@just-every/ensemble 0.2.182 → 0.2.183

package/dist/model_providers/gemini.js

@@ -4,7 +4,7 @@ import { BaseModelProvider } from './base_provider.js';
 import { costTracker } from '../utils/cost_tracker.js';
 import { log_llm_error, log_llm_request, log_llm_response } from '../utils/llm_logger.js';
 import { isPaused } from '../utils/pause_controller.js';
-import { appendMessageWithImage, normalizeImageDataUrl, resizeAndTruncateForGemini } from '../utils/image_utils.js';
+import { appendMessageWithImage, normalizeImageDataUrl, resizeAndTruncateForGemini, resizeDataUrl, } from '../utils/image_utils.js';
 import { hasEventHandler } from '../utils/event_controller.js';
 import { truncateLargeValues } from '../utils/truncate_utils.js';
 function convertParameterToGeminiFormat(param) {
@@ -225,6 +225,56 @@ function formatGroundingChunks(chunks) {
         .map((c, i) => `${i + 1}. ${c.web.title || 'Untitled'} – ${c.web.uri}`)
         .join('\n');
 }
+function normalizeGroundingChunk(chunk) {
+    if (!chunk || typeof chunk !== 'object')
+        return null;
+    const webUri = chunk?.web?.uri;
+    const webTitle = chunk?.web?.title;
+    const imageUri = chunk?.image?.imageUri || chunk?.image?.image_uri || chunk?.image_uri;
+    const imageLandingUri = chunk?.image?.uri || chunk?.uri;
+    const uri = webUri || imageLandingUri;
+    if (!uri && !imageUri)
+        return null;
+    return {
+        ...(uri ? { uri } : {}),
+        ...(imageUri ? { image_uri: imageUri } : {}),
+        ...(webTitle ? { title: webTitle } : {}),
+    };
+}
+function dedupeGroundingChunks(chunks) {
+    const seen = new Set();
+    const out = [];
+    for (const chunk of chunks) {
+        const key = `${chunk.uri || ''}|${chunk.image_uri || ''}|${chunk.title || ''}`;
+        if (seen.has(key))
+            continue;
+        seen.add(key);
+        out.push(chunk);
+    }
+    return out;
+}
+function mergeImageMetadata(target, source) {
+    const next = {
+        ...target,
+        model: source.model || target.model,
+    };
+    if (source.grounding) {
+        const t = target.grounding || {};
+        const s = source.grounding;
+        next.grounding = {
+            ...t,
+            ...s,
+            imageSearchQueries: Array.from(new Set([...(t.imageSearchQueries || []), ...(s.imageSearchQueries || [])])),
+            webSearchQueries: Array.from(new Set([...(t.webSearchQueries || []), ...(s.webSearchQueries || [])])),
+            groundingChunks: dedupeGroundingChunks([...(t.groundingChunks || []), ...(s.groundingChunks || [])]),
+            groundingSupports: [...(t.groundingSupports || []), ...(s.groundingSupports || [])],
+        };
+    }
+    next.thought_signatures = Array.from(new Set([...(target.thought_signatures || []), ...(source.thought_signatures || [])]));
+    next.thoughts = [...(target.thoughts || []), ...(source.thoughts || [])];
+    next.citations = dedupeGroundingChunks([...(target.citations || []), ...(source.citations || [])]);
+    return next;
+}
 async function addImagesToInput(input, images, source) {
     for (const [image_id, imageData] of Object.entries(images)) {
         const processedImageData = await resizeAndTruncateForGemini(imageData);
@@ -396,6 +446,54 @@ const THINKING_BUDGET_CONFIGS = {
     '-high': 12288,
     '-max': 24576,
 };
+const GEMINI_31_FLASH_IMAGE_05K_DIMENSIONS = {
+    '1:1': { width: 512, height: 512 },
+    '1:4': { width: 256, height: 1024 },
+    '1:8': { width: 192, height: 1536 },
+    '2:3': { width: 424, height: 632 },
+    '3:2': { width: 632, height: 424 },
+    '3:4': { width: 448, height: 600 },
+    '4:1': { width: 1024, height: 256 },
+    '4:3': { width: 600, height: 448 },
+    '4:5': { width: 464, height: 576 },
+    '5:4': { width: 576, height: 464 },
+    '8:1': { width: 1536, height: 192 },
+    '9:16': { width: 384, height: 688 },
+    '16:9': { width: 688, height: 384 },
+    '21:9': { width: 792, height: 168 },
+};
+const GEMINI_3_PRO_IMAGE_DIMENSION_PRESETS = {
+    '1024x1024': { ar: '1:1', imageSize: '1K' },
+    '848x1264': { ar: '2:3', imageSize: '1K' },
+    '1264x848': { ar: '3:2', imageSize: '1K' },
+    '896x1200': { ar: '3:4', imageSize: '1K' },
+    '1200x896': { ar: '4:3', imageSize: '1K' },
+    '928x1152': { ar: '4:5', imageSize: '1K' },
+    '1152x928': { ar: '5:4', imageSize: '1K' },
+    '768x1376': { ar: '9:16', imageSize: '1K' },
+    '1376x768': { ar: '16:9', imageSize: '1K' },
