YC (S26) | Open Computer History | Continuously record your company computer work, map your workflows, help you find work worth automating, and power your agents' context
Rust agentsagiaiai-memoryaudio-recordingcomputer-visionhermeshermes-agentllmlocal-ailocal-firstmachine-learningmcpmultimodalopenclawprivacyrewindscreen-recordingspeech-to-textycombinator
fix(app): seed default AI preset on auto instead of claude-opus-4-8 (#4065)
Louis Beaumontd02e6daaec69
diff --git a/apps/screenpipe-app-tauri/lib/hooks/use-settings.tsx b/apps/screenpipe-app-tauri/lib/hooks/use-settings.tsxindex 423f97ef4..e8ba06abe 100644--- a/apps/screenpipe-app-tauri/lib/hooks/use-settings.tsx+++ b/apps/screenpipe-app-tauri/lib/hooks/use-settings.tsx@@ -423,17 +423,15 @@ const DEFAULT_IGNORED_WINDOWS_PER_OS: Record<string, string[]> = {linux: ["Info center", "Discover", "Parted"],};-// Two default screenpipe-cloud presets on first install:-// - "Chat": Claude Opus 4.7 if the user is pro, Claude Sonnet 4.5 otherwise.-// Opus is gated in the ai-gateway (subscribed tier), so pushing-// it to non-pro users would 403 their first message.-// - "Pipes": Claude Haiku 4.5 — cheap/fast for recurring pipe runs.-// Pipes default to this preset; users can override per-pipe.+// Default screenpipe-cloud presets on first install — every seed runs on+// "auto": the ai-gateway routes to the best model the user's tier allows, so+// nobody starts pinned to a specific (possibly tier-gated) Claude model.+// - "Chat": default preset, used by interactive chat.+// - "Pipes": same routing, separate id so recurring pipe runs can be+// customized independently; users can override per-pipe.const CHAT_PRESET_ID = "chat";const PIPES_PRESET_ID = "pipes";-// Pro users get the chat / pipes pair (opus for interactive chat, auto for-// pipe runs that pick the cheapest model that fits the task).// Non-pro users get a single "screenpipe" preset on auto — auto handles// model routing without needing the user to know what to pick.const SCREENPIPE_PRESET_ID = "screenpipe";@@ -445,7 +443,7 @@ export function makeDefaultPresets(isPro: boolean): AIPreset[] {id: CHAT_PRESET_ID,provider: "screenpipe-cloud",url: "",- model: "claude-opus-4-8",+ model: "auto",maxContextChars: 200000,defaultPreset: true,prompt: "",diff --git a/apps/screenpipe-app-tauri/lib/utils/pick-pipe-preset.ts b/apps/screenpipe-app-tauri/lib/utils/pick-pipe-preset.tsindex fa020b18b..dc2313b14 100644--- a/apps/screenpipe-app-tauri/lib/utils/pick-pipe-preset.ts+++ b/apps/screenpipe-app-tauri/lib/utils/pick-pipe-preset.ts@@ -6,12 +6,12 @@* Pick the AI preset to assign to a freshly installed pipe.** Pipes should run on the dedicated "pipes" preset (seeded on `auto` — cheap- * and tier-safe). The user's `defaultPreset` for pro users is the "chat"- * preset on Opus, a premium model NOT in lower tiers' allow-lists — pinning a- * pipe to it makes the pipe fail with "model not available for your tier" the- * moment tier resolution flickers to logged_in/anonymous (token refresh,- * sidecar restart). `auto` lets the gateway pick an allowed model and never- * 403s.+ * and tier-safe). The user's `defaultPreset` may be pinned to a premium model+ * (e.g. Opus — older installs seeded it that way, and users can pick it) NOT+ * in lower tiers' allow-lists — pinning a pipe to it makes the pipe fail with+ * "model not available for your tier" the moment tier resolution flickers to+ * logged_in/anonymous (token refresh, sidecar restart). `auto` lets the+ * gateway pick an allowed model and never 403s.** Resolution order:* 1. the dedicated "pipes" preset (id === "pipes")diff --git a/docs/mockups/default-preset-auto.png b/docs/mockups/default-preset-auto.pngnew file mode 100644index 000000000..5cc15d209Binary files /dev/null and b/docs/mockups/default-preset-auto.png differ