fix(core): bump nitro to 3.0.260610-beta to fix dev-server cold-start crash

[teleaziz]

Bumps nitro 3.0.260415-beta → 3.0.260610-beta to fix a dev-server
cold-start crash (Vite environment "nitro" is unavailable / UND_ERR_SOCKET).
The Nitro v3-beta worker raced its entry import on slow startup; the newer beta
resolves it

[github-actions]

Here's a visual recap of what changed:

Open the full interactive recap

[builder-io-integration]

⚠️ Review Agent ran into a problem and couldn't finish reviewing the latest commit.

We've been automatically notified and are looking into it. Push a new commit to re-trigger the review, or contact support@builder.io if this keeps happening.

Error ID: 31ae79ebf5b847a99cf987e409f0cdc0

[builder-io-integration]

Builder reviewed your changes and found 1 potential issue 🟡

Review Details

This incremental update expands the original Nitro bump into a fuller cold-start fix for framework routes. In addition to aligning nitro, jiti, and rollup, the PR now patches h3.config.onRequest in framework-request-handler.ts so framework-route readiness is awaited before h3 snapshots middleware, and it adds targeted regression tests for the dev-runtime race. The overall direction is sound: the new gate addresses the timing problem at the right lifecycle point, and the dependency metadata is now aligned for downstream consumers.

I classified this as standard risk because it changes core runtime/request-dispatch infrastructure rather than product UI. After re-reviewing the updated code and verifying the dependency alignment, I found one confirmed behavioral issue in the failure-path handling for broad tracked prefixes.

Key findings

• 🟡 MEDIUM: a failed plugin tracked under /_agent-native will cause the existing 503 fallback to shadow unrelated healthy routes under that prefix, so one transient core-routes failure can take down sibling endpoints like /_agent-native/mcp or /_agent-native/agent-chat.

Overall assessment

• ✅ Correct lifecycle point for the cold-start race fix
• ✅ Dependency and changeset alignment look good
• ⚠️ Failure fallback remains too broad for prefix-scoped plugin tracking

🧪 Browser testing: Skipped — PR only modifies backend/config/docs, no UI impact.

[builder-io-integration]

🟡 Broad tracked prefix can shadow unrelated healthy framework routes

Tracking this init under /_agent-native means the existing 503 failure fallback will match every sibling route under that prefix if init rejects once. In practice a transient core-routes failure can start returning 503 for already-healthy endpoints like /_agent-native/mcp or /_agent-native/agent-chat, so the failure handling needs narrower ownership or path scoping.

Additional Info

Independently reproduced against the current branch by registering a healthy `/_agent-native/agent-chat` handler, then calling `trackPluginInit(nitroApp, rejectedPromise, { paths: ["/_agent-native"] })` and dispatching `/_agent-native/agent-chat`; the placeholder returned `{ error: "agent-native route is initializing or unavailable: boom" }` with HTTP 503 instead of the healthy handler. Root cause is the unchanged placeholder logic in framework-request-handler.ts that matches failures by prefix via `resolveMountMatch(reqPath, failedPath)` before falling through.

[builder-io-integration]

Builder reviewed your changes and has a few items to flag 🟡

Review Details

This incremental update extends the original Nitro bump with two additional pieces of infrastructure work: a dev-only Vite forwarder for extension-bearing framework paths in packages/core/src/vite/client.ts, and extra tests around framework route readiness plus the new dev-path classification. I re-reviewed the latest diff, reran the targeted framework-request-handler and vite/client specs successfully, and re-checked the dependency alignment (nitro, jiti, rollup) introduced earlier.

I classified this as standard risk because it continues to modify core runtime and dev-server behavior rather than product UI. The newly added Vite forwarding logic is directionally sound for the reported cold-start / dev-routing failures, and I did not find a new regression in the latest delta. However, the previously posted medium-severity comment about broad /_agent-native failure scoping still appears to be present and unchanged, so I am not re-posting it.

Overall assessment

• ✅ Nitro/jiti/rollup alignment remains correct
• ✅ New dev forwarder and tests are coherent with the intended fix
• ⚠️ Previously open framework-route failure-scope issue still appears unresolved (not re-posted)
• ✅ No additional actionable issues found in this incremental update

🧪 Browser testing: Skipped — PR only modifies backend/config/docs, no UI impact.