[{"_1":2,"_181":-5,"_182":-5},"loaderData",{"_3":4,"_93":94},"root",{"_5":6,"_36":37,"_66":67},"ctx",{"_7":8,"_23":-5,"_24":25,"_26":27,"_28":20,"_29":20,"_30":20,"_31":-5,"_32":33,"_34":20,"_35":20},"auth",{"_9":-5,"_10":-5,"_11":-5,"_12":-5,"_13":-5,"_14":-5,"_15":-5,"_16":-5,"_17":-5,"_18":-5,"_19":20,"_21":22},"sessionClaims","sessionId","sessionStatus","userId","actor","orgId","orgRole","orgSlug","orgPermissions","factorVerificationAge","isAuthenticated",false,"tokenType","session_token","user","preferencesKeybindings","default","preferencesTheme","light","preferencesFixers","preferencesCopilot","preferencesSelectionAi","preferencesChatPanelState","notificationCount",0,"flagDenols","isTeamOrgMember","ENV",{"_38":39,"_40":41,"_42":43,"_44":45,"_46":47,"_48":49,"_50":51,"_52":53,"_54":55,"_56":57,"_58":59,"_60":61,"_62":63,"_64":65},"REACT_APP_SENTRY_DSN","https://586020d86f05489db63a554690acf1e9@o1398359.ingest.sentry.io/6724499","VAL_RUNTIME_NAME","prod","RENDER_GIT_COMMIT","08df2158b567c8dc30dd332965dd8203fb644318","RENDER_GIT_BRANCH","main","CLERK_PUBLISHABLE_KEY","pk_live_Y2xlcmsudmFsLnRvd24k","POSTHOG_API_TOKEN","phc_nmoyuAkStEZuUp69rYlg1DZgrIt19Gx0Y4Hwf8tGgJP","POSTHOG_API_HOST","https://wart.val.town","WEB_API_TEMPLATE","https://{val}.web.val.run","APP_URL","https://www.val.town","MODULE_URL","https://esm.town","VANITY_WEB_API_TEMPLATE","https://{label}.val.run/","DENOLS_URL","https://lsp.val.town","NODE_ENV","production","API_URL","https://api.val.town","clerkState",{"_68":69},"__internal_clerk_state",{"_70":71,"_72":47,"_73":74,"_75":74,"_76":20,"_77":74,"_78":74,"_79":74,"_80":74,"_81":82,"_91":20,"_92":20},"__clerk_ssr_state",{"_21":22,"_9":-5,"_10":-5,"_11":-5,"_12":-5,"_13":-5,"_14":-5,"_15":-5,"_16":-5,"_17":-5,"_18":-5,"_19":20},"__publishableKey","__proxyUrl","","__domain","__isSatellite","__signInUrl","__signUpUrl","__afterSignInUrl","__afterSignUpUrl","__clerk_debug",{"_83":20,"_19":20,"_84":74,"_85":86,"_87":74,"_88":47,"_89":20,"_90":74},"isSignedIn","proxyUrl","reason","session-token-and-uat-missing","message","publishableKey","isSatellite","domain","__telemetryDisabled","__telemetryDebug","routes/embed.x.$handle.$name.$",{"_95":96,"_97":98,"_99":100,"_101":102,"_155":156,"_178":179,"_180":-5},"handle","kamenxrider","name","readwise-mastra","path","lib/llm_pollinations.ts","project",{"_103":104,"_97":98,"_105":106,"_107":-5,"_108":20,"_109":110,"_111":-5,"_112":-5,"_12":113,"_114":-5,"_115":20,"_116":-5,"_117":-5,"_118":33,"_14":119,"_120":121,"_122":123,"_124":125,"_126":125,"_127":128,"_129":128,"_130":128,"_131":125,"_132":125,"_23":133,"_136":137,"_141":142,"_154":20},"id","cbd319ae-ffd3-11f0-bcd8-42dde27851f2","createdAt",["D",1769994557115],"deletedAt","legacyEnvVars","privacy","public","description","imageUrl","d4de366f-8281-484e-8856-3824b6bdfb0b","pinnedFileId","pinPreview","pinMigrated","screenshotVersion","likeCount","9bb2e2b6-8aa1-4b6a-b4ef-ad7adab2b868","tursoName","cbd319aeffd311f0bcd842dde27851f2","tursoHostname","cbd319aeffd311f0bcd842dde27851f2-val-town.aws-us-east-2.turso.io","userPerm","r","valPerm","blobPerm","rw","emailPerm","sqlitePerm","projectPerm","telemetryPerm",{"_95":96,"_134":135},"profileImageUrl","https://img.clerk.com/eyJ0eXBlIjoicHJveHkiLCJzcmMiOiJodHRwczovL2ltYWdlcy5jbGVyay5kZXYvb2F1dGhfZ2l0aHViL2ltZ18ybFJkNm8zSzFzbGE5RzhsTUt3OGJMZFdGZFkifQ","org",{"_95":96,"_138":135,"_12":113,"_139":140},"avatar","tier","pro","branch",{"_143":144,"_97":45,"_103":145,"_146":104,"_147":-5,"_148":-5,"_149":-5,"_150":-5,"_151":-5,"_152":98,"_153":96},"version",51,"cbd4b750-ffd3-11f0-bcd8-42dde27851f2","projectId","forkedBranchId","forkedBranchVersion","forkedProjectPrivacy","forkedProjectId","forkedProjectOrgId","projectName","userHandle","isOwned","file",{"_103":157,"_97":158,"_159":160,"_161":162,"_163":145,"_164":165,"_166":167,"_105":168,"_143":169,"_170":171,"_172":173,"_174":20,"_175":-5,"_99":176},"c0237f64-ffd7-11f0-b645-42dde27851f2","llm_pollinations.ts","fileId","d729e61a-ffd4-11f0-953d-42dde27851f2","parentId","99160818-ffd4-11f0-97f3-42dde27851f2","branchId","branchVersion",40,"latest",true,["D",1769996255498],1,"fileType","script","content","/**\n * Pollinations LLM wrapper with model discovery and smart fallbacks\n * Canonical endpoint: https://gen.pollinations.ai/v1/chat/completions\n * Docs: https://github.com/pollinations/pollinations/blob/main/APIDOCS.md\n */\n\nimport { loadEnv } from \"./env.ts\";\nimport {\n upsertModelCache,\n getModelCache,\n getModelByName,\n clearModelCache,\n getModelConfig,\n setModelConfig,\n getAllModelConfig,\n runMigrations,\n type ModelCache,\n} from \"./db.ts\";\n\n// ============ Types ============\n\nexport interface ChatMessage {\n role: \"system\" | \"user\" | \"assistant\";\n content: string | MessageContentPart[];\n}\n\nexport interface MessageContentPart {\n type: \"text\" | \"image_url\";\n text?: string;\n image_url?: { url: string; detail?: \"auto\" | \"low\" | \"high\" };\n}\n\nexport interface ChatCompletionRequest {\n model: string;\n messages: ChatMessage[];\n temperature?: number;\n max_tokens?: number;\n top_p?: number;\n stream?: boolean;\n response_format?: { type: \"text\" | \"json_object\" };\n}\n\nexport interface ChatCompletionChoice {\n index: number;\n message: {\n role: \"assistant\";\n content: string | null;\n content_blocks?: ContentBlock[];\n };\n finish_reason: string;\n}\n\nexport interface ContentBlock {\n type: \"text\" | \"image_url\" | \"thinking\";\n text?: string;\n image_url?: { url: string };\n}\n\nexport interface ChatCompletionResponse {\n id: string;\n object: string;\n created: number;\n model: string;\n choices: ChatCompletionChoice[];\n usage?: {\n prompt_tokens: number;\n completion_tokens: number;\n total_tokens: number;\n };\n}\n\nexport interface GenerateResult {\n text: string;\n raw: ChatCompletionResponse;\n modelUsed: string;\n fallbacksAttempted: string[];\n}\n\nexport interface GenerateOptions {\n model?: string;\n temperature?: number;\n maxTokens?: number;\n useTools?: boolean;\n complexTask?: boolean;\n retryWithFallback?: boolean;\n}\n\n// ============ Model Discovery Types ============\n\ninterface PollinationsTextModel {\n name: string;\n aliases?: string[];\n description?: string;\n tools?: boolean;\n reasoning?: boolean;\n context_window?: number;\n pricing?: Record;\n input_modalities?: string[];\n output_modalities?: string[];\n}\n\ninterface PollinationsImageModel {\n name: string;\n aliases?: string[];\n description?: string;\n pricing?: Record;\n}\n\ninterface OpenAIModelListResponse {\n object: string;\n data: Array<{ id: string; object: string; created: number }>;\n}\n\n// ============ Constants ============\n\nconst DEFAULT_TEXT_MODEL_PRIORITY = [\"gemini\", \"openai\", \"gemini-large\"];\nconst DEFAULT_IMAGE_MODEL = \"flux\";\n\n// ============ Core API Functions ============\n\n/**\n * Get the full chat completions URL\n */\nfunction getChatCompletionsUrl(): string {\n const env = loadEnv();\n return `${env.POLLINATIONS_BASE_URL}${env.POLLINATIONS_CHAT_PATH}`;\n}\n\n/**\n * Get authorization headers\n */\nfunction getHeaders(): HeadersInit {\n const env = loadEnv();\n return {\n \"Content-Type\": \"application/json\",\n \"Authorization\": `Bearer ${env.POLLINATIONS_TOKEN}`,\n };\n}\n\n// ============ Model Discovery ============\n\n/**\n * Discover available text models from Pollinations\n */\nexport async function discoverTextModels(): Promise {\n const env = loadEnv();\n const baseUrl = env.POLLINATIONS_BASE_URL;\n\n // First get the OpenAI-compatible list to know what's available\n const modelsUrl = `${baseUrl}/v1/models`;\n const modelsRes = await fetch(modelsUrl, { headers: getHeaders() });\n\n if (!modelsRes.ok) {\n console.error(`[pollinations] Failed to fetch /v1/models: ${modelsRes.status}`);\n return [];\n }\n\n const modelsList: OpenAIModelListResponse = await modelsRes.json();\n const availableIds = new Set(modelsList.data.map((m) => m.id));\n\n // Then get rich metadata from /text/models\n const textModelsUrl = `${baseUrl}/text/models`;\n const textRes = await fetch(textModelsUrl, { headers: getHeaders() });\n\n if (!textRes.ok) {\n console.error(`[pollinations] Failed to fetch /text/models: ${textRes.status}`);\n // Fall back to basic list\n return modelsList.data.map((m) => ({ name: m.id }));\n }\n\n const textModels: PollinationsTextModel[] = await textRes.json();\n\n // Filter to only models that are actually available\n return textModels.filter(\n (m) => availableIds.has(m.name) || m.aliases?.some((a) => availableIds.has(a))\n );\n}\n\n/**\n * Discover available image models from Pollinations\n */\nexport async function discoverImageModels(): Promise {\n const env = loadEnv();\n const url = `${env.POLLINATIONS_BASE_URL}/image/models`;\n\n const res = await fetch(url, { headers: getHeaders() });\n\n if (!res.ok) {\n console.error(`[pollinations] Failed to fetch /image/models: ${res.status}`);\n return [];\n }\n\n return res.json();\n}\n\n/**\n * Refresh the model cache by discovering models from Pollinations\n */\nexport async function refreshModelCache(): Promise<{\n textModels: number;\n imageModels: number;\n}> {\n console.log(\"[pollinations] Refreshing model cache...\");\n\n // Ensure tables exist\n await runMigrations();\n\n const now = new Date().toISOString();\n let textCount = 0;\n let imageCount = 0;\n\n try {\n // Discover and cache text models\n const textModels = await discoverTextModels();\n for (const model of textModels) {\n await upsertModelCache({\n name: model.name,\n type: \"text\",\n aliases: model.aliases || null,\n description: model.description || null,\n tools: model.tools || false,\n reasoning: model.reasoning || false,\n context_window: model.context_window || null,\n pricing_json: model.pricing ? JSON.stringify(model.pricing) : null,\n input_modalities: model.input_modalities || null,\n output_modalities: model.output_modalities || null,\n last_updated: now,\n });\n textCount++;\n }\n\n // Discover and cache image models\n const imageModels = await discoverImageModels();\n for (const model of imageModels) {\n await upsertModelCache({\n name: model.name,\n type: \"image\",\n aliases: model.aliases || null,\n description: model.description || null,\n tools: false,\n reasoning: false,\n context_window: null,\n pricing_json: model.pricing ? JSON.stringify(model.pricing) : null,\n input_modalities: null,\n output_modalities: [\"image\"],\n last_updated: now,\n });\n imageCount++;\n }\n\n // Update last discovery timestamp\n await setModelConfig(\"last_discovery\", now);\n\n // Auto-select primary model if not set\n const currentPrimary = await getModelConfig(\"primary_text_model\");\n if (!currentPrimary) {\n const primary = await selectBestModel({ preferCheap: true });\n if (primary) {\n await setModelConfig(\"primary_text_model\", primary);\n console.log(`[pollinations] Auto-selected primary