+    '1584x672': { ar: '21:9', imageSize: '1K' },
+    '2048x2048': { ar: '1:1', imageSize: '2K' },
+    '1696x2528': { ar: '2:3', imageSize: '2K' },
+    '2528x1696': { ar: '3:2', imageSize: '2K' },
+    '1792x2400': { ar: '3:4', imageSize: '2K' },
+    '2400x1792': { ar: '4:3', imageSize: '2K' },
+    '1856x2304': { ar: '4:5', imageSize: '2K' },
+    '2304x1856': { ar: '5:4', imageSize: '2K' },
+    '1536x2752': { ar: '9:16', imageSize: '2K' },
+    '2752x1536': { ar: '16:9', imageSize: '2K' },
+    '3168x1344': { ar: '21:9', imageSize: '2K' },
+    '4096x4096': { ar: '1:1', imageSize: '4K' },
+    '3392x5056': { ar: '2:3', imageSize: '4K' },
+    '5056x3392': { ar: '3:2', imageSize: '4K' },
+    '3584x4800': { ar: '3:4', imageSize: '4K' },
+    '4800x3584': { ar: '4:3', imageSize: '4K' },
+    '3712x4608': { ar: '4:5', imageSize: '4K' },
+    '4608x3712': { ar: '5:4', imageSize: '4K' },
+    '3072x5504': { ar: '9:16', imageSize: '4K' },
+    '5504x3072': { ar: '16:9', imageSize: '4K' },
+    '6336x2688': { ar: '21:9', imageSize: '4K' },
+};
 export class GeminiProvider extends BaseModelProvider {
     _client;
     apiKey;
@@ -893,14 +991,45 @@ export class GeminiProvider extends BaseModelProvider {
             if (hasOtherTools) {
                 console.warn('[Gemini] Image generation ignores function tools; only google_web_search is supported.');
             }
+            const explicitWebGrounding = opts?.grounding?.web_search;
+            const explicitImageGrounding = opts?.grounding?.image_search;
+            const enableWebGrounding = explicitWebGrounding ?? hasGoogleWebSearch ?? false;
+            const isGemini31FlashImageModel = model.includes('gemini-3.1-flash-image-preview');
+            const enableImageGrounding = explicitImageGrounding === true && isGemini31FlashImageModel;
+            if (explicitImageGrounding && !isGemini31FlashImageModel) {
+                console.warn('[Gemini] Image Search grounding is only available for gemini-3.1-flash-image-preview. Ignoring image_search=true.');
+            }
+            const includeThoughts = opts?.thinking?.include_thoughts === true;
+            const requestedThinkingLevel = opts?.thinking?.level;
+            const thinkingLevel = requestedThinkingLevel === 'high' ? 'High' : requestedThinkingLevel ? 'Minimal' : undefined;
+            if (requestedThinkingLevel && !isGemini31FlashImageModel) {
+                console.warn('[Gemini] thinking.level is currently supported for gemini-3.1-flash-image-preview only. Ignoring thinking level.');
+            }
             let aspectRatio = '1:1';
             if (opts?.size === 'landscape')
                 aspectRatio = '16:9';
             else if (opts?.size === 'portrait')
                 aspectRatio = '9:16';
             console.log(`[Gemini] Generating ${numberOfImages} image(s) with model ${model}, prompt: "${prompt.substring(0, 100)}${prompt.length > 100 ? '...' : ''}"`);
-            if (model.includes('gemini-2.5-flash-image-preview') || model.includes('gemini-3-pro-image-preview')) {
+            if (model.includes('gemini-2.5-flash-image-preview') ||
+                model.includes('gemini-3.1-flash-image-preview') ||
+                model.includes('gemini-3-pro-image-preview')) {
+                let aggregateMetadata = { model };
                 const sizeMap = {
+                    '1:1': { ar: '1:1' },
+                    '1:4': { ar: '1:4' },
+                    '1:8': { ar: '1:8' },
+                    '2:3': { ar: '2:3' },
+                    '3:2': { ar: '3:2' },
+                    '3:4': { ar: '3:4' },
+                    '4:1': { ar: '4:1' },
+                    '4:3': { ar: '4:3' },
+                    '4:5': { ar: '4:5' },
+                    '5:4': { ar: '5:4' },
+                    '8:1': { ar: '8:1' },
+                    '9:16': { ar: '9:16' },
+                    '16:9': { ar: '16:9' },
+                    '21:9': { ar: '21:9' },
                     square: { ar: '1:1' },
                     landscape: { ar: '16:9' },
                     portrait: { ar: '9:16' },
@@ -915,20 +1044,85 @@ export class GeminiProvider extends BaseModelProvider {
                     '1024x1792': { ar: '9:16' },
                 };
                 const sm = opts?.size ? sizeMap[String(opts.size)] : undefined;
+                const gemini3ProDimensionPreset = model.includes('gemini-3-pro-image-preview')
