modelfusion 0.106.0 → 0.108.0

package/model-function/generate-text/prompt-template/MistralInstructPromptTemplate.test.cjs

@@ -0,0 +1,60 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+const MistralInstructPromptTemplate_js_1 = require("./MistralInstructPromptTemplate.cjs");
+describe("text prompt", () => {
+    it("should format prompt", () => {
+        const prompt = (0, MistralInstructPromptTemplate_js_1.text)().format("prompt");
+        expect(prompt).toMatchSnapshot();
+    });
+});
+describe("instruction prompt", () => {
+    it("should format prompt with instruction", () => {
+        const prompt = (0, MistralInstructPromptTemplate_js_1.instruction)().format({
+            instruction: "instruction",
+        });
+        expect(prompt).toMatchSnapshot();
+    });
+    it("should format prompt with system and instruction", () => {
+        const prompt = (0, MistralInstructPromptTemplate_js_1.instruction)().format({
+            system: "system",
+            instruction: "instruction",
+        });
+        expect(prompt).toMatchSnapshot();
+    });
+    it("should format prompt with instruction and response prefix", () => {
+        const prompt = (0, MistralInstructPromptTemplate_js_1.instruction)().format({
+            instruction: "instruction",
+            responsePrefix: "response prefix",
+        });
+        expect(prompt).toMatchSnapshot();
+    });
+});
+describe("chat prompt", () => {
+    it("should format prompt with user message", () => {
+        const prompt = (0, MistralInstructPromptTemplate_js_1.chat)().format({
+            messages: [{ role: "user", content: "user message" }],
+        });
+        expect(prompt).toMatchSnapshot();
+    });
+    it("should format prompt with user-assistant-user messages", () => {
+        const prompt = (0, MistralInstructPromptTemplate_js_1.chat)().format({
+            messages: [
+                { role: "user", content: "1st user message" },
+                { role: "assistant", content: "assistant message" },
+                { role: "user", content: "2nd user message" },
+            ],
+        });
+        expect(prompt).toMatchSnapshot();
+    });
+    it("should format prompt with system message and user-assistant-user messages", () => {
+        const prompt = (0, MistralInstructPromptTemplate_js_1.chat)().format({
+            system: "you are a chatbot",
+            messages: [
+                { role: "user", content: "1st user message" },
+                { role: "assistant", content: "assistant message" },
+                { role: "user", content: "2nd user message" },
+            ],
+        });
+        expect(prompt).toMatchSnapshot();
+    });
+});

package/model-function/generate-text/prompt-template/MistralInstructPromptTemplate.test.js

@@ -0,0 +1,58 @@
+import { chat, instruction, text } from "./MistralInstructPromptTemplate.js";
+describe("text prompt", () => {
+    it("should format prompt", () => {
+        const prompt = text().format("prompt");
+        expect(prompt).toMatchSnapshot();
+    });
+});
+describe("instruction prompt", () => {
+    it("should format prompt with instruction", () => {
+        const prompt = instruction().format({
+            instruction: "instruction",
+        });
+        expect(prompt).toMatchSnapshot();
+    });
+    it("should format prompt with system and instruction", () => {
+        const prompt = instruction().format({
+            system: "system",
+            instruction: "instruction",
+        });
+        expect(prompt).toMatchSnapshot();
+    });
+    it("should format prompt with instruction and response prefix", () => {
+        const prompt = instruction().format({
+            instruction: "instruction",
+            responsePrefix: "response prefix",
+        });
+        expect(prompt).toMatchSnapshot();
+    });
+});
+describe("chat prompt", () => {
+    it("should format prompt with user message", () => {
+        const prompt = chat().format({
+            messages: [{ role: "user", content: "user message" }],
+        });
+        expect(prompt).toMatchSnapshot();
+    });
+    it("should format prompt with user-assistant-user messages", () => {
+        const prompt = chat().format({
+            messages: [
+                { role: "user", content: "1st user message" },
+                { role: "assistant", content: "assistant message" },
+                { role: "user", content: "2nd user message" },
+            ],
+        });
+        expect(prompt).toMatchSnapshot();
+    });
+    it("should format prompt with system message and user-assistant-user messages", () => {
+        const prompt = chat().format({
+            system: "you are a chatbot",
+            messages: [
+                { role: "user", content: "1st user message" },
+                { role: "assistant", content: "assistant message" },
+                { role: "user", content: "2nd user message" },
+            ],
+        });
+        expect(prompt).toMatchSnapshot();
+    });
+});