model: ${primary}`);\n }\n }\n\n // Set fallback models\n const fallbacks = await determineFallbackModels();\n await setModelConfig(\"fallback_text_models\", JSON.stringify(fallbacks));\n\n console.log(`[pollinations] Cached ${textCount} text models, ${imageCount} image models`);\n } catch (err) {\n console.error(\"[pollinations] Error refreshing model cache:\", err);\n }\n\n return { textModels: textCount, imageModels: imageCount };\n}\n\n/**\n * Determine fallback models based on what's available\n */\nasync function determineFallbackModels(): Promise {\n const cached = await getModelCache(\"text\");\n const available = new Set(cached.map((m) => m.name));\n\n const fallbacks: string[] = [];\n for (const model of DEFAULT_TEXT_MODEL_PRIORITY) {\n if (available.has(model)) {\n fallbacks.push(model);\n }\n }\n\n return fallbacks;\n}\n\n/**\n * Select the best model based on requirements\n */\nexport async function selectBestModel(options: {\n needsTools?: boolean;\n needsReasoning?: boolean;\n minContextWindow?: number;\n preferCheap?: boolean;\n complexTask?: boolean;\n}): Promise {\n const env = loadEnv();\n\n // If explicit override is set, use it\n if (env.POLLINATIONS_PRIMARY_MODEL) {\n return env.POLLINATIONS_PRIMARY_MODEL;\n }\n\n const cached = await getModelCache(\"text\");\n if (cached.length === 0) {\n // No cache, return default\n return \"gemini\";\n }\n\n // Filter by requirements\n let candidates = cached;\n\n if (options.needsTools) {\n candidates = candidates.filter((m) => m.tools);\n }\n if (options.needsReasoning) {\n candidates = candidates.filter((m) => m.reasoning);\n }\n if (options.minContextWindow) {\n candidates = candidates.filter(\n (m) => m.context_window && m.context_window >= options.minContextWindow!\n );\n }\n\n if (candidates.length === 0) {\n // No models match, fall back to default priority\n for (const model of DEFAULT_TEXT_MODEL_PRIORITY) {\n if (cached.some((m) => m.name === model)) {\n return model;\n }\n }\n return cached[0]?.name || \"gemini\";\n }\n\n // For complex tasks, prefer gemini-large if available\n if (options.complexTask) {\n const large = candidates.find((m) => m.name === \"gemini-large\");\n if (large) return large.name;\n }\n\n // For cheap preference, use default priority order\n if (options.preferCheap) {\n for (const model of DEFAULT_TEXT_MODEL_PRIORITY) {\n if (candidates.some((m) => m.name === model)) {\n return model;\n }\n }\n }\n\n return candidates[0].name;\n}\n\n/**\n * Get the current model configuration with fallbacks\n */\nexport async function getModelWithFallbacks(): Promise<{\n primary: string;\n fallbacks: string[];\n}> {\n const env = loadEnv();\n\n // Check for env override first\n if (env.POLLINATIONS_PRIMARY_MODEL) {\n return {\n primary: env.POLLINATIONS_PRIMARY_MODEL,\n fallbacks: DEFAULT_TEXT_MODEL_PRIORITY.filter(\n (m) => m !== env.POLLINATIONS_PRIMARY_MODEL\n ),\n };\n }\n\n // Get from config\n const primary = (await getModelConfig(\"primary_text_model\")) || \"gemini\";\n const fallbacksJson = await getModelConfig(\"fallback_text_models\");\n const fallbacks = fallbacksJson\n ? JSON.parse(fallbacksJson).filter((m: string) => m !== primary)\n : DEFAULT_TEXT_MODEL_PRIORITY.filter((m) => m !