+                    ? GEMINI_3_PRO_IMAGE_DIMENSION_PRESETS[String(opts?.size)]
+                    : undefined;
                 const imageConfig = {};
                 if (sm?.ar)
                     imageConfig.aspectRatio = sm.ar;
+                if (gemini3ProDimensionPreset?.ar)
+                    imageConfig.aspectRatio = gemini3ProDimensionPreset.ar;
                 const qualityKey = typeof opts?.quality === 'string' ? opts.quality.toLowerCase() : '';
-                const imageSizeMap = {
-                    low: '1K',
-                    standard: '2K',
-                    medium: '2K',
-                    hd: '4K',
-                    high: '4K',
-                };
-                const imageSize = imageSizeMap[qualityKey];
-                if (imageSize)
-                    imageConfig.imageSize = imageSize;
+                const imageSizeMap = isGemini31FlashImageModel
+                    ? {
+                        low: '0.5K',
+                        standard: '1K',
+                        medium: '2K',
+                        hd: '4K',
+                        high: '4K',
+                    }
+                    : {
+                        low: '1K',
+                        standard: '2K',
+                        medium: '2K',
+                        hd: '4K',
+                        high: '4K',
+                    };
+                let imageSize = imageSizeMap[qualityKey];
+                if (gemini3ProDimensionPreset?.imageSize) {
+                    imageSize = gemini3ProDimensionPreset.imageSize;
+                }
+                if (isGemini31FlashImageModel && opts?.size === '512x512') {
+                    imageSize = '0.5K';
+                }
+                const requestImageSize = imageSize === '0.5K' ? undefined : imageSize;
+                if (requestImageSize)
+                    imageConfig.imageSize = requestImageSize;
+                const thinkingConfig = {};
+                if (opts?.thinking && 'include_thoughts' in opts.thinking) {
+                    thinkingConfig.includeThoughts = includeThoughts;
+                }
+                if (thinkingLevel && isGemini31FlashImageModel) {
+                    thinkingConfig.thinkingLevel = thinkingLevel;
+                }
+                const searchTypes = {};
+                if (enableWebGrounding)
+                    searchTypes.webSearch = {};
+                if (enableImageGrounding)
+                    searchTypes.imageSearch = {};
+                const googleSearchTool = Object.keys(searchTypes).length > 0
+                    ? {
+                        googleSearch: {
+                            searchTypes,
+                        },
+                    }
+                    : undefined;
+                const halfKTargetDimensions = (() => {
+                    if (!isGemini31FlashImageModel || imageSize !== '0.5K')
+                        return undefined;
+                    const ar = imageConfig.aspectRatio || '1:1';
+                    const exactDimensions = GEMINI_31_FLASH_IMAGE_05K_DIMENSIONS[ar];
+                    if (exactDimensions)
+                        return exactDimensions;
+                    const match = /^(\d+):(\d+)$/.exec(ar);
+                    if (!match)
+                        return { width: 512, height: 512 };
+                    const wRatio = Number(match[1]);
+                    const hRatio = Number(match[2]);
+                    if (!Number.isFinite(wRatio) || !Number.isFinite(hRatio) || wRatio <= 0 || hRatio <= 0) {
+                        return { width: 512, height: 512 };
+                    }
+                    if (wRatio >= hRatio) {
+                        return {
+                            width: Math.max(1, Math.round((wRatio / hRatio) * 512)),
+                            height: 512,
+                        };
+                    }
+                    return {
+                        width: 512,
+                        height: Math.max(1, Math.round((hRatio / wRatio) * 512)),
+                    };
+                })();
                 const perImageCost = this.getImageCost(model, imageSize);
                 const makeOne = async () => {
                     const requestParams = {
@@ -989,13 +1183,15 @@ export class GeminiProvider extends BaseModelProvider {
                         config: {
                             responseModalities: [Modality.IMAGE, Modality.TEXT],