package/model-function/generate-text/prompt-template/NeuralChatPromptTemplate.test.cjs

@@ -46,4 +46,15 @@ describe("chat prompt", () => {
         });
         expect(prompt).toMatchSnapshot();
     });
+    it("should format prompt with system message and user-assistant-user messages", () => {
+        const prompt = (0, NeuralChatPromptTemplate_js_1.chat)().format({
+            system: "you are a chatbot",
+            messages: [
+                { role: "user", content: "1st user message" },
+                { role: "assistant", content: "assistant message" },
+                { role: "user", content: "2nd user message" },
+            ],
+        });
+        expect(prompt).toMatchSnapshot();
+    });
 });

package/model-function/generate-text/prompt-template/NeuralChatPromptTemplate.test.js

@@ -44,4 +44,15 @@ describe("chat prompt", () => {
         });
         expect(prompt).toMatchSnapshot();
     });
+    it("should format prompt with system message and user-assistant-user messages", () => {
+        const prompt = chat().format({
+            system: "you are a chatbot",
+            messages: [
+                { role: "user", content: "1st user message" },
+                { role: "assistant", content: "assistant message" },
+                { role: "user", content: "2nd user message" },
+            ],
+        });
+        expect(prompt).toMatchSnapshot();
+    });
 });

package/model-function/generate-text/prompt-template/TextPromptTemplate.test.cjs

@@ -46,4 +46,15 @@ describe("chat prompt", () => {
         });
         expect(prompt).toMatchSnapshot();
     });
+    it("should format prompt with system message and user-assistant-user messages", () => {
+        const prompt = (0, TextPromptTemplate_js_1.chat)().format({
+            system: "you are a chatbot",
+            messages: [
+                { role: "user", content: "1st user message" },
+                { role: "assistant", content: "assistant message" },
+                { role: "user", content: "2nd user message" },
+            ],
+        });
+        expect(prompt).toMatchSnapshot();
+    });
 });

package/model-function/generate-text/prompt-template/TextPromptTemplate.test.js

@@ -44,4 +44,15 @@ describe("chat prompt", () => {
         });
         expect(prompt).toMatchSnapshot();
     });
+    it("should format prompt with system message and user-assistant-user messages", () => {
+        const prompt = chat().format({
+            system: "you are a chatbot",
+            messages: [
+                { role: "user", content: "1st user message" },
+                { role: "assistant", content: "assistant message" },
+                { role: "user", content: "2nd user message" },
+            ],
+        });
+        expect(prompt).toMatchSnapshot();
+    });
 });

package/model-function/generate-text/prompt-template/VicunaPromptTemplate.test.cjs

@@ -18,4 +18,15 @@ describe("chat prompt", () => {
         });
         expect(prompt).toMatchSnapshot();
     });
+    it("should format prompt with system message and user-assistant-user messages", () => {
+        const prompt = (0, VicunaPromptTemplate_js_1.chat)().format({
+            system: "you are a chatbot",
+            messages: [
+                { role: "user", content: "1st user message" },
+                { role: "assistant", content: "assistant message" },
+                { role: "user", content: "2nd user message" },
+            ],
+        });
+        expect(prompt).toMatchSnapshot();
+    });
 });

package/model-function/generate-text/prompt-template/VicunaPromptTemplate.test.js

