Skip to content

Add pinned MiniCPM5 and Spark GGUF models to v0.12 runtime - #43

Merged
pirumpi merged 2 commits into
mainfrom
feat/minicpm-spark-current
Sep 18, 2026
Merged

pirumpi merged 2 commits into
mainfrom
feat/minicpm-spark-current

Conversation

@pirumpi

@pirumpi pirumpi commented Sep 18, 2026

Copy link
Copy Markdown
Contributor

Summary

Based on released main/v0.12.0. Adds official pinned MiniCPM5-2B and Spark-X2.5-4B Q4_K_M GGUF models, with immutable revisions, exact file sizes and SHA-256. Bundles llama.cpp b10964 (needed for Spark) on the CLI and desktop. Archive checksums are verified before extraction, and the desktop builder detects a stale cached runtime.

Research date: 2026-09-18. This supersedes draft PRs #41 and #42, which were based on the pre-v0.12 main branch and conflicted with the released Granite and download-integrity implementation.

Validation

  • Downloaded the exact current official MiniCPM5 and Spark files and verified full SHA-256 and sizes.
  • Verified all five official b10964 platform archive hashes and extraction layouts.
  • Apple Silicon CPU/Metal basic chat passed for both models; English/Chinese tested. Existing Phi-4 Mini and GLM-4.7-Flash Q8 were regression-smoked with b10964 Metal.
  • tapioca pull cached-file integrity checks passed for both; Spark tapioca serve answered an OpenAI-compatible chat request.
  • Catalog manifest validation, go test ./..., full desktop Vitest suite and TypeScript typecheck passed locally.

Remaining qualification

Windows/Linux Vulkan behavior, low-memory measurements, long context, and structured tool use have not been qualified. This PR remains draft; its catalog guidance says so. Do not close #37/#38 until platform qualification and required review are complete. Existing Granite 4.2 remains unchanged.

Sources

Related: #37, #38.

@coderabbitai

coderabbitai Bot commented Sep 18, 2026 •

Copy link
Copy Markdown

Warning

Review limit reached

Next included review available in 48 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used the included review currently available.

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

Learn how review limits work.

Review configuration:

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Advanced

Run ID: c40aa62b-445f-4f13-8b2f-33848e8a7dfd

📥 Commits

Reviewing files that changed from the base of the PR and between efe7a55 and 47340e5.

📒 Files selected for processing (9)
  • .github/workflows/build.yml
  • catalog/catalog.json
  • catalog/catalog.json.sha256
  • desktop/scripts/prepare-runtime.test.ts
  • desktop/scripts/prepare-runtime.ts
  • docs/guides/choosing-models.md
  • internal/catalog/catalog.go
  • internal/catalog/compact_models.go
  • internal/catalog/compact_models_test.go

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.

❤️ Share

Comment @coderabbitai help to get the list of available commands.

@pirumpi
pirumpi marked this pull request as ready for review September 18, 2026 22:58
@pirumpi

pirumpi commented Sep 18, 2026

Copy link
Copy Markdown
Contributor Author

All platform bundle jobs, Go tests, and website/docs checks are green on the latest commit. Apple CPU/Metal inference and Tapioca pull/serve passed locally. Requesting review now; Windows/Linux runtime inference and low-memory qualification remain open items, so this should not be treated as a fully qualified public release solely from build success.

@pirumpi
pirumpi merged commit a0551bb into main Sep 18, 2026
9 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Add MiniCPM5-2B support via llama.cpp

1 participant