pi-llama-cpp 0.11.0 → 0.13.0

package/tests/legacyModel.test.ts

@@ -1,6 +1,5 @@
 import { beforeEach, describe, expect, it } from "vitest";
 import { Mode } from "../src/enums/mode";
-import { Status } from "../src/enums/status";
 import { DataProperty } from "../src/interfaces/endpoints/models";
 import { LegacyModel } from "../src/models/legacyModel";
 import { createMockServer, mockRpc } from "./mocks";
@@ -47,28 +46,14 @@ describe("LegacyModel capabilities", () => {
 
     expect(capabilities).toEqual(["text"]);
   });
-});
-
-describe("LegacyModel getStatus", () => {
-  it("should return LOADED when not sleeping", async () => {
-    mockRpc.mockResolvedValueOnce({ is_sleeping: false });
-
-    const model = createModel();
-    const status = await model.getStatus();
 
-    expect(status).toBe(Status.LOADED);
-    expect(mockRpc).toHaveBeenCalledWith(
-      `/props?model=${model.id}&autoload=false`,
-    );
-  });
-
-  it("should return SLEEPING when is_sleeping is true", async () => {
-    mockRpc.mockResolvedValueOnce({ is_sleeping: true });
+  it("should fall back to text-only when auth fails", async () => {
+    mockRpc.mockRejectedValue(new Error("401 Unauthorized"));
 
     const model = createModel();
-    const status = await model.getStatus();
+    const capabilities = await model.getCapabilities();
 
-    expect(status).toBe(Status.SLEEPING);
+    expect(capabilities).toEqual(["text"]);
   });
 });
 

package/tests/mocks.ts

@@ -12,7 +12,7 @@ import {
 } from "../src/constants";
 import { Mode } from "../src/enums/mode";
 import { Status } from "../src/enums/status";
-import type { LlamaServer } from "../src/interfaces/settings";
+import type { ModelOverride } from "../src/interfaces/settings";
 import type { LlamaSettingsManager } from "../src/managers/settings";
 import { BaseModel } from "../src/models/baseModel";
 import { Server } from "../src/server";
@@ -32,18 +32,19 @@ export const makeSettingsStub = (
   overrides: Partial<LlamaSettingsManager> = {},
 ): LlamaSettingsManager =>
   ({
-    resolveTimeouts: vi.fn(() => ({
+    getLlamaSettings: vi.fn(async () => ({})),
+    getLlamaServers: vi.fn(async () => []),
+    resolveTimeouts: vi.fn(async () => ({
       pollingTimeout: POLLING_TIMEOUT,
       serverTimeout: SERVER_TIMEOUT,
     })),
-    resolveServers: vi.fn((): Server[] => []),
-    resolveSortBy: vi.fn(() => SORT_BY),
+    resolveServers: vi.fn(async () => []),
+    resolveSortBy: vi.fn(async () => SORT_BY),
     resolveApiKey: vi.fn(() => API_KEY_PLACEHOLDER),
-    resolveReactToModelSelect: vi.fn(() => REACT_TO_MODEL_SELECT),
-    resolveAutoloadOnMessage: vi.fn(() => AUTOLOAD_ON_MESSAGE),
+    resolveReactToModelSelect: vi.fn(async () => REACT_TO_MODEL_SELECT),
+    resolveAutoloadOnMessage: vi.fn(async () => AUTOLOAD_ON_MESSAGE),
     resolveThinkingLevel: vi.fn(() => undefined),
     resolveThinkingBudgets: vi.fn(() => ({ ...THINKING_BUDGETS })),
-    llamaServers: [] as LlamaServer[],
     takeWarnings: vi.fn((): string[] => []),
     setLlamaSetting: vi.fn(() => Promise.resolve()),
     ...overrides,
@@ -102,6 +103,7 @@ export type MockServerOverrides = Partial<
   models?: BaseModel[];
   pollingTimeout?: number;
   serverTimeout?: number;
+  overrides?: Record<string, ModelOverride>;
 };
 
