speechflow 1.7.1 → 2.0.0

package/speechflow-cli/src/speechflow-util-llm.ts

@@ -0,0 +1,367 @@
+/*
+**  SpeechFlow - Speech Processing Flow Graph
+**  Copyright (c) 2024-2025 Dr. Ralf S. Engelschall <rse@engelschall.com>
+**  Licensed under GPL 3.0 <https://spdx.org/licenses/GPL-3.0-only>
+*/
+
+/*  standard dependencies  */
+import EventEmitter                  from "node:events"
+
+/*  external dependencies  */
+import OpenAI                        from "openai"
+import Anthropic                     from "@anthropic-ai/sdk"
+import { GoogleGenAI }               from "@google/genai"
+import { Ollama, type ListResponse } from "ollama"
+import * as Transformers             from "@huggingface/transformers"
+
+/*  own utility types  */
+export type LLMCompleteMessage = {
+    role:           "system" | "user" | "assistant"
+    content:        string
+}
+export type LLMConfig = {
+    provider?:      "openai" | "anthropic" | "google" | "ollama" | "transformers"
+    api?:           string
+    model?:         string
+    key?:           string
+    timeout?:       number
+    temperature?:   number
+    maxTokens?:     number
+    topP?:          number
+    cacheDir?:      string
+}
+export type LLMCompleteOptions = {
+    system?:        string
+    messages?:      LLMCompleteMessage[]
+    prompt:         string
+}
+
+/*  LLM class for unified LLM access  */
+export class LLM extends EventEmitter {
+    /*  internal state  */
+    private config:      Required<LLMConfig>
+    private openai:      OpenAI                              | null = null
+    private anthropic:   Anthropic                           | null = null
+    private google:      GoogleGenAI                         | null = null
+    private ollama:      Ollama                              | null = null
+    private transformer: Transformers.TextGenerationPipeline | null = null
+    private initialized                                             = false
+
+    /*  construct LLM instance  */
+    constructor (config: LLMConfig) {
+        /*  pass-through to EventEmitter  */
+        super()
+
+        /*  provide configuration defaults  */
+        this.config = {
+            provider:      "openai",
+            api:           "",
+            model:         "",
+            key:           "",
+            timeout:       30 * 1000,
+            temperature:   0.7,
+            maxTokens:     1024,
+            topP:          0.5,
+            cacheDir:      "",
+            ...config
+        } as Required<LLMConfig>
+
+        /*  validate configuration options  */
+        if (this.config.key === "") {
+            if (this.config.provider === "openai")
+                this.config.key = process.env.SPEECHFLOW_OPENAI_KEY ?? ""
+            else if (this.config.provider === "anthropic")
+                this.config.key = process.env.SPEECHFLOW_ANTHROPIC_KEY ?? ""
+            else if (this.config.provider === "google")
+                this.config.key = process.env.SPEECHFLOW_GOOGLE_KEY ?? ""
+            if (this.config.provider.match(/^(?:openai|anthropic|google)$/) && this.config.key === "")
+                throw new Error(`API key is required for provider "${this.config.provider}"`)
+        }
+        if (this.config.model === "")
+            throw new Error("model is required")
+    }
+
+    /*  internal logging helper  */
+    private log (level: "info" | "warning" | "error", message: string): void {
+        this.emit("log", level, message)
+    }
+
+    /*  initialize the LLM client  */
+    async open (): Promise<void> {
+        if (this.initialized)
+            return
+        if (this.config.provider === "openai") {
+            /*  instantiate OpenAI API  */
+            this.openai = new OpenAI({
+                ...(this.config.api !== "" ? { baseURL: this.config.api } : {}),
+                apiKey:  this.config.key,
+                timeout: this.config.timeout
+            })
+        }
+        else if (this.config.provider === "anthropic") {
+            /*  instantiate Anthropic API  */
+            this.anthropic = new Anthropic({
+                ...(this.config.api !== "" ? { baseURL: this.config.api } : {}),
+                apiKey:  this.config.key,
+                timeout: this.config.timeout
+            })
+        }
+        else if (this.config.provider === "google") {
+            /*  instantiate Google API  */
+            this.google = new GoogleGenAI({
+                apiKey:      this.config.key,
+                httpOptions: {
+                    timeout: this.config.timeout,
+                    ...(this.config.api !== "" ? { baseUrl: this.config.api } : {})
+                }
+            })
+        }
+        else if (this.config.provider === "ollama") {
+            /*  instantiate Ollama API  */
+            this.ollama = new Ollama({ host: this.config.api })
+
+            /*  ensure the model is available  */
+            let models: ListResponse
+            try {
+                models = await this.ollama.list()
+            }
+            catch (err) {
+                throw new Error(`failed to connect to Ollama API at ${this.config.api}: ${err}`, { cause: err })
+            }
+            const exists = models.models.some((m) => m.name === this.config.model)
+            if (!exists) {
+                this.log("info", `LLM: model "${this.config.model}" still not present in Ollama -- ` +
