）
-| **同時獨立 Kit 進程（≈ Kit Application Instance）** | 1（撞牆） | **6-9 個**（24 GB ÷ 3 GB） | 14-30 個（48-96 GB） |
+| **同時獨立 Kit 進程（≈ Kit Application Instance）** | 1（撞牆） | **6-8 個**（24 GB 理論上限約 8；建議預留餘裕） | 14-30 個（48-96 GB） |
-Kit Streaming（GPU） : 6-9 個獨立 Kit 進程／容器（各 ≈ 1 × Kit Application Instance；假設每進程 1 × stage、per-slot 3 GB VRAM cap）
+Kit Streaming（GPU） : 6-8 個獨立 Kit 進程／容器（各 ≈ 1 × Kit Application Instance；假設每進程 1 × stage、per-slot 3 GB VRAM cap）
-→ 「6-9」指的是 **6-9 條可並行的獨立 GPU streaming capacity（每台 Kit 進程一條）**；
+→ 「6-8」指的是 **6-8 條可並行的獨立 GPU streaming capacity（每台 Kit 進程一條）**；
-→ 結論：**24 GB 級開發機**可承擔 Phase 4 約 **6-9 個獨立 Kit 進程**並發（外加 spectator／AOV 會吃掉額外 encoder／頻寬／VRAM），並支援 Phase 5 多數能力的 dev 驗證。
+→ 結論：**24 GB 級開發機**可承擔 Phase 4 約 **6-8 個獨立 Kit 進程**並發（外加 spectator／AOV 會吃掉額外 encoder／頻寬／VRAM），並支援 Phase 5 多數能力的 dev 驗證。Also applies to: 689-700, 749-749
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@docs/plans/AI-BIM-governance-saas-roadmap-2026-05.md` at line 671, The
capacity estimate for "同時獨立 Kit 進程（≈ Kit Application Instance）" is inconsistent:
replace the current "6-9 個（24 GB ÷ 3 GB）" entry with the correct result of the
formula ("8 個（24 GB ÷ 3 GB）") and similarly update the other occurrences
referenced (around the same paragraph at lines ~689-700 and 749) so all mentions
of the 24GB estimate match the formula and any reserve assumptions; locate the
table row that contains the header string "同時獨立 Kit 進程（≈ Kit Application
Instance）" and the numeric cell "6-9 個（24 GB ÷ 3 GB）" and change the text to "8
個（24 GB ÷ 3 GB）".
There was a problem hiding this comment.
Pull request overview
This PR updates the May 2026 SaaS roadmap planning document to improve OpenSpec traceability (specs vs archived changes) and to tighten the terminology alignment for NVIDIA Omniverse Kit/OVAS and WebRTC capacity planning assumptions.
Changes:
- Clarifies the roadmap baseline (
mainHEAD) and explicitly points to OpenSpec sources of truth (openspec/specs/andopenspec/changes/archive/), adding a new §1.4 mapping table. - Rewrites/extends §9 with a new §9.0 terminology section to distinguish Kit OS processes/instances vs WebRTC endpoints (primary/spectator/AOV) and connects this to capacity estimates.
- Adds an explicit cross-reference note at the start of §11.4 to explain how §9 and §11.4 split responsibilities/definitions.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| | `web-viewer-sample/` | 5173 | Browser client + Demo Control Panel | OK | | ||
|
|
||
| ### 1.2 已歸檔的 9 個 OpenSpec specs | ||
| ### 1.2 已歸檔的 OpenSpec specs（權威：`openspec/specs/`） |
摘要
驗證
相關
Made with Cursor
Summary by CodeRabbit
Documentation