 /**
@@ -124,6 +126,7 @@ export const createMockServer = (
     models,
     pollingTimeout,
     serverTimeout,
+    overrides: modelOverrides,
     initialize,
     ...members
   } = overrides;
@@ -131,7 +134,7 @@ export const createMockServer = (
   const settings = makeSettingsStub({
     ...(apiKey !== undefined && { resolveApiKey: vi.fn(() => apiKey) }),
     ...((pollingTimeout !== undefined || serverTimeout !== undefined) && {
-      resolveTimeouts: vi.fn(() => ({
+      resolveTimeouts: vi.fn(async () => ({
         pollingTimeout: pollingTimeout ?? POLLING_TIMEOUT,
         serverTimeout: serverTimeout ?? SERVER_TIMEOUT,
       })),
@@ -145,6 +148,7 @@ export const createMockServer = (
       baseUrl: baseUrl ?? "http://127.0.0.1:8080",
       customId,
       customName,
+      overrides: modelOverrides,
     },
     {
       createApiClient: () => apiClient,

package/tests/overrides.test.ts

@@ -0,0 +1,384 @@
+import { beforeEach, describe, expect, it } from "vitest";
+import { FALLBACK_CTX } from "../src/constants";
+import type { LlamaServer, ModelOverride } from "../src/interfaces/settings";
+import { SingleModel } from "../src/models/singleModel";
+import { Server } from "../src/server";
+import {
+  addOverrideEntry,
+  applyCostFieldValue,
+  formatOverrideSummary,
+  parseCostValue,
+  removeOverrideEntry,
+  updateOverrideEntry,
+} from "../src/ui/overrideEntryEditor";
+import { overrideFieldValue } from "../src/ui/overrideSettingsList";
+import { createMockServer, mockRpc } from "./mocks";
+
+beforeEach(() => {
+  mockRpc.mockReset();
+});
+
+const createModel = (overrides?: Record<string, ModelOverride>): SingleModel =>
+  new SingleModel(
+    {
+      id: "test",
+      tags: [],
+      object: "model",
+      owned_by: "test",
+      created: Date.now(),
+    },
+    createMockServer({ overrides }),
+  );
+
+/** Standard server responses: /props for capabilities, /v1/models for n_ctx */
+const mockDetection = (nCtx: number | undefined) =>
+  mockRpc.mockImplementation((endpoint: string) => {
+    if (endpoint.startsWith("/props")) {
+      return Promise.resolve({ modalities: { vision: false } });
+    }
+    return Promise.resolve({
+      data: [{ id: "test", meta: nCtx === undefined ? {} : { n_ctx: nCtx } }],
+    });
+  });
+
+// ─── Pattern matching (Server.findOverrideForModel) ───────────────────────
+
+describe("Server.findOverrideForModel", () => {
+  const makeServer = (overrides: Record<string, ModelOverride>): Server =>
+    createMockServer({ overrides });
+
+  it("should match a model by id prefix", () => {
+    const server = makeServer({ lla: { reasoning: false } });
+    expect(server.findOverrideForModel("llama-3")).toEqual({
+      reasoning: false,
+    });
+  });
+
+  it("should return undefined when no pattern matches", () => {
+    const server = makeServer({ qwen: { reasoning: false } });
+    expect(server.findOverrideForModel("llama-3")).toBeUndefined();
+  });
+
+  it("should prefer the longest (most specific) matching pattern", () => {
+    const server = makeServer({
+      llama: { reasoning: false },
+      "llama-3-8b": { reasoning: true, contextSize: 128 },
+    });
+    expect(server.findOverrideForModel("llama-3-8b")).toEqual({
+      reasoning: true,
+      contextSize: 128,
+    });
+    expect(server.findOverrideForModel("llama-3-70b")).toEqual({
+      reasoning: false,
+    });
+  });
+
+  it("should ignore empty patterns", () => {
+    const server = makeServer({ "": { reasoning: false } });
+    expect(server.findOverrideForModel("llama-3")).toBeUndefined();
+  });
+});
+
+// ─── reasoning override ────────────────────────────────────────────────────
+
+describe("BaseModel reasoning override", () => {
+  it("should default to true when no override matches", () => {
+    expect(createModel().reasoning).toBe(true);
+  });
+
+  it("should let the override win over the default", () => {
+    const model = createModel({ test: { reasoning: false } });
+    expect(model.reasoning).toBe(false);
+  });
+});
+
+// ─── capabilities override ────────────────────────────────────────────────
+
+describe("BaseModel capabilities override", () => {
+  it("should fully replace detection without contacting the server", async () => {
+    mockRpc.mockImplementation(() => {
+      throw new Error("override must not trigger a fetch");
+    });
+
+    const model = createModel({ test: { capabilities: ["text", "image"] } });
+    const capabilities = await model.getCapabilities();
+
+    expect(capabilities).toEqual(["text", "image"]);
+    expect(mockRpc).not.toHaveBeenCalled();
+  });
+
