modelfusion 0.66.1 → 0.67.0

package/model-provider/ollama/OllamaTextGenerationModel.cjs

@@ -72,6 +72,8 @@ class OllamaTextGenerationModel extends AbstractModel_js_1.AbstractModel {
             "system",
             "template",
             "context",
+            "format",
+            "raw",
         ];
         return Object.fromEntries(Object.entries(this.settings).filter(([key]) => eventSettingProperties.includes(key)));
     }
@@ -140,7 +142,7 @@ const ollamaTextStreamingResponseSchema = zod_1.z.discriminatedUnion("done", [
         context: zod_1.z.array(zod_1.z.number()),
     }),
 ]);
-async function callOllamaTextGenerationAPI({ api = new OllamaApiConfiguration_js_1.OllamaApiConfiguration(), abortSignal, responseFormat, prompt, model, contextWindowSize, maxCompletionTokens, mirostat, mirostat_eta, mirostat_tau, num_gpu, num_gqa, num_threads, repeat_last_n, repeat_penalty, seed, stopSequences, temperature, tfs_z, top_k, top_p, system, template, context, }) {
+async function callOllamaTextGenerationAPI({ api = new OllamaApiConfiguration_js_1.OllamaApiConfiguration(), abortSignal, responseFormat, prompt, model, format, contextWindowSize, maxCompletionTokens, mirostat, mirostat_eta, mirostat_tau, num_gpu, num_gqa, num_threads, repeat_last_n, repeat_penalty, seed, stopSequences, temperature, tfs_z, top_k, top_p, system, template, context, raw, }) {
     return (0, postToApi_js_1.postJsonToApi)({
         url: api.assembleUrl(`/api/generate`),
         headers: api.headers,
@@ -148,6 +150,7 @@ async function callOllamaTextGenerationAPI({ api = new OllamaApiConfiguration_js
             stream: responseFormat.stream,
             model,
             prompt,
+            format,
             options: {
                 mirostat,
                 mirostat_eta,
@@ -169,6 +172,7 @@ async function callOllamaTextGenerationAPI({ api = new OllamaApiConfiguration_js
             system,
             template,
             context,
+            raw,
         },
         failedResponseHandler: OllamaError_js_1.failedOllamaCallResponseHandler,
         successfulResponseHandler: responseFormat.handler,

package/model-provider/ollama/OllamaTextGenerationModel.d.ts

@@ -7,27 +7,104 @@ import { Delta } from "../../model-function/Delta.js";
 import { PromptFormatTextStreamingModel } from "../../model-function/generate-text/PromptFormatTextStreamingModel.js";
 import { TextGenerationModelSettings, TextStreamingModel } from "../../model-function/generate-text/TextGenerationModel.js";
 import { TextGenerationPromptFormat } from "../../model-function/generate-text/TextGenerationPromptFormat.js";
+/**
+ * @see https://github.com/jmorganca/ollama/blob/main/docs/api.md#generate-a-completion
+ */
 export interface OllamaTextGenerationModelSettings<CONTEXT_WINDOW_SIZE extends number | undefined> extends TextGenerationModelSettings {
     api?: ApiConfiguration;
+    /**
+     * The name of the model to use. For example, 'mistral'.
+     *
+     * @see https://ollama.ai/library
+     */
     model: string;
+    /**
+     * The temperature of the model. Increasing the temperature will make the model
+     * answer more creatively. (Default: 0.8)
+     */
     temperature?: number;
     /**
      * Specify the context window size of the model that you have loaded in your
-     * Ollama server.
+     * Ollama server. (Default: 2048)
      */
     contextWindowSize?: CONTEXT_WINDOW_SIZE;
+    /**
+     * Enable Mirostat sampling for controlling perplexity.
+     * (default: 0, 0 = disabled, 1 = Mirostat, 2 = Mirostat 2.0)
+     */
     mirostat?: number;
+    /**
+     * Influences how quickly the algorithm responds to feedback from the generated text.
+     * A lower learning rate will result in slower adjustments,
+     * while a higher learning rate will make the algorithm more responsive. (Default: 0.1)
+     */
     mirostat_eta?: number;
+    /**
+     * Controls the balance between coherence and diversity of the output.
+     * A lower value will result in more focused and coherent text. (Default: 5.0)
+     */
     mirostat_tau?: number;
+    /**
+     * The number of GQA groups in the transformer layer. Required for some models,
+     * for example it is 8 for llama2:70b
+     */
     num_gqa?: number;
+    /**
+     * The number of layers to send to the GPU(s). On macOS it defaults to 1 to
+     * enable metal support, 0 to disable.
+     */
     num_gpu?: number;
+    /**
+     * Sets the number of threads to use during computation. By default, Ollama will
+     * detect this for optimal performance. It is recommended to set this value to the
+     * number of physical CPU cores your system has (as opposed to the logical number of cores).
+     */
     num_threads?: number;
+    /**
+     * Sets how far back for the model to look back to prevent repetition.
+     * (Default: 64, 0 = disabled, -1 = num_ctx)
+     */
     repeat_last_n?: number;
+    /**
+     * Sets how strongly to penalize repetitions. A higher value (e.g., 1.5)
+     * will penalize repetitions more strongly, while a lower value (e.g., 0.9)
+     * will be more lenient. (Default: 1.1)
+     */
     repeat_penalty?: number;
+    /**
+     * Sets the random number seed to use for generation. Setting this to a
+     * specific number will make the model generate the same text for the same prompt.
+     * (Default: 0)
+     */
     seed?: number;
+    /**
+     * Tail free sampling is used to reduce the impact of less probable tokens
+     * from the output. A higher value (e.g., 2.0) will reduce the impact more,
+     * while a value of 1.0 disables this setting. (default: 1)
+     */
     tfs_z?: number;
+    /**
+     * Reduces the probability of generating nonsense. A higher value (e.g. 100)
+     * will give more diverse answers, while a lower value (e.g. 10) will be more
+     *  conservative. (Default: 40)
+     */
     top_k?: number;
+    /**
+     * Works together with top-k. A higher value (e.g., 0.95) will lead to more
+     * diverse text, while a lower value (e.g., 0.5) will generate more focused
+     * and conservative text. (Default: 0.9)
+     */
     top_p?: number;
+    /**
+     * When set to true, no formatting will be applied to the prompt and no context
+     * will be returned.
+     */
+    raw?: boolean;
+    /**
+     * The format to return a response in. Currently the only accepted value is 'json'.
+     * Leave undefined to return a string.
+     */
+    format?: "json";
     system?: string;
     template?: string;
     context?: number[];