+                    "automatically downloading model")
+                let artifact = ""
+                let percent  = 0
+                let lastLoggedPercent = -1
+                const interval = setInterval(() => {
+                    if (percent !== lastLoggedPercent) {
+                        this.log("info", `LLM: downloaded ${percent.toFixed(2)}% of artifact "${artifact}"`)
+                        lastLoggedPercent = percent
+                    }
+                }, 1000)
+                try {
+                    const progress = await this.ollama.pull({ model: this.config.model, stream: true })
+                    for await (const event of progress) {
+                        if (event.digest)
+                            artifact = event.digest
+                        if (event.completed && event.total)
+                            percent = (event.completed / event.total) * 100
+                    }
+                }
+                finally {
+                    clearInterval(interval)
+                }
+            }
+        }
+        else if (this.config.provider === "transformers") {
+            /*  track download progress when instantiating Transformers pipeline  */
+            const progressState = new Map<string, number>()
+            const progressCallback: Transformers.ProgressCallback = (progress: any) => {
+                let artifact = this.config.model
+                if (typeof progress.file === "string")
+                    artifact += `:${progress.file}`
+                let percent = 0
+                if (typeof progress.loaded === "number" && typeof progress.total === "number")
+                    percent = (progress.loaded / progress.total) * 100
+                else if (typeof progress.progress === "number")
+                    percent = progress.progress
+                if (percent > 0)
+                    progressState.set(artifact, percent)
+            }
+            const interval = setInterval(() => {
+                for (const [ artifact, percent ] of progressState) {
+                    this.log("info", `LLM: downloaded ${percent.toFixed(2)}% of artifact "${artifact}"`)
+                    if (percent >= 100.0)
+                        progressState.delete(artifact)
+                }
+            }, 1000)
+
+            /*  instantiate HuggingFace Transformers text generation pipeline  */
+            try {
+                const pipelinePromise = Transformers.pipeline("text-generation", this.config.model, {
+                    ...(this.config.cacheDir !== "" ? { cache_dir: this.config.cacheDir } : {}),
+                    dtype:             "q4",
+                    device:            "auto",
+                    progress_callback: progressCallback
+                })
+                this.transformer = await pipelinePromise
+            }
+            catch (err) {
+                throw new Error(`failed to instantiate HuggingFace Transformers pipeline: ${err}`, { cause: err })
+            }
+            finally {
+                clearInterval(interval)
+            }
+            if (this.transformer === null)
+                throw new Error("failed to instantiate HuggingFace Transformers pipeline")
+        }
+        else {
+            const exhaustive: never = this.config.provider
+            throw new Error(`unsupported LLM provider: ${exhaustive}`)
+        }
+        this.log("info", `LLM: initialized ${this.config.provider} client ` +
+            `(${this.config.api !== "" ? `api: ${this.config.api}, ` : ""}model: ${this.config.model})`)
+        this.initialized = true
+    }
+
+    /*  perform a completion  */
+    async complete (options: LLMCompleteOptions): Promise<string> {
+        if (!this.initialized)
+            throw new Error("LLM still not initialized")
+
+        /*  build messages array  */
+        const messages: LLMCompleteMessage[] = []
+        if (options.system)
+            messages.push({ role: "system", content: options.system })
+        if (options.messages)
+            messages.push(...options.messages)
+        messages.push({ role: "user", content: options.prompt })
+
+        /*  perform LLM query  */
+        if (this.config.provider === "openai") {
+            if (!this.openai)
+                throw new Error("OpenAI client not available")
+
+            /*  perform OpenAI chat completion  */
+            const completion = await this.openai.chat.completions.create({
+                model:       this.config.model,
+                max_tokens:  this.config.maxTokens,
+                temperature: this.config.temperature,
+                top_p:       this.config.topP,
+                messages:    messages as OpenAI.ChatCompletionMessageParam[]
+            }).catch((err) => {
+                throw new Error(`failed to perform OpenAI chat completion: ${err}`, { cause: err })
+            })
+            const content = completion?.choices?.[0]?.message?.content
+            if (!content)
+                throw new Error("OpenAI API returned empty content")
+            return content
+        }
+        else if (this.config.provider === "anthropic") {
+            if (!this.anthropic)
+                throw new Error("Anthropic client not available")
+
+            /*  separate system message from other messages for Anthropic API  */
