pi-llama-cpp 0.9.0 → 0.9.2

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-llama-cpp",
-  "version": "0.9.0",
+  "version": "0.9.2",
   "description": "Pi extension for llama.cpp integration. Supports router, single and legacy models. Supports multiple servers.",
   "keywords": [
     "pi",
@@ -32,13 +32,13 @@
     ]
   },
   "peerDependencies": {
-    "@earendil-works/pi-ai": "*",
-    "@earendil-works/pi-coding-agent": "*",
-    "@earendil-works/pi-tui": "*"
+    "@earendil-works/pi-ai": ">=0.84.0",
+    "@earendil-works/pi-coding-agent": ">=0.84.0",
+    "@earendil-works/pi-tui": ">=0.84.0"
   },
   "devDependencies": {
-    "@types/node": "^26.1.1",
+    "@types/node": "^26.2.0",
     "prettier-plugin-organize-imports": "^4.3.0",
-    "vitest": "^4.1.10"
+    "vitest": "^4.1.11"
   }
 }

package/src/index.ts

@@ -47,8 +47,8 @@ export default async function (pi: ExtensionAPI) {
 
   pi.on(
     "before_provider_request",
-    async (event: BeforeProviderRequestEvent) =>
-      await eventManager.onBeforeProviderRequest(event),
+    async (event: BeforeProviderRequestEvent, ctx: ExtensionContext) =>
+      await eventManager.onBeforeProviderRequest(event, ctx),
   );
 
   pi.on(

package/src/managers/command.ts

@@ -177,15 +177,20 @@ export class CommandManager {
         }
       };
 
+      const onFinished = async () => {
+        cleanupProgress();
+        EventManager.resetInflightModel();
+
+        // Re-scan providers to ensure accuracy of loaded models
+        await this.serverManager.update(pi);
+
+        // Force TUI refresh so Pi picks up the updated model states
+        ctx.ui.setStatus(PROVIDER_NAME, " ");
+        ctx.ui.setStatus(PROVIDER_NAME, undefined);
+      };
+
       // Load the model without blocking the UI
-      model
-        .load()
-        .then(onSuccess)
-        .catch(onFailure)
-        .finally(() => {
-          cleanupProgress();
-          EventManager.resetInflightModel();
-        });
+      model.load().then(onSuccess).catch(onFailure).finally(onFinished);
     }
   }
 

package/src/managers/events.ts

@@ -72,7 +72,10 @@ export class EventManager {
    * @param event Request event
    * @returns Updated payload
    */
-  async onBeforeProviderRequest(event: BeforeProviderRequestEvent) {
+  async onBeforeProviderRequest(
+    event: BeforeProviderRequestEvent,
+    ctx: ExtensionContext,
+  ) {
     const payload = event.payload as { model?: string };
     const { model } = payload;
     if (!model) return payload;
@@ -86,7 +89,8 @@ export class EventManager {
 
     // Retrieve pi's current thinking level, so we can setup a budget
     const resolver = new ConfigResolver();
-    const level = resolver.resolveThinkingLevel() ?? "medium";
+    const level =
+      ctx.thinkingLevel ?? resolver.resolveThinkingLevel() ?? "medium";
     const budgets = resolver.resolveThinkingBudgets();
     const thinking_budget_tokens = budgets[level];
 

package/tests/events.test.ts

@@ -44,6 +44,21 @@ const createNonLlamaPayload = () => ({
   messages: [{ role: "user", content: "hello" }],
 });
 
+const createMockCtx = (
+  thinkingLevel?:
+    | "off"
+    | "minimal"
+    | "low"
+    | "medium"
+    | "high"
+    | "xhigh"
+    | "max",
+) =>
+  ({
+    thinkingLevel,
+    ui: { notify: vi.fn() },
+  }) as any;
+
 describe("EventManager.onBeforeProviderRequest", () => {
   describe("normal usage — each thinking level", () => {
     it.each([
@@ -68,8 +83,10 @@ describe("EventManager.onBeforeProviderRequest", () => {
         const eventManager = new EventManager([server]);
         const event = { payload: createPayload("model-a") };
 
