fix: catalog sync, validation, and models list

[stefanoverna]

What

Four fixes for the catalog and models list command:

1. catalog sync now migrates JSON to SQLite — previously it only downloaded the JSON file, so models list (which reads from SQLite) would always show "catalog not found".

2. validateCatalog() handles orphan models gracefully — the upstream catalog may contain models for providers not yet in the catalog. Instead of failing the entire sync, orphan models are now skipped with a warning. Currently 7 models are affected:

   • deepreinforce/ornith-1.0-31b
   • deepreinforce/ornith-1.0-35b
   • deepreinforce/ornith-1.0-397b
   • deepreinforce/ornith-1.0-9b
   • meituan/longcat-2.0
   • microsoft/mai-code-1-flash
   • tencent/hy3-preview

   These reference providers (deepreinforce, meituan, microsoft, tencent) not yet present in the upstream providers map.

3. ListProviderModels() uses the pre-built index — the new catalog format uses a ProviderModels map built during load. Previously it fell through to a strings.HasPrefix scan that did not match the new key format, returning 0 models.

4. models list without --provider shows catalog providers — instead of returning routing intermediaries with no models, it now lists all catalog providers that have models.

Testing

• All 72 catalog tests pass
• All 32 main command tests pass (31 existing + 1 updated)
• Manually verified: catalog sync then models list shows all providers

[kilo-code-bot]

Code Review Roast 🔥

Verdict: No Issues Found | Recommendation: Merge

Oh wait, this PR is actually clean. I need to sit down. I had my flamethrower warmed up and everything.

📊 Overall: Like finding a unicorn in production — I didn't think clean PRs existed anymore, but here we are.

Files Reviewed (1 file)

• cmd/routatic-proxy/main.go

Previous Review Summaries (4 snapshots, latest commit 3ae12c1)

Current summary above is authoritative. Previous snapshots are kept for context only.

Previous review (commit 3ae12c1)

Verdict: No Issues Found | Recommendation: Merge

Oh wait, this PR is actually clean. I need to sit down. I had my flamethrower warmed up and everything.

📊 Overall: Like finding a unicorn in production — I didn't think clean PRs existed anymore, but here we are.

Files Reviewed (1 file)

• cmd/routatic-proxy/main.go

Previous review (commit c5455df)

Verdict: 1 Issue Found | Recommendation: Address before merge

Overview

Severity	Count
🚨 critical	0
⚠️ warning	0
💡 suggestion	1
🤏 nitpick	0

Issue Details (click to expand)

File	Line	Roast
cmd/routatic-proxy/main.go	392	if-body at lines 392–394 still at 4 tabs after the if-statement was bumped to 4 tabs — gofmt violation introduced by partial fix

---

🏆 Best part: The previous compile-error (openGUI 2-arg call) and the gofmt indentation on lines 390–391 are properly fixed this time. You didn't half-ass it — well, almost.

💀 Worst part: The if-body at lines 392–394 is now at the same indent level as the if-statement (4 tabs), when it should be at 5 tabs. The fix moved the header up but forgot the body. gofmt would catch this on the first pass.

📊 Overall: Like a mechanic who fixed the engine alignment but left the hood slightly ajar. The catalog fixes are genuinely good.

Files Reviewed (1 file)

• cmd/routatic-proxy/main.go - 1 issue

Fix these issues in Kilo Cloud

Previous review (commit 4ab0308)

Verdict: 1 Issue Found | Recommendation: Address before merge

Overview

Severity	Count
🤏 nitpick	1
🚨 critical	0
⚠️ warning	0
💡 suggestion	0

Issue Details (click to expand)

File	Line	Roast
cmd/routatic-proxy/main.go	389	// Open GUI comment and if err := openGUI(guiURL) block indented at 3 tabs inside a 4-tab scope — gofmt violation left by the fix commit

---

🏆 Best part: The openGUI(guiURL, cancel) compile-error from the previous review is properly dead — you fixed it in a second patch instead of arguing about it. Unusual for a reviewer, but I'll take it.

💀 Worst part: After fixing the compile error, you left the indentation one tab short on the now-corrected lines. gofmt would yell about this — it reads like the same short-attention-span merge conflict that introduced the 2-arg bug in the first place.

📊 Overall: Like a mechanic who fixed the engine but left the hood slightly ajar. The catalog fixes are genuinely good — JSON→SQLite migration on sync and graceful unknown-provider handling are solid wins.

Files Reviewed (5 files)

• cmd/routatic-proxy/catalog.go - 0 issues
• cmd/routatic-proxy/main.go - 1 issue (gofmt indentation at lines 389–390)
• cmd/routatic-proxy/main_models_test.go - 0 issues
• internal/catalog/load.go - 0 issues
• internal/catalog/resolve.go - 0 issues

---

Fix these issues in Kilo Cloud

Previous review (commit 5db0a19)

Verdict: 1 Issue Found | Recommendation: Address before merge

Overview

Severity	Count
🚨 critical	1
⚠️ warning	0
💡 suggestion	0
🤏 nitpick	0

Issue Details (click to expand)

File	Line	Roast
cmd/routatic-proxy/main.go	391	openGUI is called with 2 args but all platform implementations still take 1 — guaranteed compile failure

---

🏆 Best part: The catalog migration logic actually solves a real pain point — syncing JSON but not migrating to SQLite was a user-facing disaster that this PR finally fixes.

💀 Worst part: openGUI(guiURL, cancel) in main.go:391 — you changed the call signature but forgot to update any of the three platform-specific implementations. This PR won't compile on Linux, macOS with CGO, or macOS without CGO. You broke the build everywhere.

📊 Overall: Like a chef who perfectly seasons the steak but sets the kitchen on fire. The catalog fixes are solid; the broken openGUI call should have been caught by any go build before pushing.

Files Reviewed (5 files)

• cmd/routatic-proxy/catalog.go - 1 issue
• cmd/routatic-proxy/main.go - 1 issue
• cmd/routatic-proxy/main_models_test.go
• internal/catalog/load.go
• internal/catalog/resolve.go

---

Fix these issues in Kilo Cloud

---

_{Reviewed by step-3.7-flash · Input: 32.4K · Output: 4.2K · Cached: 121.5K}

_{Review guidance: REVIEW.md from base branch main (truncated)}

[stefanoverna]

All issues addressed:

1. openGUI(guiURL, cancel) → fixed: removed the stray cancel argument. The catalog branch no longer touches openGUI at all — the signature change belonged to a different PR. Build and all 32 tests pass.