== primary);\n\n return { primary, fallbacks };\n}\n\n// ============ Chat Completion with Fallbacks ============\n\n/**\n * Generate a chat completion with automatic fallbacks\n */\nexport async function generateChat(\n messages: ChatMessage[],\n options: GenerateOptions = {}\n): Promise {\n const {\n model: explicitModel,\n temperature = 0.7,\n maxTokens = 2048,\n useTools,\n complexTask,\n retryWithFallback = true,\n } = options;\n\n // Determine which model to use\n let modelToUse: string;\n if (explicitModel) {\n modelToUse = explicitModel;\n } else if (complexTask) {\n modelToUse = (await selectBestModel({ complexTask: true, needsTools: useTools })) || \"gemini\";\n } else if (useTools) {\n modelToUse = (await selectBestModel({ needsTools: true, preferCheap: true })) || \"gemini\";\n } else {\n const config = await getModelWithFallbacks();\n modelToUse = config.primary;\n }\n\n const fallbacksAttempted: string[] = [];\n const config = await getModelWithFallbacks();\n const modelsToTry = [modelToUse, ...config.fallbacks.filter((m) => m !== modelToUse)];\n\n let lastError: Error | null = null;\n\n for (const currentModel of modelsToTry) {\n try {\n const result = await makeCompletionRequest(currentModel, messages, temperature, maxTokens);\n\n return {\n text: extractTextFromResponse(result),\n raw: result,\n modelUsed: currentModel,\n fallbacksAttempted,\n };\n } catch (err) {\n lastError = err instanceof Error ? err : new Error(String(err));\n console.warn(`[pollinations] Model ${currentModel} failed: ${lastError.message}`);\n fallbacksAttempted.push(currentModel);\n\n if (!retryWithFallback) {\n throw lastError;\n }\n // Continue to next fallback\n }\n }\n\n throw lastError || new Error(\"All models failed\");\n}\n\n/**\n * Make the actual API request\n */\nasync function makeCompletionRequest(\n model: string,\n messages: ChatMessage[],\n temperature: number,\n maxTokens: number\n): Promise {\n const url = getChatCompletionsUrl();\n\n const requestBody: ChatCompletionRequest = {\n model,\n messages,\n temperature,\n max_tokens: maxTokens,\n stream: false,\n };\n\n const response = await fetch(url, {\n method: \"POST\",\n headers: getHeaders(),\n body: JSON.stringify(requestBody),\n });\n\n if (!response.ok) {\n const errorText = await response.text();\n throw new PollinationsError(\n `Pollinations API error: ${response.status} - ${errorText}`,\n response.status\n );\n }\n\n return response.json();\n}\n\n/**\n * Extract text content from a completion response\n * Handles both simple content and content_blocks (Gemini)\n */\nfunction extractTextFromResponse(response: ChatCompletionResponse): string {\n const choice = response.choices?.[0];\n if (!choice) return \"\";\n\n // Simple content\n if (typeof choice.message.content === \"string\") {\n return choice.message.content;\n }\n\n // Content blocks (Gemini with code_execution, etc.)\n if (choice.message.content_blocks) {\n const textParts: string[] = [];\n for (const block of choice.message.content_blocks) {\n if (block.type === \"text\" && block.text) {\n textParts.push(block.text);\n }\n }\n return textParts.join(\"\\n\");\n }\n\n return choice.message.content || \"\";\n}\n\n// ============ Convenience Functions ============\n\n/**\n * Simple text generation (single user prompt)\n */\nexport async function generateText(\n prompt: string,\n options: GenerateOptions & { systemPrompt?: string } = {}\n): Promise {\n const messages: ChatMessage[] = [];\n\n if (options.systemPrompt) {\n messages.push({ role: \"system\", content: options.systemPrompt });\n }\n\n messages.push({ role: \"user\", content: prompt });\n\n return generateChat(messages, options);\n}\n\n/**\n * Generate with JSON output (prompts the model to respond with valid JSON)\n */\nexport async function generateJSON(\n prompt: string,\n options: GenerateOptions & { systemPrompt?: string } = {}\n): Promise<{ data: T; raw: ChatCompletionResponse; modelUsed: string }> {\n const systemPrompt = [\n options.systemPrompt || \"\",\n \"\\nYou must respond with valid JSON only. No markdown, no explanation, just the JSON object.