Loading workspace insights... Statistics interval
7 days30 daysLatest CI Pipeline Executions
3687f97e docs: audit and update Vercel AI SDK comparison
Audited the comparison against current source on both sides.
Vercel AI SDK column (was describing ~v4/v5; current is v6):
- note the comparison targets ai@6.x (ESM-only)
- agent loop: maxSteps -> stopWhen/stepCountIs + the v6 Agent (ToolLoopAgent) class
- Solid is supported by both (no longer a TanStack-only differentiator); Angular stays Vercel-only
- providers: 50+ direct / 100+ via AI Gateway
- MCP client is now stable (@ai-sdk/mcp)
- media: generateImage stable, video/speech/transcribe experimental
- tool calling: acknowledge client execution via onToolCall/addToolResult
- RSC documented as experimental / not the recommended production path
- add devToolsMiddleware, wrapEmbeddingModel
TanStack AI column (understated / new features):
- connection adapters: 6 built-in (incl. XHR for React Native) + fetcher
- middleware hooks: onStructuredOutputConfig, onIteration, onToolPhaseComplete; built-ins contentGuardMiddleware and otelMiddleware
- per-activity provider coverage: image (+Grok/OpenRouter), TTS (+Grok/fal), transcription (+Grok/ElevenLabs/fal); realtime voice across OpenAI/Grok/ElevenLabs
- new: generateAudio, openaiCompatible adapter, structured-output part, client persistence + typed runtime context, headless UI packages, code-mode skills, React Native support; provider-routed mcpTool nuance
- refresh image example to gpt-image-2 (1536x1024) 8017f622 docs: audit and update Vercel AI SDK comparison
Audited the comparison against current source on both sides.
Vercel AI SDK column (was describing ~v4/v5; current is v6):
- note the comparison targets ai@6.x (ESM-only)
- agent loop: maxSteps -> stopWhen/stepCountIs + the v6 Agent (ToolLoopAgent) class
- Solid is supported by both (no longer a TanStack-only differentiator); Angular stays Vercel-only
- providers: 50+ direct / 100+ via AI Gateway
- MCP client is now stable (@ai-sdk/mcp)
- media: generateImage stable, video/speech/transcribe experimental
- tool calling: acknowledge client execution via onToolCall/addToolResult
- RSC documented as experimental / not the recommended production path
- add devToolsMiddleware, wrapEmbeddingModel
TanStack AI column (understated / new features):
- connection adapters: 6 built-in (incl. XHR for React Native) + fetcher
- middleware hooks: onStructuredOutputConfig, onIteration, onToolPhaseComplete; built-ins contentGuardMiddleware and otelMiddleware
- per-activity provider coverage: image (+Grok/OpenRouter), TTS (+Grok/fal), transcription (+Grok/ElevenLabs/fal); realtime voice across OpenAI/Grok/ElevenLabs
- new: generateAudio, openaiCompatible adapter, structured-output part, client persistence + typed runtime context, headless UI packages, code-mode skills, React Native support; provider-routed mcpTool nuance
- refresh image example to gpt-image-2 (1536x1024) 678fb219 docs: audit and update Vercel AI SDK comparison
Audited the comparison against current source on both sides.
Vercel AI SDK column (was describing ~v4/v5; current is v6):
- note the comparison targets ai@6.x (ESM-only)
- agent loop: maxSteps -> stopWhen/stepCountIs + the v6 Agent (ToolLoopAgent) class
- Solid is supported by both (no longer a TanStack-only differentiator); Angular stays Vercel-only
- providers: 50+ direct / 100+ via AI Gateway
- MCP client is now stable (@ai-sdk/mcp)
- media: generateImage stable, video/speech/transcribe experimental
- tool calling: acknowledge client execution via onToolCall/addToolResult
- RSC documented as experimental / not the recommended production path
- add devToolsMiddleware, wrapEmbeddingModel
TanStack AI column (understated / new features):
- connection adapters: 6 built-in (incl. XHR for React Native) + fetcher
- middleware hooks: onStructuredOutputConfig, onIteration, onToolPhaseComplete; built-ins contentGuardMiddleware and otelMiddleware
- per-activity provider coverage: image (+Grok/OpenRouter), TTS (+Grok/fal), transcription (+Grok/ElevenLabs/fal); realtime voice across OpenAI/Grok/ElevenLabs
- new: generateAudio, openaiCompatible adapter, structured-output part, client persistence + typed runtime context, headless UI packages, code-mode skills, React Native support; provider-routed mcpTool nuance
- refresh image example to gpt-image-2 (1536x1024)