package/model-provider/ollama/OllamaTextGenerationModel.js

@@ -69,6 +69,8 @@ export class OllamaTextGenerationModel extends AbstractModel {
             "system",
             "template",
             "context",
+            "format",
+            "raw",
         ];
         return Object.fromEntries(Object.entries(this.settings).filter(([key]) => eventSettingProperties.includes(key)));
     }
@@ -136,7 +138,7 @@ const ollamaTextStreamingResponseSchema = z.discriminatedUnion("done", [
         context: z.array(z.number()),
     }),
 ]);
-async function callOllamaTextGenerationAPI({ api = new OllamaApiConfiguration(), abortSignal, responseFormat, prompt, model, contextWindowSize, maxCompletionTokens, mirostat, mirostat_eta, mirostat_tau, num_gpu, num_gqa, num_threads, repeat_last_n, repeat_penalty, seed, stopSequences, temperature, tfs_z, top_k, top_p, system, template, context, }) {
+async function callOllamaTextGenerationAPI({ api = new OllamaApiConfiguration(), abortSignal, responseFormat, prompt, model, format, contextWindowSize, maxCompletionTokens, mirostat, mirostat_eta, mirostat_tau, num_gpu, num_gqa, num_threads, repeat_last_n, repeat_penalty, seed, stopSequences, temperature, tfs_z, top_k, top_p, system, template, context, raw, }) {
     return postJsonToApi({
         url: api.assembleUrl(`/api/generate`),
         headers: api.headers,
@@ -144,6 +146,7 @@ async function callOllamaTextGenerationAPI({ api = new OllamaApiConfiguration(),
             stream: responseFormat.stream,
             model,
             prompt,
+            format,
             options: {
                 mirostat,
                 mirostat_eta,
@@ -165,6 +168,7 @@ async function callOllamaTextGenerationAPI({ api = new OllamaApiConfiguration(),
             system,
             template,
             context,
+            raw,
         },
         failedResponseHandler: failedOllamaCallResponseHandler,
         successfulResponseHandler: responseFormat.handler,

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "modelfusion",
   "description": "Build multimodal applications, chatbots, and agents with JavaScript and TypeScript.",
-  "version": "0.66.1",
+  "version": "0.67.0",
   "author": "Lars Grammel",
   "license": "MIT",
   "keywords": [

package/util/streaming/parseJsonStream.cjs

@@ -8,9 +8,9 @@ function parseJsonStream({ schema, stream, process, onDone, }) {
     }
     return (async () => {
         try {
-            let unprocessedText = "";
             const reader = new ReadableStreamDefaultReader(stream);
             const utf8Decoder = new TextDecoder("utf-8");
+            let unprocessedText = "";
             // eslint-disable-next-line no-constant-condition
             while (true) {
                 const { value: chunk, done } = await reader.read();
@@ -18,8 +18,8 @@ function parseJsonStream({ schema, stream, process, onDone, }) {
                     break;
                 }
                 unprocessedText += utf8Decoder.decode(chunk, { stream: true });
-                const processableLines = unprocessedText.split(/\r\n|\n|\r/g);
-                unprocessedText = processableLines.pop() || "";
+                const processableLines = unprocessedText.split("\n");
+                unprocessedText = processableLines.pop() ?? "";
                 processableLines.forEach(processLine);
             }
             // processing remaining text:

package/util/streaming/parseJsonStream.js

@@ -5,9 +5,9 @@ export function parseJsonStream({ schema, stream, process, onDone, }) {
     }
     return (async () => {
         try {
-            let unprocessedText = "";
             const reader = new ReadableStreamDefaultReader(stream);
             const utf8Decoder = new TextDecoder("utf-8");
+            let unprocessedText = "";
             // eslint-disable-next-line no-constant-condition
             while (true) {
                 const { value: chunk, done } = await reader.read();
@@ -15,8 +15,8 @@ export function parseJsonStream({ schema, stream, process, onDone, }) {
                     break;
                 }
                 unprocessedText += utf8Decoder.decode(chunk, { stream: true });
-                const processableLines = unprocessedText.split(/\r\n|\n|\r/g);
-                unprocessedText = processableLines.pop() || "";
+                const processableLines = unprocessedText.split("\n");
+                unprocessedText = processableLines.pop() ?? "";
                 processableLines.forEach(processLine);
             }
             // processing remaining text: