@nshiab/simple-data-analysis 5.19.11 → 5.19.13

package/esm/deps/jsr.io/@nshiab/journalism-ai/1.2.22/src/ai/askAI.js

@@ -0,0 +1,760 @@
+import { existsSync, mkdirSync, readFileSync, writeFileSync } from "node:fs";
+import process from "node:process";
+import { GenerateContentResponse, GoogleGenAI, HarmBlockThreshold, HarmCategory, ThinkingLevel, } from "@google/genai";
+import { formatNumber, prettyDuration } from "../../../../journalism-format/1.1.9/src/index.js";
+import crypto from "node:crypto";
+import ollama, { Ollama } from "ollama";
+// Implementation
+export default async function askAI(prompt, options = {}) {
+    if (options.screenshotFrom) {
+        throw new Error("The 'screenshotFrom' option has been removed to reduce dependencies. Please take a screenshot yourself and pass it via the 'image' option.");
+    }
+    const start = Date.now();
+    let client;
+    const ollamaVar = options.ollama === true ||
+        options.ollama instanceof Ollama || process.env.OLLAMA;
+    const defaults = {
+        returnJson: options.returnJson || options.schemaJson ? true : false,
+        parseJson: options.returnJson || options.schemaJson ? true : false,
+    };
+    options = { ...defaults, ...options };
+    // Initialize detailed response tracking
+    const detailedResponse = {
+        response: undefined,
+        prompt: prompt,
+        rawResponse: undefined,
+        fromCache: false,
+        model: "",
+        promptTokenCount: 0,
+        outputTokenCount: 0,
+        totalTokens: 0,
+        tokensPerSecond: 0,
+        durationMs: 0,
+        thoughts: "",
+        thoughtsTokenCount: 0,
+    };
+    if (ollamaVar) {
+        client = options.ollama instanceof Ollama ? options.ollama : ollama;
+    }
+    else if (options.vertex || options.apiKey || options.project || options.location) {
+        client = new GoogleGenAI({
+            apiKey: options.apiKey,
+            vertexai: options.vertex,
+            project: options.project,
+            location: options.location,
+        });
+    }
+    else if (process.env.AI_PROJECT && process.env.AI_LOCATION) {
+        client = new GoogleGenAI({
+            vertexai: true,
+            project: process.env.AI_PROJECT,
+            location: process.env.AI_LOCATION,
+        });
+    }
+    else if (process.env.AI_KEY) {
+        client = new GoogleGenAI({
+            apiKey: process.env.AI_KEY,
+        });
+    }
+    if (!client) {
+        throw new Error("No API key or project/location or Ollama found. Please set AI_KEY, AI_PROJECT, AI_LOCATION or OLLAMA in your environment variables or pass them as options.");
+    }
+    const model = options.model ?? process.env.AI_MODEL;
+    if (!model) {
+        throw new Error("Model not specified. Use the AI_MODEL environment variable or pass it as an option.");
+    }
+    detailedResponse.model = model;
+    if (options.verbose) {
+        if (options.systemPrompt) {
+            console.log(`\nSystem prompt:`);
+            console.log(options.systemPrompt);
+        }
+        console.log(`\nPrompt to ${model}:`);
+        console.log(prompt);
+        if (options.schemaJson) {
+            console.log(`JSON schema for response:`);
+            console.log(JSON.stringify(options.schemaJson, null, 2));
+        }
+    }
+    // Google GenAI
+    const contents = [];
+    // Ollama
+    const message = {
+        role: "user",
+        content: "",
+    };
+    let promptToBeSent = prompt;
+    if (options.HTMLFrom) {
+        const urls = Array.isArray(options.HTMLFrom)
+            ? options.HTMLFrom
+            : [options.HTMLFrom];
+        for (const url of urls) {
+            try {
+                const start = options.verbose ? new Date() : null;
+                const response = await fetch(url, {
+                    headers: {
+                        "User-Agent": "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36",
+                    },
+                });
+                const fullHtml = await response.text();
+                const bodyMatch = fullHtml.match(/<body[^>]*>([\s\S]*)<\/body>/i);
+                const html = bodyMatch ? bodyMatch[1] : fullHtml;
+                promptToBeSent += `\n\nHTML content from ${url}:\n${html}`;
+                if (start) {
+                    console.log(`\nRetrieved body HTML from ${url} in ${prettyDuration(start)}`);
+                }
+            }
+            catch (error) {
+                console.log(`Problem retrieving body HTML from ${url}:`, JSON.stringify(error));
+            }
+        }
+    }
+    if (options.text) {
+        const textFiles = Array.isArray(options.text)
+            ? options.text
+            : [options.text];
+        for (const textFile of textFiles) {
+            if (textFile.startsWith("gs://")) {
+                if (!ollamaVar) {
+                    contents.push({
+                        fileData: {
+                            fileUri: textFile,
+                            mimeType: "text/plain",
+                        },
+                    });
+                }
+                else {
+                    throw new Error("Ollama does not support Google Cloud Storage files. Please use local file paths.");
+                }
+            }
+            else {
+                const textContent = readFileSync(textFile, { encoding: "utf-8" });
+                promptToBeSent += `\n\nContent from ${textFile}:\n${textContent}`;
+            }
+        }
+    }
+    if (ollamaVar) {
+        message.content = promptToBeSent;
+    }
+    else {
+        contents.push(promptToBeSent);
+    }
+    if (options.audio) {
+        if (ollamaVar) {
+            throw new Error("Ollama does not support audio files.");
+        }
+        else {
+            const audioFiles = Array.isArray(options.audio)
+                ? options.audio
+                : [options.audio];
+            for (const audioFile of audioFiles) {
+                if (audioFile.startsWith("gs://")) {
+                    contents.push({
+                        fileData: {
+                            fileUri: audioFile,
+                            mimeType: "audio/mp3",
+                        },
+                    });
+                }
+                else {
+                    const base64Audio = readFileSync(audioFile, {
+                        encoding: "base64",
+                    });
+                    contents.push({
+                        inlineData: { data: base64Audio, mimeType: "audio/mp3" },
+                    });
+                }
+            }
+        }
+    }
+    if (options.video) {
+        if (ollamaVar) {
+            throw new Error("Ollama does not support video files.");
+        }
+        else {
+            const videoFiles = Array.isArray(options.video)
+                ? options.video
+                : [options.video];
+            for (const videoFile of videoFiles) {
+                if (videoFile.startsWith("gs://")) {
+                    contents.push({
+                        fileData: {
+                            fileUri: videoFile,
+                            mimeType: "video/mp4",
+                        },
+                    });
+                }
+                else {
+                    const base64Video = readFileSync(videoFile, {
+                        encoding: "base64",
+                    });
+                    contents.push({
+                        inlineData: { data: base64Video, mimeType: "video/mp4" },
+                    });
+                }
+            }
+        }
+    }
+    if (options.pdf) {
+        if (ollamaVar) {
+            throw new Error("Ollama does not support PDF files.");
+        }
+        else {
+            const pdfFiles = Array.isArray(options.pdf) ? options.pdf : [options.pdf];
+            for (const pdfFile of pdfFiles) {
+                if (pdfFile.startsWith("gs://")) {
+                    contents.push({
+                        fileData: {
+                            fileUri: pdfFile,
+                            mimeType: "application/pdf",
+                        },
+                    });
+                }
+                else {
+                    const base64Pdf = readFileSync(pdfFile, { encoding: "base64" });
+                    contents.push({
+                        inlineData: { data: base64Pdf, mimeType: "application/pdf" },
+                    });
+                }
+            }
+        }
+    }
+    if (options.image) {
+        const imageFiles = Array.isArray(options.image)
+            ? options.image
+            : [options.image];
+        if (ollamaVar) {
+            message.images = imageFiles.map((imageFile) => readFileSync(imageFile, {
+                encoding: "base64",
+            }));
+        }
+        else {
+            for (const imageFile of imageFiles) {
+                if (imageFile.startsWith("gs://")) {
+                    contents.push({
+                        fileData: {
+                            fileUri: imageFile,
+                            mimeType: "image/jpeg",
+                        },
+                    });
+                }
+                else {
+                    const base64Image = readFileSync(imageFile, {
+                        encoding: "base64",
+                    });
+                    contents.push({
+                        inlineData: { data: base64Image, mimeType: "image/jpeg" },
+                    });
+                }
+            }
+        }
+    }
+    // Update the prompt in detailedResponse to reflect what was actually sent
+    detailedResponse.prompt = promptToBeSent;
+    const safetyEnabled = options.safetyEnabled ??
