Inngest $ npx @buildinternet/releases get inngest $ npx @buildinternet/releases get inngest-agent-kit Activity Timeline
All 90d 30d
May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May
Releases 9 Avg 3/mo Versions @inngest/use-agent@0.2.0 → @inngest/agent-kit@0.13.2
Latest
@inngest/agent-kit@0.13.2
Tracking since Nov 22, 2024
Tracking since Nov 20, 2023
Changelog file packages/agent-kit/CHANGELOG.md packages/use-agent/CHANGELOG.md Last fetched 21m ago View on GitHub @inngest/agent-kit
0.13.2
Patch Changes
3818d37: Support multiple AsyncContext shapes following an update in inngest@3.45.0
0.13.1
Patch Changes
07ae4dd: Remove unused json-schema-to-zod require from bundled cjs
0.13.0
Minor Changes
c9b0b16: support for latest version of inngest (3.43.1)
integrated azure-openai model from @inngest/ai
migrated to zod v4 and removed zod-to-json-schema package in favor of native z.toJSONSchema()
0.12.1
Patch Changes
4a81376: replacing the static import of json-schema-to-zod with a dynamic import() inside the function where it's used to resolve crashing when loading agentkit in a cjs project using require()
0.12.0
Minor Changes
b175718: # Comprehensive AgentKit Enhancements
Major improvements to AgentKit with enhanced documentation, new API routes, comprehensive UI components, and example applications.
📚 Documentation Enhancements
New Advanced Pattern Guides:
Added legacy-ui-streaming.mdx - Guide for UI streaming with useAgent hook
Added use-chat.mdx - Comprehensive guide for building chat interfaces
Added use-threads.mdx - Documentation for managing conversation threads
Added use-agent.mdx - Updated agent integration patterns
Documentation Reorganization:
Moved UI streaming guides to dedicated UI Integration section
Enhanced advanced patterns with practical examples
Added sequence diagrams and usage guides
⚡ Revolutionary Automatic Event Streaming System
Comprehensive Streaming Architecture:
StreamingContext : Hierarchical context management for network/agent runs with shared sequence counters
Event Schema : 15+ event types covering complete agent lifecycle (run.started, part.created, text.delta, tool calls, HITL, etc.)
Automatic Enrichment : Events auto-enriched with threadId, userId, and context metadata
Sequence Management : Monotonic sequence numbering for perfect event ordering across contexts
Parent/Child Contexts : Seamless context inheritance for agent runs within network runs
Proxy-based Step Wrapper : Transparent integration with Inngest steps without breaking existing code
Best-effort Publishing : Graceful error handling that never breaks agent execution
OpenAI-Compatible IDs : Automatic generation of tool call IDs within OpenAI's 40-character limit
Event Types Supported:
Lifecycle Events : run.started, run.completed, run.failed, run.interrupted
Content Streaming : text.delta, reasoning.delta, data.delta
Tool Integration : tool_call.arguments.delta, tool_call.output.delta
Part Management : part.created, part.completed, part.failed
HITL Support : hitl.requested, hitl.resolved
Metadata & Control : usage.updated, metadata.updated, stream.ended
Developer Experience:
Zero Configuration : Automatic context extraction from network state
Debug Logging : Comprehensive debug output for development
Shared Sequence Counters : Perfect event ordering across multiple contexts
Flexible Publishing : Configurable publish functions for any transport
This streaming system enables real-time UI updates that perfectly match the useAgent hook expectations, creating seamless agent-to-UI communication.
