percher 0.5.22 → 0.5.24

package/dist/templates/capsule/src/lib/ai.ts

@@ -1,91 +1,15 @@
+import { createManagedInferenceClient } from "./ai-client";
 import { pb } from "./db";
 
-// Managed inference endpoint, injected at build time (a public URL, not a
-// secret). The Dockerfile declares ARG VITE_PERCHER_AI_URL.
-const AI_URL = import.meta.env.VITE_PERCHER_AI_URL;
-
-/**
- * Keyless managed inference — NO API key, billed to the app owner's Percher
- * plan. Only works for SIGNED-IN users (the proxy validates the current PB
- * session against this app's sidecar). Returns the completion text.
- *
- *   const reply = await complete("Summarize: " + text);
- */
-export async function complete(
-  prompt: string,
-  opts?: { model?: string; maxTokens?: number },
-): Promise<string> {
-  if (!AI_URL) throw new Error("Managed inference isn't configured for this app.");
-  if (!pb.authStore.isValid) throw new Error("Sign in to use AI.");
-  const res = await fetch(AI_URL, {
-    method: "POST",
-    headers: { "content-type": "application/json", authorization: pb.authStore.token },
-    body: JSON.stringify({
-      model: opts?.model,
-      max_tokens: opts?.maxTokens ?? 512,
-      messages: [{ role: "user", content: prompt }],
-    }),
-  });
-  if (!res.ok) throw new Error(`AI request failed (${res.status})`);
-  const data = (await res.json()) as { choices?: Array<{ message?: { content?: string } }> };
-  return data.choices?.[0]?.message?.content ?? "";
-}
-
-/**
- * Streaming variant of {@link complete} — calls `onToken` with each text delta
- * as it arrives and resolves with the full text. Same auth + billing as
- * `complete` (keyless, billed to the app owner). Use it for a typewriter UI:
- *
- *   const full = await streamComplete("Write a haiku", (t) => { output.value += t; });
- */
-export async function streamComplete(
-  prompt: string,
-  onToken: (delta: string) => void,
-  opts?: { model?: string; maxTokens?: number },
-): Promise<string> {
-  if (!AI_URL) throw new Error("Managed inference isn't configured for this app.");
-  if (!pb.authStore.isValid) throw new Error("Sign in to use AI.");
-  const res = await fetch(AI_URL, {
-    method: "POST",
-    headers: { "content-type": "application/json", authorization: pb.authStore.token },
-    body: JSON.stringify({
-      model: opts?.model,
-      max_tokens: opts?.maxTokens ?? 512,
-      messages: [{ role: "user", content: prompt }],
-      stream: true,
-    }),
-  });
-  if (!res.ok || !res.body) throw new Error(`AI request failed (${res.status})`);
-
-  const reader = res.body.getReader();
-  const decoder = new TextDecoder();
-  let buffer = "";
-  let full = "";
-  for (;;) {
-    const { done, value } = await reader.read();
-    if (done) break;
-    buffer += decoder.decode(value, { stream: true });
-    // SSE events are separated by a blank line; parse each complete one.
-    let sep = buffer.indexOf("\n\n");
-    while (sep !== -1) {
-      const event = buffer.slice(0, sep);
-      buffer = buffer.slice(sep + 2);
-      sep = buffer.indexOf("\n\n");
-      const line = event.startsWith("data:") ? event.slice(5).trim() : "";
-      if (!line || line === "[DONE]") continue;
-      try {
-        const chunk = JSON.parse(line) as {
-          choices?: Array<{ delta?: { content?: string } }>;
-        };
-        const delta = chunk.choices?.[0]?.delta?.content ?? "";
-        if (delta) {
-          full += delta;
-          onToken(delta);
-        }
-      } catch {
-        /* ignore keep-alive / non-JSON lines */
-      }
-    }
-  }
-  return full;
-}
+// This is a public URL, never a provider credential. Percher injects it only
+// when the shared server-side feature gate is operational.
+const client = createManagedInferenceClient({
+  endpoint: import.meta.env.VITE_PERCHER_AI_URL as string | undefined,
+  getSession: () => ({ valid: pb.authStore.isValid, token: pb.authStore.token }),
+});
+
+export { ManagedInferenceError } from "./ai-client";
+export type { ManagedInferenceOptions } from "./ai-client";
+export const isManagedInferenceConfigured = client.isConfigured;
+export const complete = client.complete;
+export const streamComplete = client.streamComplete;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "percher",
-  "version": "0.5.22",
+  "version": "0.5.24",
   "description": "CLI for the Percher hosting platform — publish apps with one command",
   "type": "module",
   "bin": {
@@ -27,9 +27,9 @@
     "typecheck": "tsc --noEmit"
   },
   "dependencies": {
-    "@percher/client": "^0.2.21",
-    "@percher/core": "^0.4.27",
-    "@percher/toml": "^0.1.10",
+    "@percher/client": "^0.3.1",
+    "@percher/core": "^0.4.29",
+    "@percher/toml": "^0.1.11",
     "cac": "^7.0.0",
     "picocolors": "^1.1.1"
   },