14 docs tagged with "AI"

RN 0.74+ mobile apps use a two-stage 'frontend fetch + backend proxy' pattern: a backend Route Handler (Next.js / Cloud Function / Cloud Run, your choice) uses @cloudbase/node-sdk's app.ai().createModel('cloudbase').streamText({model: 'deepseek-v4-flash'}) to return a text/plain streaming Response; the RN frontend reads the stream with fetch + body.getReader() (or react-native-fetch-api polyfill / XHR onprogress fallback) and renders text character by character in a FlatList. SecretId / SecretKey must never appear in the JS bundle — RN bundles are plaintext and can be decompiled by anyone.

Use wx.cloud.extend.AI to call DeepSeek / Hunyuan and other CloudBase platform models directly from a Mini Program frontend, consume textStream with for-await for streaming output, no self-hosted LLM gateway required.

Use @cloudbase/node-sdk in a Next.js Route Handler to call CloudBase AI streamText, convert AsyncIterable to ReadableStream for the frontend, and consume the streaming response in a Client Component with fetch + getReader.

When calling the TCB AI large model, you may encounter 429 Too Many Requests or EXCEEDTOKENQUOTA_LIMIT errors. This article introduces common causes and solutions.

claude mcp add cloudbase -- npx -y @cloudbase/cloudbase-mcp@latest — gives Claude Code 36 CloudBase tools (database / cloud functions / static hosting / cloud run / cloud storage / search / AI Agent) so you can go from a single natural-language prompt to a fully deployed CloudBase application.

Write an agent and tools with Mastra 1.x, use mastra build to produce a Hono server, write a multi-stage Dockerfile, and deploy to CloudBase Run with tcb cloudrun deploy — including OpenAI key injection and CloudBase database tool integration.

Write an MCP Server with @modelcontextprotocol/sdk + Hono that supports Streamable HTTP transport, package it with a multi-stage Dockerfile, deploy it to CloudBase Run via tcb cloudrun deploy, and connect Cursor / Claude Code / Windsurf directly via URL.

TCB AI's hunyuan-image image generation model supports generating images from text descriptions. This document describes the parameters supported by the model.

Parse PDF / entire codebases / Excel files into plain text, feed the full content into the prompt, and call CloudBase AI streamText({ model: 'deepseek-v4-pro' }) to answer questions in one shot — skipping embeddings and vector databases, ideal for one-off long-document Q&A.

A Next.js Route Handler receives user-uploaded images, converts them to base64, and calls @cloudbase/node-sdk's app.ai().createModel('cloudbase').generateText with model: 'deepseek-v4-pro' using multimodal messages to get image descriptions, OCR results, and content analysis. Covers single image, multiple images, and image + text prompts.

Use @deepgram/sdk in a CloudBase Cloud Function to call nova-3, transcribe audio from Cloud Storage with timestamps and speaker diarization, and write results back to the database.

Call fal.ai's FLUX model from a CloudBase Cloud Function using @fal-ai/client, save the generated image to Cloud Storage, and return a temporary URL to the frontend.

Use @tavily/core in a CloudBase Cloud Function to call the Tavily Search API, wrap real-time web retrieval for LLM agents into an HTTP-triggered function, and let frontends / Mini Programs call it directly without exposing the API key.

Use @ai-sdk/codemod to auto-migrate 4.x → 6.x, verify streamText / useChat compatibility, and rewrite with ToolLoopAgent + tool approval for a cleaner implementation.