+  it("should not intercept detection when no override matches", async () => {
+    mockDetection(8192);
+
+    const model = createModel({ other: { capabilities: ["text"] } });
+    const capabilities = await model.getCapabilities();
+
+    expect(capabilities).toEqual(["text"]);
+    expect(mockRpc).toHaveBeenCalled();
+  });
+});
+
+// ─── contextSize override ─────────────────────────────────────────────────
+
+describe("BaseModel.getContextSize with contextSize override", () => {
+  it("should return the overridden context size instead of the detected one", async () => {
+    mockDetection(8192);
+
+    const model = createModel({ test: { contextSize: 32768 } });
+    const ctxSize = await model.getContextSize();
+
+    expect(ctxSize).toBe(32768);
+  });
+
+  it("should treat a stored 0 as absent and autodetect", async () => {
+    mockDetection(8192);
+
+    const model = createModel({ test: { contextSize: 0 } });
+    const ctxSize = await model.getContextSize();
+
+    expect(ctxSize).toBe(8192);
+  });
+
+  it("should autodetect when no override matches", async () => {
+    mockDetection(8192);
+
+    const model = createModel({ other: { contextSize: 32768 } });
+    const ctxSize = await model.getContextSize();
+
+    expect(ctxSize).toBe(8192);
+  });
+});
+
+// ─── toProviderConfig (cost, maxTokens, compat, precedence) ───────────────
+
+describe("toProviderConfig overrides", () => {
+  it("should merge a partial cost override with zero defaults", async () => {
+    mockDetection(8192);
+
+    const model = createModel({ test: { cost: { input: 0.2 } } });
+    const config = await model.toProviderConfig();
+
+    expect(config.cost).toEqual({
+      input: 0.2,
+      output: 0,
+      cacheRead: 0,
+      cacheWrite: 0,
+    });
+  });
+
+  it("should default every cost field to zero without an override", async () => {
+    mockDetection(8192);
+
+    const model = createModel();
+    const config = await model.toProviderConfig();
+
+    expect(config.cost).toEqual({
+      input: 0,
+      output: 0,
+      cacheRead: 0,
+      cacheWrite: 0,
+    });
+  });
+
+  it("should pass a compat override through to the provider config", async () => {
+    mockDetection(8192);
+
+    const compat = {
+      supportsStore: false,
+      maxTokensField: "max_tokens",
+    } as const;
+    const model = createModel({ test: { compat } });
+    const config = await model.toProviderConfig();
+
+    expect(config.compat).toEqual(compat);
+  });
+
+  it("should leave compat undefined without an override", async () => {
+    mockDetection(8192);
+
+    const model = createModel();
+    const config = await model.toProviderConfig();
+
+    expect(config.compat).toBeUndefined();
+  });
+
+  it("should use the overridden context size for contextWindow and as the maxTokens fallback", async () => {
+    mockDetection(8192);
+
+    const model = createModel({ test: { contextSize: 32768 } });
+    const config = await model.toProviderConfig();
+
+    expect(config.contextWindow).toBe(32768);
+    expect(config.maxTokens).toBe(32768);
+  });
+
+  it("should prefer an explicit maxTokens override over the contextSize override", async () => {
+    mockDetection(8192);
+
+    const model = createModel({
+      test: { contextSize: 32768, maxTokens: 4096 },
+    });
+    const config = await model.toProviderConfig();
+
+    expect(config.contextWindow).toBe(32768);
+    expect(config.maxTokens).toBe(4096);
+  });
+
+  it("should use the detected context size when no override matches", async () => {
+    mockDetection(8192);
+
+    const model = createModel();
+    const config = await model.toProviderConfig();
+
+    expect(config.contextWindow).toBe(8192);
+    expect(config.maxTokens).toBe(8192);
+  });
+
+  it("should fall back to FALLBACK_CTX when the model is missing from detection", async () => {
+    mockRpc.mockImplementation((endpoint: string) => {
+      if (endpoint.startsWith("/props")) {
+        return Promise.resolve({ modalities: { vision: false } });
+      }
+      // /v1/models: no entry for this model → getContextSize throws → fallback
+      return Promise.resolve({ data: [] });
+    });
+
+    const model = createModel();
+    const config = await model.toProviderConfig();
+
+    expect(config.contextWindow).toBe(FALLBACK_CTX);
+    expect(config.maxTokens).toBe(FALLBACK_CTX);
+  });
+});
+
+// ─── formatOverrideSummary ────────────────────────────────────────────────
+
+describe("formatOverrideSummary", () => {
+  it("should render every configured field", () => {
+    const summary = formatOverrideSummary({
+      cost: { input: 0.2, output: 0.6, cacheRead: 0.01 },
+      capabilities: ["text", "image"],