@@ -16,4 +16,15 @@ describe("chat prompt", () => {
         });
         expect(prompt).toMatchSnapshot();
     });
+    it("should format prompt with system message and user-assistant-user messages", () => {
+        const prompt = chat().format({
+            system: "you are a chatbot",
+            messages: [
+                { role: "user", content: "1st user message" },
+                { role: "assistant", content: "assistant message" },
+                { role: "user", content: "2nd user message" },
+            ],
+        });
+        expect(prompt).toMatchSnapshot();
+    });
 });

package/model-function/generate-text/prompt-template/index.cjs

@@ -26,7 +26,7 @@ var __exportStar = (this && this.__exportStar) || function(m, exports) {
     for (var p in m) if (p !== "default" && !Object.prototype.hasOwnProperty.call(exports, p)) __createBinding(exports, m, p);
 };
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.VicunaPrompt = exports.TextPrompt = exports.NeuralChatPrompt = exports.Llama2Prompt = exports.ChatMLPrompt = exports.AlpacaPrompt = void 0;
+exports.VicunaPrompt = exports.TextPrompt = exports.NeuralChatPrompt = exports.MistralInstructPrompt = exports.Llama2Prompt = exports.ChatMLPrompt = exports.AlpacaPrompt = void 0;
 exports.AlpacaPrompt = __importStar(require("./AlpacaPromptTemplate.cjs"));
 exports.ChatMLPrompt = __importStar(require("./ChatMLPromptTemplate.cjs"));
 __exportStar(require("./ChatPrompt.cjs"), exports);
@@ -34,6 +34,7 @@ __exportStar(require("./ContentPart.cjs"), exports);
 __exportStar(require("./InstructionPrompt.cjs"), exports);
 __exportStar(require("./InvalidPromptError.cjs"), exports);
 exports.Llama2Prompt = __importStar(require("./Llama2PromptTemplate.cjs"));
+exports.MistralInstructPrompt = __importStar(require("./MistralInstructPromptTemplate.cjs"));
 exports.NeuralChatPrompt = __importStar(require("./NeuralChatPromptTemplate.cjs"));
 exports.TextPrompt = __importStar(require("./TextPromptTemplate.cjs"));
 exports.VicunaPrompt = __importStar(require("./VicunaPromptTemplate.cjs"));

package/model-function/generate-text/prompt-template/index.d.ts

@@ -5,6 +5,7 @@ export * from "./ContentPart.js";
 export * from "./InstructionPrompt.js";
 export * from "./InvalidPromptError.js";
 export * as Llama2Prompt from "./Llama2PromptTemplate.js";
+export * as MistralInstructPrompt from "./MistralInstructPromptTemplate.js";
 export * as NeuralChatPrompt from "./NeuralChatPromptTemplate.js";
 export * as TextPrompt from "./TextPromptTemplate.js";
 export * as VicunaPrompt from "./VicunaPromptTemplate.js";

package/model-function/generate-text/prompt-template/index.js

@@ -5,6 +5,7 @@ export * from "./ContentPart.js";
 export * from "./InstructionPrompt.js";
 export * from "./InvalidPromptError.js";
 export * as Llama2Prompt from "./Llama2PromptTemplate.js";
+export * as MistralInstructPrompt from "./MistralInstructPromptTemplate.js";
 export * as NeuralChatPrompt from "./NeuralChatPromptTemplate.js";
 export * as TextPrompt from "./TextPromptTemplate.js";
 export * as VicunaPrompt from "./VicunaPromptTemplate.js";

