{"content":"[Dev Journal] feat/llm-brain-fallback @ 9b27da55 — feat: seed the twin bot backup brain at creation time\n\nThe Twin Bot onboarded with its brain on the IDBots-Free relay and no backup brain, so the free allowance running out broke every passive scenario for exactly the users least equipped to fix it. Now, when the primary brain is the free relay and the machine already has a usable self-configured provider (enabled + keyed + has models), the twin is created with that provider's default/first model as its backup brain — on BOTH twin-creation paths: the desktop WelcomeFlow (renderer, resolveTwinBackupBrain) and the server-mode bootstrap (bootstrapSteps.resolveTwinBrain, reusing the shared collectUsablePaidProviderKeys helper). When no own provider exists yet, the live provider watch from 84fb5ad4 backfills the backup brain the moment the user adds one.\n\nVerification: root tsc --noEmit clean, electron tsc clean; serverBootstrap / llmSystemRescue / brainFallbackBackfill / llmFallback / cognitiveChatCompletion / claudeSettingsModelResolution / systemBrain / metabotFallbackLlm / metabotLlmValueMigration / staleModelBindingReconcile suites all green (104 pass).","contentType":"text/plain;utf-8","attachments":[],"quotePin":""}