+      reasoning: false,
+      contextSize: 32768,
+      maxTokens: 4096,
+    });
+
+    expect(summary).toContain("input: $0.2");
+    expect(summary).toContain("output: $0.6");
+    expect(summary).toContain("cache read: $0.01");
+    expect(summary).toContain("capabilities: text,image");
+    expect(summary).toContain("reasoning: false");
+    expect(summary).toContain("contextSize: 32768");
+    expect(summary).toContain("maxTokens: 4096");
+  });
+
+  it("should omit zero, undefined and empty fields", () => {
+    const summary = formatOverrideSummary({
+      cost: { input: 0, output: 0.6 },
+      reasoning: undefined,
+    });
+
+    expect(summary).not.toContain("input:");
+    expect(summary).toContain("output: $0.6");
+    expect(summary).not.toContain("capabilities");
+    expect(summary).not.toContain("reasoning");
+    expect(summary).not.toContain("contextSize");
+    expect(summary).not.toContain("maxTokens");
+  });
+
+  it("should render an em dash for an empty override", () => {
+    expect(formatOverrideSummary({})).toBe("—");
+  });
+});
+
+// ─── override entry list helpers (overrideEntryEditor) ────────────────────
+
+describe("override entry list helpers", () => {
+  const servers = [
+    { url: "http://a", overrides: { "a-1": { reasoning: false } } },
+    { url: "http://b" },
+  ] as LlamaServer[];
+
+  it("addOverrideEntry should append a pattern → override entry", () => {
+    const next = addOverrideEntry(servers, 0, "a-2", { contextSize: 4096 });
+    expect(next[0].overrides).toEqual({
+      "a-1": { reasoning: false },
+      "a-2": { contextSize: 4096 },
+    });
+  });
+
+  it("updateOverrideEntry should replace in place without reordering", () => {
+    const next = updateOverrideEntry(servers, 0, 0, "a-renamed", {
+      reasoning: true,
+    });
+    expect(Object.keys(next[0].overrides!)).toEqual(["a-renamed"]);
+    expect(next[0].overrides!["a-renamed"]).toEqual({ reasoning: true });
+    expect(next[1]).toBe(servers[1]);
+  });
+
+  it("removeOverrideEntry should drop only the targeted entry", () => {
+    const withTwo = addOverrideEntry(servers, 0, "a-2", {});
+    const next = removeOverrideEntry(withTwo, 0, 0);
+    expect(Object.keys(next[0].overrides!)).toEqual(["a-2"]);
+  });
+
+  it("parseCostValue should accept blank, non-negative numbers; reject the rest", () => {
+    expect(parseCostValue("")).toBe(0);
+    expect(parseCostValue("  0.5 ")).toBe(0.5);
+    expect(parseCostValue("-1")).toBeNull();
+    expect(parseCostValue("abc")).toBeNull();
+    expect(parseCostValue("Infinity")).toBeNull();
+  });
+
+  it("applyCostFieldValue should set a positive value, keeping other fields", () => {
+    const next = applyCostFieldValue({ input: 0.2 }, "output", 0.6);
+    expect(next).toEqual({ input: 0.2, output: 0.6 });
+  });
+
+  it("applyCostFieldValue should remove the field on zero", () => {
+    const next = applyCostFieldValue({ input: 0.2, output: 0.6 }, "input", 0);
+    expect(next).toEqual({ output: 0.6 });
+  });
+
+  it("applyCostFieldValue should return undefined when the last field is zeroed", () => {
+    expect(applyCostFieldValue({ input: 0.2 }, "input", 0)).toBeUndefined();
+    expect(applyCostFieldValue(undefined, "input", 0)).toBeUndefined();
+  });
+
+  it("applyCostFieldValue should treat an empty override as unset cost", () => {
+    expect(applyCostFieldValue(undefined, "input", 0.5)).toEqual({
+      input: 0.5,
+    });
+  });
+});
+
+describe("overrideFieldValue", () => {
+  it("formats cost fields, defaulting to 0", () => {
+    expect(overrideFieldValue("cost.input", {})).toBe("0");
+    expect(overrideFieldValue("cost.input", { cost: { input: 0.2 } })).toBe(
+      "0.2",
+    );
+    expect(overrideFieldValue("cost.cacheWrite", {})).toBe("0");
+  });
+
+  it("formats capabilities and reasoning labels", () => {
+    expect(overrideFieldValue("capabilities", {})).toBe("text");
+    expect(
+      overrideFieldValue("capabilities", { capabilities: ["text", "image"] }),
+    ).toBe("text | image");
+    expect(overrideFieldValue("reasoning", {})).toBe("true");
+    expect(overrideFieldValue("reasoning", { reasoning: false })).toBe("false");
+  });
+
+  it("formats maxTokens/contextSize, defaulting to 0", () => {
+    expect(overrideFieldValue("maxTokens", {})).toBe("0");
+    expect(overrideFieldValue("maxTokens", { maxTokens: 4096 })).toBe("4096");
+    expect(overrideFieldValue("contextSize", { contextSize: 8192 })).toBe(
+      "8192",
+    );
+  });
+
+  it("returns an empty label for unknown fields", () => {
+    expect(overrideFieldValue("other", {})).toBe("");
+  });
+});