package/model-provider/llamacpp/LlamaCppBakLLaVA1PromptTemplate.d.ts

@@ -1,11 +1,11 @@
 import { TextGenerationPromptTemplate } from "../../model-function/generate-text/TextGenerationPromptTemplate.js";
 import { ChatPrompt } from "../../model-function/generate-text/prompt-template/ChatPrompt.js";
 import { InstructionPrompt } from "../../model-function/generate-text/prompt-template/InstructionPrompt.js";
-import { LlamaCppTextGenerationPrompt } from "./LlamaCppTextGenerationModel.js";
+import { LlamaCppCompletionPrompt } from "./LlamaCppCompletionModel.js";
 /**
  * BakLLaVA 1 uses a Vicuna 1 prompt. This mapping combines it with the LlamaCpp prompt structure.
  *
  * @see https://github.com/SkunkworksAI/BakLLaVA
  */
-export declare function instruction(): TextGenerationPromptTemplate<InstructionPrompt, LlamaCppTextGenerationPrompt>;
-export declare function chat(): TextGenerationPromptTemplate<ChatPrompt, LlamaCppTextGenerationPrompt>;
+export declare function instruction(): TextGenerationPromptTemplate<InstructionPrompt, LlamaCppCompletionPrompt>;
+export declare function chat(): TextGenerationPromptTemplate<ChatPrompt, LlamaCppCompletionPrompt>;

package/model-provider/llamacpp/{LlamaCppTextGenerationModel.cjs → LlamaCppCompletionModel.cjs}