+        (options.vertex ? false : true);
+    const safetySettings = safetyEnabled === false
+        ? [
+            {
+                category: HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT,
+                threshold: HarmBlockThreshold.BLOCK_NONE,
+            },
+            {
+                category: HarmCategory.HARM_CATEGORY_HARASSMENT,
+                threshold: HarmBlockThreshold.BLOCK_NONE,
+            },
+            {
+                category: HarmCategory.HARM_CATEGORY_HATE_SPEECH,
+                threshold: HarmBlockThreshold.BLOCK_NONE,
+            },
+            {
+                category: HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT,
+                threshold: HarmBlockThreshold.BLOCK_NONE,
+            },
+            {
+                category: HarmCategory.HARM_CATEGORY_UNSPECIFIED,
+                threshold: HarmBlockThreshold.BLOCK_NONE,
+            },
+        ]
+        : undefined;
+    // Just everything here
+    const params = {
+        model,
+        contents: contents,
+        messages: [message],
+        system: options.systemPrompt,
+        format: options.schemaJson
+            ? options.schemaJson
+            : options.returnJson
+                ? "json"
+                : undefined,
+        temperature: options.temperature ?? 0,
+        config: {
+            systemInstruction: options.systemPrompt,
+            safetySettings,
+            temperature: options.temperature ?? 0,
+            responseMimeType: options.returnJson ? "application/json" : undefined,
+            responseJsonSchema: options.schemaJson,
+            thinkingConfig: options.thinkingLevel
+                ? {
+                    thinkingLevel: ThinkingLevel[options.thinkingLevel.toUpperCase()],
+                    includeThoughts: options.includeThoughts,
+                }
+                : typeof options.thinkingBudget === "number"
+                    ? {
+                        thinkingBudget: options.thinkingBudget ?? 0,
+                        includeThoughts: options.includeThoughts,
+                    }
+                    : {
+                        thinkingBudget: 0,
+                        includeThoughts: options.includeThoughts,
+                    },
+            tools: options.webSearch
+                ? [
+                    {
+                        googleSearch: {},
+                    },
+                ]
+                : undefined,
+        },
+    };
+    if (safetyEnabled === false) {
+        delete params.config?.safetySettings;
+    }
+    let cacheFileJSON;
+    let cacheFileText;
+    if (options.cache) {
+        const cachePath = "./.journalism-cache";
+        if (!existsSync(cachePath)) {
+            mkdirSync(cachePath);
+        }
+        const hash = crypto
+            .createHash("sha256")
+            // Passing clean too because cleaned data is cached
+            .update(JSON.stringify({ ...params, clean: options.clean }))
+            .digest("hex");
+        cacheFileJSON = `${cachePath}/askAI-${hash}.json`;
+        cacheFileText = `${cachePath}/askAI-${hash}.txt`;
+        if (existsSync(cacheFileJSON)) {
+            const cachedResponse = JSON.parse(readFileSync(cacheFileJSON, "utf-8"));
+            if (options.verbose) {
+                console.log("\nReturning cached JSON response.");
+            }
+            if (options.test) {
+                if (Array.isArray(options.test)) {
+                    options.test.forEach((test) => test(cachedResponse));
+                }
+                else {
+                    options.test(cachedResponse);
+                }
+            }
+            if (options.verbose) {
+                console.log("\nResponse:");
+                console.log(cachedResponse);
+            }
+            if (options.detailedResponse) {
+                return {
+                    response: cachedResponse,
+                    rawResponse: undefined,
+                    fromCache: true,
+                    prompt: prompt,
+                    model: model,
+                    promptTokenCount: 0,
+                    outputTokenCount: 0,
+                    totalTokens: 0,
+                    tokensPerSecond: 0,
+                    durationMs: 0,
+                };
+            }
+            return cachedResponse;
+        }
+        else if (existsSync(cacheFileText)) {
+            const cachedResponse = readFileSync(cacheFileText, "utf-8");
+            if (options.verbose) {
+                console.log("\nReturning cached text response.");
+            }
+            if (options.test) {
+                if (Array.isArray(options.test)) {
+                    options.test.forEach((test) => test(cachedResponse));
+                }
+                else {
+                    options.test(cachedResponse);
+                }
+            }
+            if (options.verbose) {
+                console.log("\nResponse:");
+                console.log(cachedResponse);
+            }
+            if (options.detailedResponse) {
+                return {
+                    response: cachedResponse,
+                    rawResponse: undefined,
+                    fromCache: true,
+                    prompt: prompt,
+                    model: model,
+                    promptTokenCount: 0,
+                    outputTokenCount: 0,
+                    totalTokens: 0,
+                    tokensPerSecond: 0,
+                    durationMs: 0,
+                };
+            }
+            return cachedResponse;
+        }
+        else {
+            if (options.verbose) {
+                console.log("\nCache missed. Generating new response...");
+            }
+        }
+    }
+    const response = client instanceof GoogleGenAI
+        ? await client.models.generateContentStream({
+            ...params,
+            ...(options.geminiParameters ?? {}),
+        })
+        : await client.chat({
+            model,
+            messages: options.systemPrompt
+                ? [{ role: "system", content: options.systemPrompt }, message]
+                : [message],
+            format: params.format,
+            options: {
+                temperature: 0,
+                num_ctx: options.contextWindow,
+            },
+            think: options.thinkingLevel === "minimal"
+                ? "low"
+                : options.thinkingLevel ?? (options.thinkingBudget ?? 0) > 0,
+            ...(options.ollamaParameters ?? {}),
+            stream: true,
+        });
+    let thoughts = "";
+    let returnedResponse = "";
+    let finalUsageMetadata = null;
+    let finalOllamaResponse = null;
+    try {
+        for await (const chunk of response) {
+            if (chunk instanceof GenerateContentResponse) {
+                const candidate = chunk.candidates?.at(0);
+                const parts = candidate?.content?.parts ?? [];
+                // Capture usage metadata from the final chunk
+                if (chunk.usageMetadata) {
+                    finalUsageMetadata = chunk.usageMetadata;
+                }
+                for (const p of parts) {
+                    if (!p.text) {
+                        continue;
+                    }
+                    else if (p.thought) {
+                        if (options.verbose || options.detailedResponse) {
+                            if (options.verbose && !thoughts) {
+                                process.stdout.write("\nThoughts:\n");
+                            }
+                            if (options.verbose) {
+                                process.stdout.write(p.text);
+                            }
+                            thoughts += p.text;
+                        }
+                    }
+                    else {
+                        if (options.verbose) {
+                            if (!returnedResponse) {
+                                process.stdout.write("\nResponse:\n");
+                            }
+                            process.stdout.write(p.text);
+                        }
+                        returnedResponse += p.text;
+                    }
+                }
+            }
+            else {
+                // This is an Ollama response chunk
+                finalOllamaResponse = chunk; // Keep updating with the latest chunk to get final metadata
+                if (chunk.message.thinking) {
+                    if (options.verbose || options.detailedResponse) {
+                        if (options.verbose && !thoughts) {
+                            process.stdout.write("\nThoughts:\n");
+                        }
+                        if (options.verbose) {
+                            process.stdout.write(chunk.message.thinking);
+                        }
+                        thoughts += chunk.message.thinking;
+                    }
+                }
+                else if (chunk.message.content) {
+                    if (options.verbose) {
+                        if (!returnedResponse) {
+                            process.stdout.write("\nResponse:\n");
+                        }
+                        process.stdout.write(chunk.message.content);
+                    }
+                    returnedResponse += chunk.message.content;
+                }
+            }
+        }
+    }
+    finally {
+        // Ensure the response stream is properly closed for Ollama streaming responses
+        if ("abort" in response && typeof response.abort === "function") {
+            response.abort();
+        }
+        if (options.verbose) {
+            process.stdout.write("\n");
+        }
+    }
+    if (options.parseJson) {
+        try {
+            if (typeof returnedResponse === "string") {
+                returnedResponse = JSON.parse(returnedResponse);
+            }
+        }
+        catch (error) {
+            const displayResponse = returnedResponse === ""
+                ? "[empty string]"
+                : returnedResponse;
+            throw new Error(`Failed to parse response as JSON: ${error}.\nResponse: ${displayResponse}`);
+        }
+        if (options.verbose) {
+            console.log("\nParsed JSON response:");
+            console.log(returnedResponse);
+        }
+    }
+    let cleanedResponse = returnedResponse;
+    // Store raw response before cleaning
+    detailedResponse.rawResponse = returnedResponse;
+    if (options.clean) {
+        cleanedResponse = options.clean(returnedResponse);
+    }
+    else {
+        cleanedResponse = returnedResponse;
+    }
+    if (options.test) {
+        if (Array.isArray(options.test)) {
+            options.test.forEach((test) => test(cleanedResponse));
+        }
+        else {
+            options.test(cleanedResponse);
+        }
+    }
+    if (options.cache && options.parseJson && cacheFileJSON) {
+        writeFileSync(cacheFileJSON, JSON.stringify(cleanedResponse));
+        options.verbose && console.log("\nResponse cached as JSON.");
+    }
+    else if (options.cache && cacheFileText) {
+        writeFileSync(cacheFileText, JSON.stringify(cleanedResponse));
+        options.verbose && console.log("\nResponse cached as text.");
+    }
+    if (options.verbose && options.clean) {
+        console.log("\nCleaned response:");
+        console.log(cleanedResponse, "\n");
+    }
+    // Store cleaned response
+    detailedResponse.response = cleanedResponse;
+    if (detailedResponse.rawResponse === detailedResponse.response) {
+        // If no cleaning was done, avoid duplication
+        detailedResponse.rawResponse = undefined;
+    }
+    // Calculate metrics and token usage
+    if ((options.verbose || options.metrics || options.detailedResponse) &&
+        finalUsageMetadata) {
+        // Google GenAI streaming response
+        const hasAudio = options.audio ? true : false;
+        const pricing = [
+            {
+                model: "gemini-3.1-pro-preview",
+                tiers: [
+                    { threshold: 200_000, input: 2.00, output: 12.00 },
+                    { threshold: Infinity, input: 4.00, output: 18.00 },
+                ],
+            },
+            {
+                model: "gemini-3.1-flash-preview",
+                input: hasAudio ? 1.00 : 0.50,
+                output: 3.00,
+            },
+            {
+                model: "gemini-3.1-flash-lite",
+                input: hasAudio ? 0.50 : 0.25,
+                output: 1.50,
+            },
+            {
+                model: "gemini-3-pro-preview",
+                tiers: [
+                    { threshold: 200_000, input: 2.00, output: 12.00 },
+                    { threshold: Infinity, input: 4.00, output: 18.00 },
+                ],
+            },
+            {
+                model: "gemini-3-flash-preview",
+                input: hasAudio ? 1.00 : 0.50,
+                output: 3.00,
+            },
+            {
+                model: "gemini-2.5-pro",
+                tiers: [
+                    { threshold: 200_000, input: 1.25, output: 10.00 },
+                    { threshold: Infinity, input: 2.50, output: 15.00 },
+                ],
+            },
+            {
+                model: "gemini-2.5-flash",
+                input: hasAudio ? 1.00 : 0.30,
+                output: 2.50,
+            },
+            {
+                model: "gemini-2.5-flash-lite",
+                input: hasAudio ? 0.30 : 0.10,
+                output: 0.40,
+            },
+            {
+                model: "gemini-2.0-flash",
+                input: hasAudio ? 0.70 : 0.10,
+                output: 0.40,
+            },
+            {
+                model: "gemini-2.0-flash-lite",
+                input: 0.075,
+                output: 0.30,
+            },
+            {
+                model: "gemini-1.5-pro",
+                tiers: [
+                    { threshold: 128_000, input: 1.25, output: 5.00 },
+                    { threshold: Infinity, input: 2.50, output: 10.00 },
+                ],
+            },
+            {
+                model: "gemini-1.5-flash",
+                tiers: [
+                    { threshold: 128_000, input: 0.075, output: 0.30 },
+                    { threshold: Infinity, input: 0.15, output: 0.60 },
+                ],
+            },
+        ];
+        const modelPricing = pricing.find((p) => p.model === model);
+        if (!modelPricing) {
+            if (options.verbose) {
+                console.log(`\nModel ${model} not found in pricing list.`);
+            }
+        }
+        else {
+            const promptTokenCount = finalUsageMetadata.promptTokenCount ?? 0;
+            const outputTokenCount = finalUsageMetadata.candidatesTokenCount ?? 0;
+            const thoughtsTokenCount = finalUsageMetadata.thoughtsTokenCount ?? 0;
+            let inputRate;
+            let outputRate;
+            if ("tiers" in modelPricing && modelPricing.tiers) {
+                // Find the appropriate tier based on prompt token count
+                const tier = modelPricing.tiers.find((t) => promptTokenCount <= t.threshold) || modelPricing.tiers[modelPricing.tiers.length - 1];
+                inputRate = tier.input;
+                outputRate = tier.output;
+                if (options.verbose) {
+                    const tierDescription = tier.threshold === Infinity
+                        ? `> ${formatNumber(modelPricing.tiers[0].threshold)} tokens`
+                        : `≤ ${formatNumber(tier.threshold)} tokens`;
+                    console.log(`\nPricing tier: ${tierDescription}${hasAudio ? " (audio pricing applied)" : ""}`);
+                }
+            }
+            else if ("input" in modelPricing && "output" in modelPricing) {
+                inputRate = modelPricing.input;
+                outputRate = modelPricing.output;
+            }
+            else {
+                if (options.verbose) {
+                    console.log(`\nInvalid pricing structure for model ${model}.`);
+                }
+                // Still populate basic metadata
+                const durationMs = Date.now() - start;
+                const totalTokens = promptTokenCount + outputTokenCount +
+                    thoughtsTokenCount;
+                const tokensPerSecond = totalTokens / (durationMs / 1000);
+                detailedResponse.promptTokenCount = promptTokenCount;
+                detailedResponse.outputTokenCount = outputTokenCount;
+                detailedResponse.totalTokens = totalTokens;
+                detailedResponse.tokensPerSecond = tokensPerSecond;
+                detailedResponse.durationMs = durationMs;
+                detailedResponse.thoughts = thoughts;
+                detailedResponse.thoughtsTokenCount = thoughtsTokenCount;
+                if (options.detailedResponse) {
+                    return detailedResponse;
+                }
+                else {
+                    return cleanedResponse;
+                }
+            }
+            const promptTokenCost = (promptTokenCount / 1_000_000) * inputRate;
+            const outputTokenCost = (outputTokenCount / 1_000_000) * outputRate;
+            const estimatedCost = promptTokenCost + outputTokenCost;
+            const totalTokens = promptTokenCount + outputTokenCount +
+                thoughtsTokenCount;
+            const durationMs = Date.now() - start;
+            const durationSeconds = durationMs / 1000;
+            const tokensPerSecond = totalTokens / durationSeconds;
+            // Always populate metadata
+            detailedResponse.promptTokenCount = promptTokenCount;
+            detailedResponse.outputTokenCount = outputTokenCount;
+            detailedResponse.totalTokens = totalTokens;
+            detailedResponse.tokensPerSecond = tokensPerSecond;
+            detailedResponse.estimatedCost = estimatedCost;
+            detailedResponse.durationMs = durationMs;
+            detailedResponse.thoughts = thoughts;
+            detailedResponse.thoughtsTokenCount = thoughtsTokenCount;
+            if (options.metrics) {
+                options.metrics.totalCost += estimatedCost;
+                options.metrics.totalInputTokens += promptTokenCount;
+                options.metrics.totalOutputTokens += outputTokenCount;
+                options.metrics.totalRequests += 1;
+            }
+            if (options.verbose) {
+                console.log(`\n\nTokens in:`, formatNumber(detailedResponse.promptTokenCount), "/", "Tokens out:", formatNumber(detailedResponse.outputTokenCount), "/", "Thinking tokens:", formatNumber(detailedResponse.thoughtsTokenCount), "/", "Tokens per second:", formatNumber(detailedResponse.tokensPerSecond, {
+                    significantDigits: 1,
+                }), "/", `Estimated cost${options.webSearch ? " (web search excluded)" : ""}:`, formatNumber(detailedResponse.estimatedCost, {
+                    prefix: "$",
+                    significantDigits: 1,
+                    suffix: " USD",
+                }));
+            }
+        }
+    }
+    else if ((options.verbose || options.metrics || options.detailedResponse) &&
+        finalOllamaResponse) {
+        // Ollama streaming response
+        const promptTokenCount = finalOllamaResponse.prompt_eval_count;
+        const outputTokenCount = finalOllamaResponse.eval_count;
+        const totalTokens = promptTokenCount + outputTokenCount;
+        const durationMs = Date.now() - start;
+        const durationSeconds = durationMs / 1000;
+        const tokensPerSecond = totalTokens / durationSeconds;
+        // Always populate metadata
+        detailedResponse.promptTokenCount = promptTokenCount;
+        detailedResponse.outputTokenCount = outputTokenCount;
+        detailedResponse.totalTokens = totalTokens;
+        detailedResponse.tokensPerSecond = tokensPerSecond;
+        detailedResponse.durationMs = durationMs;
+        detailedResponse.thoughts = thoughts;
+        if (options.metrics) {
+            options.metrics.totalInputTokens += promptTokenCount;
+            options.metrics.totalOutputTokens += outputTokenCount;
+            options.metrics.totalRequests += 1;
+        }
+        if (options.verbose) {
+            console.log(`\n\nTokens in:`, formatNumber(detailedResponse.promptTokenCount), "/", "Tokens out:", formatNumber(detailedResponse.outputTokenCount), "/", "Thinking tokens:", "N/A", "/", "Tokens per second:", formatNumber(detailedResponse.tokensPerSecond, {
+                significantDigits: 1,
+            }));
+        }
+    }
+    else if (options.detailedResponse) {
+        // No token metadata available, just populate duration
+        detailedResponse.durationMs = Date.now() - start;
+        detailedResponse.thoughts = thoughts;
+    }
+    if (options.verbose) {
+        console.log("Execution time:", prettyDuration(start), "\n");
+    }
+    if (options.detailedResponse) {
+        return detailedResponse;
+    }
+    else {
+        return cleanedResponse;
+    }
+}