package/tests/server.test.ts

@@ -1,4 +1,4 @@
-import { beforeEach, describe, expect, it, vi } from "vitest";
+import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
 import { POLLING_TIMEOUT, SERVER_TIMEOUT } from "../src/constants";
 import { ServerStatus } from "../src/enums/serverStatus";
 import type { LlamaSettingsManager } from "../src/managers/settings";
@@ -35,6 +35,26 @@ describe("Server providerName", () => {
   });
 });
 
+describe("Server apiBaseUrl", () => {
+  it("should append the ENDPOINT_PREFIX to baseUrl", () => {
+    const server = new Server(settings, { baseUrl: "http://127.0.0.1:8080" });
+    expect(server.apiBaseUrl).toBe("http://127.0.0.1:8080/v1");
+  });
+
+  it("should not double the prefix when baseUrl already ends with it", () => {
+    const server = new Server(settings, {
+      baseUrl: "http://127.0.0.1:8080/v1",
+    });
+    expect(server.apiBaseUrl).toBe("http://127.0.0.1:8080/v1");
+  });
+
+  it("should keep baseUrl untouched for provider identity", () => {
+    const server = new Server(settings, { baseUrl: "http://127.0.0.1:8080" });
+    expect(server.baseUrl).toBe("http://127.0.0.1:8080");
+    expect(server.providerId).toBe("llama-server=http://127.0.0.1:8080");
+  });
+});
+
 describe("Server fetchModels", () => {
   it("should call the /models endpoint", async () => {
     mockRpc.mockResolvedValueOnce({
@@ -154,16 +174,16 @@ describe("Server postRequest", () => {
   });
 });
 