🚀 New API Routes & Backend Features
Chat & Communication:
POST /api/chat - Main chat endpoint with Zod validation and Inngest integration
POST /api/chat/cancel - Chat cancellation with run interruption events
POST /api/approve-tool - Human-in-the-loop tool approval system
POST /api/realtime/token - Real-time subscription token generation
Thread Management:
GET/POST /api/threads - Thread listing and creation with pagination
GET/DELETE/PATCH /api/threads/[threadId] - Individual thread operations
Thread title generation and metadata management
Support for both authenticated and anonymous users
Integration:
/api/inngest/route - Inngest function serving with runAgentChat
PostgresHistoryAdapter integration for persistent storage
🎨 Comprehensive UI Component Library
AI-Specific Elements:
Actions & Action - Interactive action buttons with tooltips
Branch components - Conversation branching and navigation
CodeBlock - Syntax-highlighted code display with copy functionality
Conversation - Chat conversation containers with scroll management
Image - AI-generated image display components
InlineCitation - Citation cards and source referencing
Loader - Loading animations and states
Message components - Message display with avatars and content
PromptInput - Responsive chat input with model selection
Reasoning - Agent reasoning display with streaming support
Sources - Source material display and linking
Suggestion - AI suggestion chips and interactions
Task - Task display and management components
Tool - Tool call display with input/output views
WebPreview - Web page preview components
Chat Interface Components:
Chat - Main chat interface with sidebar integration
EmptyState - Welcome screen with suggestions
ChatHeader - Header with actions and agent information
ShareDialog - Thread sharing functionality
Message parts for all content types (Text, Tool, Data, File, Source, etc.)
MessageActions - Copy, edit, regenerate, like/dislike functionality
MessageEditor - In-place message editing
Sidebar components (Desktop & Mobile) with thread management
Playground & Development Tools:
SqlPlayground - Interactive SQL query interface
SqlEditor - SQL editing with syntax highlighting
EphemeralChat - Client-side only chat for demos
MultiChat - Multiple concurrent chat sessions
Tab management for multiple contexts
UI Primitives & Layout:
Complete shadcn/ui component library integration
Button, Card, Dialog, Sheet, Tabs and 30+ UI primitives
Responsive layouts and mobile-first design
Dark/light theme support with CSS custom properties
🔧 Developer Experience Improvements
Example Applications:
Multi-chat interface for concurrent conversations
SQL playground with chat integration
Thread-based routing (/chat/[threadId])
Responsive design patterns
Build & Configuration:
Next.js App Router integration
Tailwind CSS with custom design system
TypeScript throughout with strict type checking
Component composition patterns
Development Tools:
Hot reload support for rapid development
Comprehensive prop interfaces and documentation
Modular component architecture
Mobile-responsive design patterns
🎯 Key Benefits
Faster Development : Pre-built components reduce implementation time
Consistent UX : Unified design system across all AgentKit applications
Production Ready : Battle-tested components with proper error handling
Flexible Architecture : Composable components for custom implementations
Enhanced Documentation : Clear guides for common integration patterns
This release significantly enhances the AgentKit ecosystem with production-ready tools for building sophisticated AI chat applications.
0.11.0
Minor Changes
81c90df: # Comprehensive AgentKit Enhancements
Major improvements to AgentKit with enhanced documentation, new API routes, comprehensive UI components, and example applications.
📚 Documentation Enhancements
New Advanced Pattern Guides:
Added legacy-ui-streaming.mdx - Guide for UI streaming with useAgent hook
Added use-chat.mdx - Comprehensive guide for building chat interfaces
Added use-threads.mdx - Documentation for managing conversation threads
Added use-agent.mdx - Updated agent integration patterns
Documentation Reorganization:
Moved UI streaming guides to dedicated UI Integration section
Enhanced advanced patterns with practical examples
Added sequence diagrams and usage guides
⚡ Revolutionary Automatic Event Streaming System
Comprehensive Streaming Architecture:
StreamingContext : Hierarchical context management for network/agent runs with shared sequence counters
Event Schema : 15+ event types covering complete agent lifecycle (run.started, part.created, text.delta, tool calls, HITL, etc.)
