feat(4090-profile): update model inventory to 8b variants + add SPARK peer - #1442
Conversation
|
Warning Rate limit exceeded
You’ve run out of usage credits. Purchase more in the billing tab. ⌛ How to resolve this issue?After the wait time has elapsed, a review can be triggered using the We recommend that you space out your commits to avoid hitting the rate limit. 🚦 How do rate limits work?CodeRabbit enforces hourly rate limits for each developer per organization. Our paid plans have higher rate limits than the trial, open-source and free plans. In all cases, we re-allow further reviews after a brief timeout. Please see our FAQ for further information. ℹ️ Review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (2)
📝 WalkthroughWalkthroughThe PR upgrades the laptop-4090 profile to qwen3-embedding:8b with recalculated VRAM budgets, adds a pmoves-spark fleet_inference peer, updates NATS embedding capability, extends Phase 4 mesh checks to include SPARK, and inserts Phase 4b Ollama inventory health checks. ChangesLaptop 4090 Profile & Operational Health
Estimated code review effort🎯 2 (Simple) | ⏱️ ~12 minutes Poem
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 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: ab8609a0dd
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".
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 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 `@pmoves/config/profiles/laptop-4090.yaml`:
- Around line 148-156: The profile defines a new agent "pmoves-spark" but the
canonical agent registry (agent_registry.yaml) is missing that entry; add a
registry entry named pmoves-spark in pmoves/config/agent_registry.yaml with the
same canonical fields as the profile (role: fleet_inference, tailscale_host:
pmoves-spark, hardware: "GB10 Grace-Blackwell 128GB unified — ARM64", services:
[ollama:11434, nim:8200] or equivalent list form, access_via: [tailscale, nats],
and notes: "Primary 120B+ inference node...") so the registry remains the single
source of truth and matches the laptop-4090.yaml pmoves-spark definition.
In `@pmoves/configs/tac_trees/node-4090-laptop.tac.yaml`:
- Around line 192-197: Summary: The ping check uses the old hostname
pmoves-gb10-spark causing false negatives; update it to pmoves-spark. Edit the
node with id n4090.mesh.reach-spark: in the action.command field replace
"pmoves-gb10-spark" with "pmoves-spark", and update the action.expect string to
match the new host (e.g., "Pong from pmoves-spark via direct connection") so the
command and expected output stay consistent.
🪄 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: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
Run ID: 76296877-0bb8-4f22-a1d6-522201070e71
📒 Files selected for processing (2)
pmoves/config/profiles/laptop-4090.yamlpmoves/configs/tac_trees/node-4090-laptop.tac.yaml
b29ce52 to
fdce045
Compare
… peer - laptop-4090.yaml: swap qwen3-embedding:4b → :8b (4700MB, installed), swap qwen3-vl:4b → :8b in multimodal_minimal group; recalculate all co-residency VRAM budgets to match actual 8b footprint; update nats_announce.caps.embeddings to qwen3-embed-8b - laptop-4090.yaml: add pmoves-gb10-spark as fleet_inference peer (GB10 Grace-Blackwell 128GB — primary 120B+ inference node) - node-4090-laptop.tac.yaml: add Phase 4b Ollama health phase (service health, embedding resident check, coder check, stale model audit); extend Phase 4 mesh with SPARK ping check and Tailscale SSH check Operator actions required (not in code): tailscale set --ssh # enable SPARK → 4090 SSH ollama rm <14 stale models> # free ~67GB ollama pull qwen3-coder:30b qwen3.5:9b qwen3.5:4b lfm2:24b Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
cross_node.peers key and tailscale_host were using pmoves-gb10-spark which doesn't resolve in Tailscale. Actual registered device name is pmoves-spark (verified via tailscale status 2026-05-10). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…idation - Fix SPARK ping check: pmoves-gb10-spark → pmoves-spark (hostname fix) - Fix SSH enabled check: Self.Capabilities reflects ACL grants not SSH state; use tailscale debug features | grep ssh with TailscaleSSHEnabled JSON fallback for accurate indicator - Fix Ollama running check: count check can pass with wrong models; replace with set-intersection against required model names so MISSING:<name> surfaces when any profile model is absent Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
fdce045 to
4a4efec
Compare
1 participant
Summary
laptop-4090.yamlto match actual Ollama inventory:qwen3-embedding:8b(4.7GB) andqwen3-vl:8b(6.1GB) are installed; original profile specified 4b variantspmoves-spark(DGX Spark GB10 Grace-Blackwell 128GB) as a fleet_inference cross-node peer — primary 120B+ model hostnode-4090-laptop.tac.yaml: service health, embedding residency, coder availability, stale model auditRelease Note
Profile update for
laptop-4090.yaml(suit-class change): acknowledges 8b embedding/VL model variants already installed, recalculates VRAM co-residency budgets, adds DGX Spark (pmoves-spark) as fleet_inference cross-node peer. TAC tree extended with Ollama health + SPARK connectivity verification phases.Operator Actions Required (not in code)
```bash
Enable Tailscale SSH on 4090 laptop (so SPARK can SSH without Windows password)
tailscale set --ssh
Tag SPARK in Tailscale admin: add tag:gpu to pmoves-spark
Delete 14 stale models (~67GB freed)
ollama rm nomic-embed-text embeddinggemma gemma3:latest gemma3:12b gemma2:latest
ollama rm llama3.2:latest llama3.2-vision qwen2.5:7b qwen2.5vl:7b
ollama rm deepseek-r1:latest deepseek-coder-v2:latest phi4:latest
ollama rm jina-embeddings-v2-base-code functiongemma
Pull 4 profile-expected models (~42GB)
ollama pull qwen3-coder:30b # ~18GB coding fallback
ollama pull qwen3.5:9b # ~6GB general fallback
ollama pull qwen3.5:4b # ~3GB multimodal minimal
ollama pull lfm2:24b # ~15GB research focus
```
Emperor-CHIT-Humility Disclosure
Have: laptop-4090.yaml full read, dgx-spark profile, TAC tree current state, Tailscale ACL policy, tailscale status verified (SPARK hostname confirmed as pmoves-spark)
Missing: Ollama running state verified 2026-05-10 (qwen3-embedding:8b and qwen3-vl:8b confirmed present)
🤖 Generated with Claude Code
Summary by CodeRabbit