+            const systemMessage = messages.find((m) => m.role === "system")
+            const chatMessages  = messages.filter((m) => m.role !== "system")
+
+            /*  perform Anthropic chat completion  */
+            const message = await this.anthropic.messages.create({
+                model:       this.config.model,
+                max_tokens:  this.config.maxTokens,
+                temperature: this.config.temperature,
+                top_p:       this.config.topP,
+                system:      systemMessage?.content,
+                messages:    chatMessages as Anthropic.MessageParam[]
+            }).catch((err) => {
+                throw new Error(`failed to perform Anthropic chat completion: ${err}`, { cause: err })
+            })
+            const content = message?.content?.[0]
+            if (!content || content.type !== "text")
+                throw new Error("Anthropic API returned empty or non-text content")
+            return content.text
+        }
+        else if (this.config.provider === "google") {
+            if (!this.google)
+                throw new Error("Google client not available")
+
+            /*  convert messages for Google API  */
+            const systemInstruction =
+                messages.find((m) => m.role === "system")?.content
+            const contents =
+                messages.filter((m) => m.role !== "system").map((m) => ({
+                    role:  m.role === "assistant" ? "model" : "user",
+                    parts: [ { text: m.content } ]
+                }))
+
+            /*  perform Google chat completion  */
+            const response = await this.google.models.generateContent({
+                model:    this.config.model,
+                contents,
+                config: {
+                    maxOutputTokens: this.config.maxTokens,
+                    temperature:     this.config.temperature,
+                    topP:            this.config.topP,
+                    ...(systemInstruction ? { systemInstruction } : {})
+                }
+            }).catch((err) => {
+                throw new Error(`failed to perform Google chat completion: ${err}`, { cause: err })
+            })
+            const content = response?.text
+            if (!content)
+                throw new Error("Google API returned empty content")
+            return content
+        }
+        else if (this.config.provider === "ollama") {
+            if (!this.ollama)
+                throw new Error("Ollama client not available")
+
+            /*  perform Ollama chat completion  */
+            const response = await this.ollama.chat({
+                model:      this.config.model,
+                messages,
+                keep_alive: "10m",
+                options: {
+                    num_predict: this.config.maxTokens,
+                    temperature: this.config.temperature,
+                    top_p:       this.config.topP
+                }
+            }).catch((err) => {
+                throw new Error(`failed to perform Ollama chat completion: ${err}`, { cause: err })
+            })
+            const content = response?.message?.content
+            if (!content)
+                throw new Error("Ollama API returned empty content")
+            return content
+        }
+        else if (this.config.provider === "transformers") {
+            if (!this.transformer)
+                throw new Error("HuggingFace Transformers pipeline not available")
+
+            /*  perform HuggingFace Transformers text generation  */
+            const result = await this.transformer(messages, {
+                max_new_tokens:  this.config.maxTokens,
+                temperature:     this.config.temperature,
+                top_p:           this.config.topP,
+                do_sample:       true
+            }).catch((err) => {
+                throw new Error(`failed to perform HuggingFace Transformers text generation: ${err}`, { cause: err })
+            })
+            const single = Array.isArray(result) ? result[0] : result
+            const generatedText = (single as Transformers.TextGenerationSingle).generated_text
+            const content = typeof generatedText === "string" ?
+                generatedText :
+                generatedText.at(-1)?.content
+            if (!content)
+                throw new Error("HuggingFace Transformers API returned empty content")
+            return content
+        }
+        else {
+            const exhaustive: never = this.config.provider
+            throw new Error(`unsupported LLM provider: ${exhaustive}`)
+        }
+    }
+
+    /*  close the LLM client  */
+    async close (): Promise<void> {
+        if (!this.initialized)
+            return
+        if (this.config.provider === "openai")
+            this.openai = null
+        else if (this.config.provider === "anthropic")
+            this.anthropic = null
+        else if (this.config.provider === "google")
+            this.google = null
+        else if (this.config.provider === "ollama") {
+            this.ollama?.abort()
+            this.ollama = null
+        }
+        else if (this.config.provider === "transformers") {
+            this.transformer?.dispose()
+            this.transformer = null
+        }
+        this.initialized = false
+    }
+}