Automatic Enrichment : Events auto-enriched with threadId, userId, and context metadata
Sequence Management : Monotonic sequence numbering for perfect event ordering across contexts
Parent/Child Contexts : Seamless context inheritance for agent runs within network runs
Proxy-based Step Wrapper : Transparent integration with Inngest steps without breaking existing code
Best-effort Publishing : Graceful error handling that never breaks agent execution
OpenAI-Compatible IDs : Automatic generation of tool call IDs within OpenAI's 40-character limit
Event Types Supported:
Lifecycle Events : run.started, run.completed, run.failed, run.interrupted
Content Streaming : text.delta, reasoning.delta, data.delta
Tool Integration : tool_call.arguments.delta, tool_call.output.delta
Part Management : , ,
0.9.0
Minor Changes
d9507fb: Added support for persistent conversation history via HistoryAdapters
Created an example NextJS app with realtime responses and thread management
0.8.4
Patch Changes
fed9545: fixed deserialization of state losing messages and results in Inngest context
0.8.3
Patch Changes
2f56454: fixed issue with openai parser not handling responses with both text and tool call parts
0.8.2
Patch Changes
f09cb8e: Adding recursive check for removing additionalProperties for gemini.
0.8.1
Patch Changes
f476961: Fixed Gemini adapter response parsing & malformed function call handling
0.8.0
Minor Changes
e59c6fd: Added support for StreamableHttp in MCP Client
Patch Changes
43a0745: Removed redundant call to this.listMCPTools(server) as we are now using a promises array to handle multiple servers concurrently
Fixed conditional in MCP client initialization and moved this._mcpClients.push(client) to the beginning of listMCPTools method to prevent duplicate clients from being registered
0.7.3
Patch Changes
5e3e74f: Export types
5e3e74f: Export types from index.ts
0.7.2
Patch Changes
b983424: Add safety checks to openai response parser
bf01b2f: fix(gemini): do not send tools and tool_config if not tools are provided
0.7.1
Patch Changes
f630722: Optimize parallelism in createServer(), removing risk of parallel indexing
0.7.0
Minor Changes
a5f2fea: Refactor AgentResult, and allow conversational history + short term mem
0.6.0
Minor Changes
e32af3d: Implement typed state management
Patch Changes
51a076c: Document typed state, re-add KV for backcompat
7eeadbd: fix(network): add back-compat for defaultRouter
0.5.1
Patch Changes
3257ff2: fix(mcp): emit valid tool name from MCP tools
0.5.0
Minor Changes
9688973: Gemini and Grok support
Patch Changes
13643d9: fix(tools): better support for strict mode + option to opt-out
0.4.1
Patch Changes
cc44a77: chore: update inngest and @inngest/ai to latest
0.4.0
Minor Changes
7f26053: chore: bump @inngest/ai for model hyper params support
Breaking change: anthropic() max_tokens options has been moved in defaultParameters
Patch Changes
c623d8a: fix(models): avoid parallel_tool_calls for o3-mini
0.3.1
Patch Changes
bbb5d1c: Dual publish ESM and CJS
0.3.0
Minor Changes
1da7554: fix(index.ts): remove server export to allow non-Node runtimes
Patch Changes
9a9f500: fix(openai): tools with no parameters
07f2634: feat(models): handle error reponses
0.2.2
Patch Changes
1720646: Resolve being unable to find async ctx when using with inngest
ae56867: Use @inngest/ai and only optionally use step tooling
6b309bb: Allow specifying Inngest functions as tools
4d6b263: Shift to pnpm workspace packages; fix linting
0.2.1
Patch Changes
d40a5c3: fix(adapters/openai): safely parse non-strong tool return value for Function calling
0.2.0
Minor Changes
c8343c0: Add basic AgentKit server to serve agents and networks as Inngest functions for easy testing
ec689b8: fix(models/anthropic): ensure that the last message isn't an assistant one
Patch Changes
d4a0d26: Update description for npm
0.1.2
Patch Changes
4e94cb2: Ensure tools mutate state, not a clone
0.1.1
Patch Changes
9f302e4: Network state flow to agents
0.1.0
Minor Changes
f7158e4: Stepless model/network/agent instantiations
0.0.3
Patch Changes
146d976: Fix README links and code examples
be20dc9: Fix tool usage failing with OpenAI requests
0.0.2
Patch Changes
36169d9: Fix GitHub link and add README.md
0.0.1
Patch Changes
5c78a6a: Initial release!