@@ -1,6 +1,6 @@
 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.LlamaCppTextGenerationResponseFormat = exports.LlamaCppTextGenerationModel = void 0;
+exports.LlamaCppCompletionResponseFormat = exports.LlamaCppCompletionModel = void 0;
 const zod_1 = require("zod");
 const callWithRetryAndThrottle_js_1 = require("../../core/api/callWithRetryAndThrottle.cjs");
 const postToApi_js_1 = require("../../core/api/postToApi.cjs");
@@ -14,7 +14,7 @@ const parseEventSourceStream_js_1 = require("../../util/streaming/parseEventSour
 const LlamaCppApiConfiguration_js_1 = require("./LlamaCppApiConfiguration.cjs");
 const LlamaCppError_js_1 = require("./LlamaCppError.cjs");
 const LlamaCppTokenizer_js_1 = require("./LlamaCppTokenizer.cjs");
-class LlamaCppTextGenerationModel extends AbstractModel_js_1.AbstractModel {
+class LlamaCppCompletionModel extends AbstractModel_js_1.AbstractModel {
     constructor(settings = {}) {
         super({ settings });
         Object.defineProperty(this, "provider", {
@@ -56,10 +56,10 @@ class LlamaCppTextGenerationModel extends AbstractModel_js_1.AbstractModel {
                             data,
                         }))
                         : undefined,
-                    cache_prompt: this.settings.cachePrompt,
                     temperature: this.settings.temperature,
                     top_k: this.settings.topK,
                     top_p: this.settings.topP,
+                    min_p: this.settings.minP,
                     n_predict: this.settings.maxGenerationTokens,
                     n_keep: this.settings.nKeep,
                     stop: this.settings.stopSequences,
@@ -68,12 +68,19 @@ class LlamaCppTextGenerationModel extends AbstractModel_js_1.AbstractModel {
                     repeat_penalty: this.settings.repeatPenalty,
                     repeat_last_n: this.settings.repeatLastN,
                     penalize_nl: this.settings.penalizeNl,
+                    presence_penalty: this.settings.presencePenalty,
+                    frequency_penalty: this.settings.frequencyPenalty,
+                    penalty_prompt: this.settings.penaltyPrompt,
                     mirostat: this.settings.mirostat,
                     mirostat_tau: this.settings.mirostatTau,
                     mirostat_eta: this.settings.mirostatEta,
+                    grammar: this.settings.grammar,
                     seed: this.settings.seed,
                     ignore_eos: this.settings.ignoreEos,
                     logit_bias: this.settings.logitBias,
+                    n_probs: this.settings.nProbs,
+                    cache_prompt: this.settings.cachePrompt,
+                    slot_id: this.settings.slotId,
                 },
                 failedResponseHandler: LlamaCppError_js_1.failedLlamaCppCallResponseHandler,
                 successfulResponseHandler: responseFormat.handler,
@@ -85,22 +92,29 @@ class LlamaCppTextGenerationModel extends AbstractModel_js_1.AbstractModel {
         const eventSettingProperties = [
             ...TextGenerationModel_js_1.textGenerationModelProperties,
             "contextWindowSize",
-            "cachePrompt",
             "temperature",
             "topK",
             "topP",
+            "minP",
             "nKeep",
             "tfsZ",
             "typicalP",
             "repeatPenalty",
             "repeatLastN",
             "penalizeNl",
+            "presencePenalty",
+            "frequencyPenalty",
+            "penaltyPrompt",
             "mirostat",
             "mirostatTau",
             "mirostatEta",
+            "grammar",
             "seed",
             "ignoreEos",
             "logitBias",
+            "nProbs",
+            "cachePrompt",
+            "slotId",
         ];
         return Object.fromEntries(Object.entries(this.settings).filter(([key]) => eventSettingProperties.includes(key)));
     }
@@ -111,7 +125,7 @@ class LlamaCppTextGenerationModel extends AbstractModel_js_1.AbstractModel {
     async doGenerateTexts(prompt, options) {
         const response = await this.callAPI(prompt, {
             ...options,
-            responseFormat: exports.LlamaCppTextGenerationResponseFormat.json,
+            responseFormat: exports.LlamaCppCompletionResponseFormat.json,
         });
         return {
             response,
@@ -135,7 +149,7 @@ class LlamaCppTextGenerationModel extends AbstractModel_js_1.AbstractModel {
     doStreamText(prompt, options) {
         return this.callAPI(prompt, {
             ...options,
-            responseFormat: exports.LlamaCppTextGenerationResponseFormat.deltaIterable,
+            responseFormat: exports.LlamaCppCompletionResponseFormat.deltaIterable,
         });
     }
     extractTextDelta(delta) {
@@ -178,10 +192,10 @@ class LlamaCppTextGenerationModel extends AbstractModel_js_1.AbstractModel {
         });
     }
     withSettings(additionalSettings) {
-        return new LlamaCppTextGenerationModel(Object.assign({}, this.settings, additionalSettings));
+        return new LlamaCppCompletionModel(Object.assign({}, this.settings, additionalSettings));
     }
 }
-exports.LlamaCppTextGenerationModel = LlamaCppTextGenerationModel;
+exports.LlamaCppCompletionModel = LlamaCppCompletionModel;
 const llamaCppTextGenerationResponseSchema = zod_1.z.object({
     content: zod_1.z.string(),
     stop: zod_1.z.literal(true),
@@ -204,7 +218,7 @@ const llamaCppTextGenerationResponseSchema = zod_1.z.object({
         seed: zod_1.z.number(),
         stop: zod_1.z.array(zod_1.z.string()),
         stream: zod_1.z.boolean(),
-        temp: zod_1.z.number(),
+        temperature: zod_1.z.number().optional(), // optional for backwards compatibility
         tfs_z: zod_1.z.number(),
         top_k: zod_1.z.number(),
         top_p: zod_1.z.number(),
@@ -221,7 +235,7 @@ const llamaCppTextGenerationResponseSchema = zod_1.z.object({
         predicted_n: zod_1.z.number(),
         predicted_per_second: zod_1.z.number().nullable(),
         predicted_per_token_ms: zod_1.z.number().nullable(),
-        prompt_ms: zod_1.z.number().nullable(),
+        prompt_ms: zod_1.z.number().nullable().optional(),
         prompt_n: zod_1.z.number(),
         prompt_per_second: zod_1.z.number().nullable(),
         prompt_per_token_ms: zod_1.z.number().nullable(),
@@ -267,7 +281,7 @@ async function createLlamaCppFullDeltaIterableQueue(stream) {
     });
     return queue;
 }
-exports.LlamaCppTextGenerationResponseFormat = {
+exports.LlamaCppCompletionResponseFormat = {
     /**
      * Returns the response as a JSON object.
      */