package/speechflow-cli/src/speechflow-util.ts

@@ -9,3 +9,4 @@ export * from "./speechflow-util-error"
 export * from "./speechflow-util-stream"
 export * from "./speechflow-util-queue"
 export * from "./speechflow-util-misc"
+export * from "./speechflow-util-llm"

package/speechflow-ui-db/package.json

@@ -18,17 +18,17 @@
         "luxon":                            "3.7.2",
         "@opensumi/reconnecting-websocket": "4.4.0",
         "axios":                            "1.13.2",
-        "typopro-web":                      "4.2.7",
+        "typopro-web":                      "4.2.8",
         "@fortawesome/fontawesome-free":    "7.1.0",
         "patch-package":                    "8.0.1",
         "@rse/stx":                         "1.1.2"
     },
     "devDependencies": {
-        "vite":                             "7.2.6",
-        "typescript-eslint":                "8.48.1",
-        "@typescript-eslint/eslint-plugin": "8.48.1",
-        "@typescript-eslint/parser":        "8.48.1",
-        "@vitejs/plugin-vue":               "6.0.2",
+        "vite":                             "7.2.7",
+        "typescript-eslint":                "8.49.0",
+        "@typescript-eslint/eslint-plugin": "8.49.0",
+        "@typescript-eslint/parser":        "8.49.0",
+        "@vitejs/plugin-vue":               "6.0.3",
         "@rollup/plugin-yaml":              "4.1.2",
         "vite-plugin-node-polyfills":       "0.24.0",
         "vite-svg-loader":                  "5.1.0",
@@ -37,8 +37,8 @@
 
         "@vue/eslint-config-typescript":    "14.6.0",
         "vue-eslint-parser":                "10.2.0",
-        "eslint":                           "9.39.1",
-        "@eslint/js":                       "9.39.1",
+        "eslint":                           "9.39.2",
+        "@eslint/js":                       "9.39.2",
         "neostandard":                      "0.12.2",
         "eslint-plugin-import":             "2.32.0",
         "eslint-plugin-vue":                "10.6.2",
@@ -58,7 +58,7 @@
         "postcss-html":                     "1.8.0",
         "stylus":                           "0.64.0",
         "typescript":                       "5.9.3",
-        "vue-tsc":                          "3.1.7",
+        "vue-tsc":                          "3.1.8",
         "delay-cli":                        "3.0.0",
         "cross-env":                        "10.1.0",
         "serve":                            "14.2.5",

package/speechflow-ui-st/package.json

@@ -19,18 +19,18 @@
         "luxon":                            "3.7.2",
         "@opensumi/reconnecting-websocket": "4.4.0",
         "axios":                            "1.13.2",
-        "typopro-web":                      "4.2.7",
+        "typopro-web":                      "4.2.8",
         "@fortawesome/fontawesome-free":    "7.1.0",
         "patch-package":                    "8.0.1",
         "@rse/stx":                         "1.1.2",
         "animejs":                          "4.2.2"
     },
     "devDependencies": {
-        "vite":                             "7.2.6",
-        "typescript-eslint":                "8.48.1",
-        "@typescript-eslint/eslint-plugin": "8.48.1",
-        "@typescript-eslint/parser":        "8.48.1",
-        "@vitejs/plugin-vue":               "6.0.2",
+        "vite":                             "7.2.7",
+        "typescript-eslint":                "8.49.0",
+        "@typescript-eslint/eslint-plugin": "8.49.0",
+        "@typescript-eslint/parser":        "8.49.0",
+        "@vitejs/plugin-vue":               "6.0.3",
         "@rollup/plugin-yaml":              "4.1.2",
         "vite-plugin-node-polyfills":       "0.24.0",
         "vite-svg-loader":                  "5.1.0",
@@ -39,8 +39,8 @@
 
         "@vue/eslint-config-typescript":    "14.6.0",
         "vue-eslint-parser":                "10.2.0",
-        "eslint":                           "9.39.1",
-        "@eslint/js":                       "9.39.1",
+        "eslint":                           "9.39.2",
+        "@eslint/js":                       "9.39.2",
         "neostandard":                      "0.12.2",
         "eslint-plugin-import":             "2.32.0",
         "eslint-plugin-vue":                "10.6.2",
@@ -60,7 +60,7 @@
         "postcss-html":                     "1.8.0",
         "stylus":                           "0.64.0",
         "typescript":                       "5.9.3",
-        "vue-tsc":                          "3.1.7",
+        "vue-tsc":                          "3.1.8",
         "delay-cli":                        "3.0.0",
         "cross-env":                        "10.1.0",
         "serve":                            "14.2.5",