\",\n ]\n .filter(Boolean)\n .join(\"\\n\");\n\n const result = await generateText(prompt, {\n ...options,\n systemPrompt,\n temperature: options.temperature ?? 0.3, // Lower temp for structured output\n });\n\n // Try to parse the JSON\n let jsonText = result.text.trim();\n\n // Strip markdown code blocks if present\n if (jsonText.startsWith(\"```json\")) {\n jsonText = jsonText.slice(7);\n } else if (jsonText.startsWith(\"```\")) {\n jsonText = jsonText.slice(3);\n }\n if (jsonText.endsWith(\"```\")) {\n jsonText = jsonText.slice(0, -3);\n }\n jsonText = jsonText.trim();\n\n try {\n const data = JSON.parse(jsonText) as T;\n return { data, raw: result.raw, modelUsed: result.modelUsed };\n } catch (err) {\n throw new PollinationsError(\n `Failed to parse JSON response: ${err}. Raw text: ${result.text.slice(0, 200)}`,\n 0\n );\n }\n}\n\n// ============ Image Generation ============\n\n/**\n * Generate an image and store it in blob storage\n */\nexport async function generateImage(\n prompt: string,\n options: {\n model?: string;\n width?: number;\n height?: number;\n } = {}\n): Promise<{ blobKey: string; url: string; model: string }> {\n const env = loadEnv();\n const model = options.model || DEFAULT_IMAGE_MODEL;\n\n // Build URL with parameters\n const encodedPrompt = encodeURIComponent(prompt);\n let url = `${env.POLLINATIONS_BASE_URL}/image/${encodedPrompt}?model=${model}`;\n\n if (options.width) url += `&width=${options.width}`;\n if (options.height) url += `&height=${options.height}`;\n\n // Fetch the image\n const response = await fetch(url, {\n headers: {\n \"Authorization\": `Bearer ${env.POLLINATIONS_TOKEN}`,\n },\n });\n\n if (!response.ok) {\n throw new PollinationsError(\n `Image generation failed: ${response.status}`,\n response.status\n );\n }\n\n // Get the image data\n const imageData = await response.arrayBuffer();\n const contentType = response.headers.get(\"Content-Type\") || \"image/png\";\n\n // Determine file extension\n const ext = contentType.includes(\"jpeg\") ? \"jpg\" : \"png\";\n\n // Store in blob storage\n const { blob } = await import(\"https://esm.town/v/std/blob\");\n const blobKey = `images/${crypto.randomUUID()}.${ext}`;\n\n await blob.set(blobKey, new Uint8Array(imageData));\n\n // Get the blob URL\n const blobUrl = `https://kamenxrider-readwise-mastra.web.val.run/blob/${blobKey}`;\n\n return {\n blobKey,\n url: blobUrl,\n model,\n };\n}\n\n/**\n * Get available image models\n */\nexport async function getImageModels(): Promise {\n return getModelCache(\"image\");\n}\n\n// ============ Utilities ============\n\n/**\n * Check if model discovery is needed (older than 24 hours)\n */\nexport async function isDiscoveryNeeded(): Promise {\n const lastDiscovery = await getModelConfig(\"last_discovery\");\n if (!lastDiscovery) return true;\n\n const oneDayAgo = Date.now() - 24 * 60 * 60 * 1000;\n return new Date(lastDiscovery).getTime() < oneDayAgo;\n}\n\n/**\n * Ensure model cache is populated (run discovery if needed)\n */\nexport async function ensureModelCache(): Promise {\n if (await isDiscoveryNeeded()) {\n await refreshModelCache();\n }\n}\n\n// ============ Error Class ============\n\nexport class PollinationsError extends Error {\n statusCode: number;\n\n constructor(message: string, statusCode: number) {\n super(message);\n this.name = \"PollinationsError\";\n this.statusCode = statusCode;\n }\n}\n","deleted","label",[177,158],"lib","permalink","https://www.val.town/x/kamenxrider/readwise-mastra/code/lib/llm_pollinations.ts","extension","actionData","errors"]