part.created
part.completed
part.failed
HITL Support : hitl.requested, hitl.resolved
Metadata & Control : usage.updated, metadata.updated, stream.ended
Zero Configuration : Automatic context extraction from network state
Debug Logging : Comprehensive debug output for development
Shared Sequence Counters : Perfect event ordering across multiple contexts
Flexible Publishing : Configurable publish functions for any transport
This streaming system enables real-time UI updates that perfectly match the useAgent hook expectations, creating seamless agent-to-UI communication.
🚀 New API Routes & Backend Features
POST /api/chat - Main chat endpoint with Zod validation and Inngest integration
POST /api/chat/cancel - Chat cancellation with run interruption events
POST /api/approve-tool - Human-in-the-loop tool approval system
POST /api/realtime/token - Real-time subscription token generation
GET/POST /api/threads - Thread listing and creation with pagination
GET/DELETE/PATCH /api/threads/[threadId] - Individual thread operations
Thread title generation and metadata management
Support for both authenticated and anonymous users
/api/inngest/route - Inngest function serving with runAgentChat
PostgresHistoryAdapter integration for persistent storage
🎨 Comprehensive UI Component Library
Actions & Action - Interactive action buttons with tooltips
Branch components - Conversation branching and navigation
CodeBlock - Syntax-highlighted code display with copy functionality
Conversation - Chat conversation containers with scroll management
Image - AI-generated image display components
InlineCitation - Citation cards and source referencing
Loader - Loading animations and states
Message components - Message display with avatars and content
PromptInput - Responsive chat input with model selection
Reasoning - Agent reasoning display with streaming support
Sources - Source material display and linking
Suggestion - AI suggestion chips and interactions
Task - Task display and management components
Tool - Tool call display with input/output views
WebPreview - Web page preview components
Chat Interface Components:
Chat - Main chat interface with sidebar integration
EmptyState - Welcome screen with suggestions
ChatHeader - Header with actions and agent information
ShareDialog - Thread sharing functionality
Message parts for all content types (Text, Tool, Data, File, Source, etc.)
MessageActions - Copy, edit, regenerate, like/dislike functionality
MessageEditor - In-place message editing
Sidebar components (Desktop & Mobile) with thread management
Playground & Development Tools:
SqlPlayground - Interactive SQL query interface
SqlEditor - SQL editing with syntax highlighting
EphemeralChat - Client-side only chat for demos
MultiChat - Multiple concurrent chat sessions
Tab management for multiple contexts
Complete shadcn/ui component library integration
Button, Card, Dialog, Sheet, Tabs and 30+ UI primitives
Responsive layouts and mobile-first design
Dark/light theme support with CSS custom properties
🔧 Developer Experience Improvements
Multi-chat interface for concurrent conversations
SQL playground with chat integration
Thread-based routing (/chat/[threadId])
Responsive design patterns
Next.js App Router integration
Tailwind CSS with custom design system
TypeScript throughout with strict type checking
Component composition patterns
Hot reload support for rapid development
Comprehensive prop interfaces and documentation
Modular component architecture
Mobile-responsive design patterns
🎯 Key Benefits
Faster Development : Pre-built components reduce implementation time
Consistent UX : Unified design system across all AgentKit applications
Production Ready : Battle-tested components with proper error handling
Flexible Architecture : Composable components for custom implementations
Enhanced Documentation : Clear guides for common integration patterns
This release significantly enhances the AgentKit ecosystem with production-ready tools for building sophisticated AI chat applications.