                             ...(Object.keys(imageConfig).length ? { imageConfig } : {}),
-                            ...(hasGoogleWebSearch ? { tools: [{ googleSearch: {} }] } : {}),
+                            ...(googleSearchTool ? { tools: [googleSearchTool] } : {}),
+                            ...(Object.keys(thinkingConfig).length ? { thinkingConfig: thinkingConfig } : {}),
                         },
                     };
                     const loggedRequestId = log_llm_request(agent.agent_id || 'default', 'gemini', model, requestParams, new Date(), requestId, agent.tags);
                     finalRequestId = loggedRequestId;
                     const response = await this.client.models.generateContentStream(requestParams);
                     const images = [];
+                    let metadata = { model };
                     let usageMetadata;
                     for await (const chunk of response) {
                         if (chunk.usageMetadata) {
@@ -1004,11 +1200,79 @@ export class GeminiProvider extends BaseModelProvider {
                         if (!chunk.candidates)
                             continue;
                         for (const cand of chunk.candidates) {
+                            const groundingMetadata = cand.groundingMetadata;
+                            if (groundingMetadata) {
+                                const chunks = Array.isArray(groundingMetadata.groundingChunks)
+                                    ? groundingMetadata.groundingChunks
+                                        .map((c) => normalizeGroundingChunk(c))
+                                        .filter((c) => !!c)
+                                    : [];
+                                const searchEntryPoint = groundingMetadata.searchEntryPoint;
+                                const imageSearchQueries = Array.isArray(groundingMetadata.imageSearchQueries)
+                                    ? groundingMetadata.imageSearchQueries
+                                        .map((q) => (typeof q === 'string' ? q : q?.query || q?.text))
+                                        .filter((q) => typeof q === 'string' && q.length > 0)
+                                    : [];
+                                const webSearchQueries = Array.isArray(groundingMetadata.webSearchQueries)
+                                    ? groundingMetadata.webSearchQueries
+                                        .map((q) => (typeof q === 'string' ? q : q?.query || q?.text))
+                                        .filter((q) => typeof q === 'string' && q.length > 0)
+                                    : [];
+                                metadata = mergeImageMetadata(metadata, {
+                                    model,
+                                    grounding: {
+                                        ...(imageSearchQueries.length ? { imageSearchQueries } : {}),
+                                        ...(webSearchQueries.length ? { webSearchQueries } : {}),
+                                        ...(chunks.length ? { groundingChunks: chunks } : {}),
+                                        ...(Array.isArray(groundingMetadata.groundingSupports)
+                                            ? { groundingSupports: groundingMetadata.groundingSupports }
+                                            : {}),
+                                        ...(searchEntryPoint ? { searchEntryPoint } : {}),
+                                    },
+                                    citations: chunks.filter(c => !!c.uri),
+                                });
+                            }
                             const parts = cand.content?.parts || [];
                             for (const part of parts) {
+                                const thoughtSignature = part.thoughtSignature || part.thought_signature;
+                                if (thoughtSignature) {
+                                    metadata = mergeImageMetadata(metadata, {
+                                        model,
+                                        thought_signatures: [thoughtSignature],
+                                    });
+                                }
+                                if (part.thought) {
+                                    if (includeThoughts) {
+                                        const thoughtPart = {