-describe("Server timeouts", () => {
-  it("should resolve timeouts from the injected settings", () => {
+describe("Server timeouts", async () => {
+  it("should resolve timeouts from the injected settings", async () => {
     const server = new Server(settings, { baseUrl: "http://127.0.0.1:8080" });
 
-    expect(server.serverTimeout).toBe(SERVER_TIMEOUT);
-    expect(server.pollingTimeout).toBe(POLLING_TIMEOUT);
+    expect(await server.getServerTimeout()).toBe(SERVER_TIMEOUT);
+    expect(await server.getPollingTimeout()).toBe(POLLING_TIMEOUT);
   });
 
-  it("should return resolved custom values", () => {
-    vi.mocked(settings.resolveTimeouts).mockReturnValue({
+  it("should return resolved custom values", async () => {
+    vi.mocked(settings.resolveTimeouts).mockResolvedValue({
       pollingTimeout: 90000,
       serverTimeout: 2000,
     });
@@ -174,29 +194,39 @@ describe("Server timeouts", () => {
       customName: "My Server",
     });
 
-    expect(server.pollingTimeout).toBe(90000);
+    expect(await server.getPollingTimeout()).toBe(90000);
   });
 
-  it("should read timeouts live from settings", () => {
+  it("should read timeouts live from settings", async () => {
     const server = new Server(settings, { baseUrl: "http://127.0.0.1:8080" });
 
-    vi.mocked(settings.resolveTimeouts).mockReturnValue({
+    vi.mocked(settings.resolveTimeouts).mockResolvedValue({
       pollingTimeout: 120000,
       serverTimeout: 3000,
     });
-    expect(server.pollingTimeout).toBe(120000);
+    expect(await server.getPollingTimeout()).toBe(120000);
 
-    vi.mocked(settings.resolveTimeouts).mockReturnValue({
+    vi.mocked(settings.resolveTimeouts).mockResolvedValue({
       pollingTimeout: 90000,
       serverTimeout: 2000,
     });
-    expect(server.pollingTimeout).toBe(90000);
+    expect(await server.getPollingTimeout()).toBe(90000);
   });
 });
 
 describe("Server isReady", () => {
+  // `isReady` delegates to the shared health probe (`utils/health`), which
+  // uses plain `fetch` instead of the ApiClient — stub it per test.
+  const stubFetch = (impl: () => Promise<unknown>) => {
+    vi.stubGlobal("fetch", vi.fn(impl));
+  };
+
+  afterEach(() => {
+    vi.unstubAllGlobals();
+  });
+
   it("should return READY when health status is ok", async () => {
-    mockRpc.mockResolvedValueOnce({ status: "ok" });
+    stubFetch(async () => ({ json: async () => ({ status: "ok" }) }));
 
     const server = createMockServer();
     const status = await server.isReady(1000);
@@ -205,7 +235,9 @@ describe("Server isReady", () => {
   });
 
   it("should return UNREACHABLE when health check fails", async () => {
-    mockRpc.mockRejectedValueOnce(new Error("connection refused"));
+    stubFetch(async () => {
+      throw new Error("connection refused");
+    });
 
     const server = createMockServer();
     const status = await server.isReady(1000);
@@ -214,11 +246,22 @@ describe("Server isReady", () => {
   });
 
   it("should return UNREACHABLE when health status is not ok", async () => {
-    mockRpc.mockResolvedValueOnce({ status: "error" });
+    stubFetch(async () => ({ json: async () => ({ status: "error" }) }));
 
     const server = createMockServer();
     const status = await server.isReady(1000);
 
     expect(status).toBe(ServerStatus.UNREACHABLE);
   });
+
+  it("should return TIMEOUT when the health check aborts", async () => {
+    stubFetch(async () => {
+      throw new DOMException("The operation timed out.", "TimeoutError");
+    });
+
+    const server = createMockServer();
+    const status = await server.isReady(1000);
+
+    expect(status).toBe(ServerStatus.TIMEOUT);
+  });
 });