feat(local-models): expand catalog with Hugging Face fallback and opt-in scanning

[AnishSarkar22]

Description

• Expand available models via hf.co/<repo>[:<quant>] fallback pulls when Ollama's registry lacks a model — ~91 → 1,400+ usable models.
• Opt-in hardware scanning: persistent fingerprinted cache skips rescans; curated + installed models show instantly, "Scan hardware" is explicit.
• Split catalog into curated (staff picks, always visible) and explore (scan-derived, ranked by fit).
• Fix quant trust: only tag pulls with best_quant when scored against llama_cpp; untrusted quants get an untagged pull + downgraded "Marginal" fit.
• Fix :latest mismatch: untagged fallback pulls now request :latest explicitly, matching what Ollama actually stores — fixes broken install verification and "already installed" detection.
• Fix display names: clean provider/repo name shown consistently (Chat summary, model picker, catalog list, Installed) before and after scanning.
• UI fixes: search + reserved-height scroll fix for "More models", long-name truncation, Models dialog clipping.
• Tests: new coverage for the fallback/quant logic, scan cache, curated manifest, and display-name consistency — full suite passing (427 backend tests).

Motivation and Context

FIX #

Screenshots

API Changes

• This PR includes API changes

Change Type

• Bug fix
• New feature
• Performance improvement
• Refactoring
• Documentation
• Dependency/Build system
• Breaking change
• Other (specify):

Testing Performed

• Tested locally
• Manual/QA verification

Checklist

• Follows project coding standards and conventions
• Documentation updated as needed
• Dependencies updated as needed
• No lint/build errors or new warnings
• All relevant tests are passing

High-level PR Summary

This PR expands the local model catalog by enabling SurfSense to offer models from Hugging Face GGUF repositories that aren't in Ollama's native registry, increasing available models from ~91 to over 1,400. It introduces a fallback mechanism where llmfit scan results with gguf_sources but no ollama_name are automatically mapped to hf.co/<repo> pull targets that Ollama natively supports. The implementation adds persistent, fingerprinted caching of llmfit scan results to avoid expensive rescans, restructures the catalog API to distinguish curated (staff picks, always visible) from explore (scan-derived, ranked by fit), and adds a scanned flag so the UI can show curated models and a "Scan hardware" prompt immediately without blocking on hardware detection. GPU verification on Windows confirmed that llmfit correctly scores these GGUF models with trustworthy quantization tags on non-Mac platforms, validating the approach.

⏱️ Estimated Review Time: 30-90 minutes

💡 Review Order Suggestion

Order	File Path
1	LLMFIT_MODEL_CATALOG_EXPANSION_SPEC.md
2	surfsense_local/backend/modules/llm/recommendations/types.py
3	surfsense_local/backend/modules/llm/recommendations/curated_models.py
4	surfsense_local/backend/modules/llm/recommendations/curated-models.json
5	surfsense_local/backend/modules/llm/recommendations/llmfit.py
6	surfsense_local/backend/modules/llm/recommendations/dependencies.py
7	surfsense_local/backend/modules/llm/recommendations/catalog.py
8	surfsense_local/backend/modules/llm/schemas.py
9	surfsense_local/backend/tests/unit/llm/recommendations/test_llmfit.py
10	surfsense_local/backend/tests/unit/llm/recommendations/test_curated_models.py
11	surfsense_local/backend/tests/unit/llm/recommendations/test_catalog.py
12	surfsense_local/backend/tests/integration/llm/test_routes.py
13	surfsense_local/frontend/src/features/model-catalog/api.ts
14	surfsense_local/frontend/src/features/model-catalog/model-catalog-page.tsx
15	surfsense_local/frontend/src/features/model-catalog/model-card.tsx
16	surfsense_local/frontend/src/features/model-catalog/model-catalog.test.tsx
17	surfsense_local/frontend/src/features/model-selection/model-selection-content.tsx
18	surfsense_local/frontend/src/features/onboarding/onboarding-page.tsx
19	surfsense_local/frontend/src/features/settings/models-settings.tsx
20	surfsense_local/frontend/src/app/app-bootstrap.test.tsx
21	surfsense_local/frontend/src/features/dashboard/dashboard-page.test.tsx
22	surfsense_local/frontend/src/features/onboarding/onboarding-page.test.tsx
23	surfsense_local/frontend/src/features/settings/settings-dialog.test.tsx

[vercel]

@AnishSarkar22 is attempting to deploy a commit to the Rohan Verma's projects Team on Vercel.

A member of the Team first needs to authorize it.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Advanced

Run ID: 73508641-9752-47c4-a146-a61e53723a84

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

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

• X
• Mastodon
• Reddit
• LinkedIn

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