+                                            thought: true,
+                                            type: part.inlineData?.data ? 'image' : 'text',
+                                            ...(part.text ? { text: part.text } : {}),
+                                            ...(part.inlineData?.mimeType ? { mime_type: part.inlineData.mimeType } : {}),
+                                            ...(part.inlineData?.data ? { data: part.inlineData.data } : {}),
+                                            ...(thoughtSignature ? { thought_signature: thoughtSignature } : {}),
+                                        };
+                                        metadata = mergeImageMetadata(metadata, {
+                                            model,
+                                            thoughts: [thoughtPart],
+                                        });
+                                    }
+                                    continue;
+                                }
                                 if (part.inlineData?.data) {
                                     const mime = part.inlineData.mimeType || 'image/png';
-                                    images.push(`data:${mime};base64,${part.inlineData.data}`);
+                                    let imageData = `data:${mime};base64,${part.inlineData.data}`;
+                                    if (halfKTargetDimensions) {
+                                        try {
+                                            imageData = await resizeDataUrl(imageData, halfKTargetDimensions.width, halfKTargetDimensions.height, {
+                                                fit: 'cover',
+                                            });
+                                        }
+                                        catch (resizeError) {
+                                            console.warn('[Gemini] Failed to resize image to 0.5K, returning original image.');
+                                            console.warn(truncateLargeValues(resizeError));
+                                        }
+                                    }
+                                    images.push(imageData);
                                 }
                             }
                         }
@@ -1057,12 +1321,13 @@ export class GeminiProvider extends BaseModelProvider {
                             });
                         }
                     }
-                    return images;
+                    return { images, metadata };
                 };
                 const allImages = [];
                 const calls = Math.max(1, numberOfImages);
                 for (let i = 0; i < calls; i++) {
-                    const imgs = await makeOne();
+                    const { images: imgs, metadata } = await makeOne();
+                    aggregateMetadata = mergeImageMetadata(aggregateMetadata, metadata);
                     for (const img of imgs) {
                         if (allImages.length < numberOfImages)
                             allImages.push(img);
@@ -1070,6 +1335,12 @@ export class GeminiProvider extends BaseModelProvider {
                     if (allImages.length >= numberOfImages)
                         break;
                 }
+                if (aggregateMetadata.grounding?.groundingChunks) {
+                    aggregateMetadata.citations = dedupeGroundingChunks(aggregateMetadata.grounding.groundingChunks.filter(c => !!c.uri));
+                }
+                if (opts?.on_metadata) {
+                    opts.on_metadata(aggregateMetadata);
+                }
                 if (allImages.length === 0) {
                     throw new Error(`No images returned from ${model} model`);
                 }
@@ -1131,7 +1402,16 @@ export class GeminiProvider extends BaseModelProvider {
         }
     }
     getImageCost(model, imageSize) {
-        if (model.includes('gemini-2.5-flash-image-preview')) {
+        if (model.includes('gemini-3.1-flash-image-preview')) {
+            if (imageSize === '4K')
+                return 0.151;
+            if (imageSize === '2K')
+                return 0.101;
+            if (imageSize === '0.5K')
+                return 0.045;
+            return 0.067;
+        }
+        else if (model.includes('gemini-2.5-flash-image-preview')) {
             return 0.039;
         }
         else if (model.includes('gemini-3-pro-image-preview')) {