+        const ctx = createMockCtx(level as any);
         const result = (await eventManager.onBeforeProviderRequest(
           event as any,
+          ctx,
         )) as Record<string, unknown>;
 
         expect(result.model).toBe("model-a");
@@ -92,8 +109,10 @@ describe("EventManager.onBeforeProviderRequest", () => {
         },
       };
 
+      const ctx = createMockCtx("low");
       const result = (await eventManager.onBeforeProviderRequest(
         event as any,
+        ctx,
       )) as Record<string, unknown>;
 
       expect(result.messages).toEqual([{ role: "user", content: "test" }]);
@@ -110,7 +129,11 @@ describe("EventManager.onBeforeProviderRequest", () => {
       const eventManager = new EventManager([server]);
       const event = { payload: createNonLlamaPayload() };
 
-      const result = await eventManager.onBeforeProviderRequest(event as any);
+      const ctx = createMockCtx();
+      const result = await eventManager.onBeforeProviderRequest(
+        event as any,
+        ctx,
+      );
 
       expect(result).toEqual(createNonLlamaPayload());
     });
@@ -124,7 +147,11 @@ describe("EventManager.onBeforeProviderRequest", () => {
       const eventManager = new EventManager([server]);
       const event = { payload: { messages: [] } };
 
-      const result = await eventManager.onBeforeProviderRequest(event as any);
+      const ctx = createMockCtx();
+      const result = await eventManager.onBeforeProviderRequest(
+        event as any,
+        ctx,
+      );
 
       expect(result).toEqual({ messages: [] });
     });
@@ -141,8 +168,10 @@ describe("EventManager.onBeforeProviderRequest", () => {
       const eventManager = new EventManager([server]);
       const event = { payload: createPayload("model-a") };
 
+      const ctx = createMockCtx("low");
       const result = (await eventManager.onBeforeProviderRequest(
         event as any,
+        ctx,
       )) as Record<string, unknown>;
 
       expect(result.thinking_budget_tokens).toBe(4096);
@@ -158,8 +187,10 @@ describe("EventManager.onBeforeProviderRequest", () => {
       const eventManager = new EventManager([server]);
       const event = { payload: createPayload("model-a") };
 
+      const ctx = createMockCtx("medium");
       const result = (await eventManager.onBeforeProviderRequest(
         event as any,
+        ctx,
       )) as Record<string, unknown>;
 
       // medium uses default since user only overrode low
@@ -185,8 +216,10 @@ describe("EventManager.onBeforeProviderRequest", () => {
       const eventManager = new EventManager([server]);
       const event = { payload: createPayload("model-a") };
 
+      const ctx = createMockCtx("medium");
       const result = (await eventManager.onBeforeProviderRequest(
         event as any,
+        ctx,
       )) as Record<string, unknown>;
 
       // Should fall back to default since "medium" is not in user budgets
@@ -207,8 +240,10 @@ describe("EventManager.onBeforeProviderRequest", () => {
       const eventManager = new EventManager([server]);
       const event = { payload: createPayload("model-a") };
 
+      const ctx = createMockCtx("off");
       const result = (await eventManager.onBeforeProviderRequest(
         event as any,
+        ctx,
       )) as Record<string, unknown>;
 
       expect(result).toMatchObject({
@@ -229,8 +264,10 @@ describe("EventManager.onBeforeProviderRequest", () => {
       const eventManager = new EventManager([server]);
       const event = { payload: createPayload("model-a") };
 
+      const ctx = createMockCtx("max");
       const result = (await eventManager.onBeforeProviderRequest(
         event as any,
+        ctx,
       )) as Record<string, unknown>;
 
       expect(result).toEqual(createPayload("model-a"));
@@ -247,8 +284,10 @@ describe("EventManager.onBeforeProviderRequest", () => {
       const eventManager = new EventManager([server]);
       const event = { payload: createPayload("model-a") };
 
+      const ctx = createMockCtx("high");
       const result = (await eventManager.onBeforeProviderRequest(
         event as any,
+        ctx,
       )) as Record<string, unknown>;
 
       expect(result.thinking_budget_tokens).toBe(DEFAULT_THINKING_BUDGETS.high);