Skip to content

Commit 26965cb

Browse files
authored
Adding a full chat example (#19)
* Copying chat example across * MCP tool calling woo! * Moving most of the debug logging behind localStorage.USE_MCP_DEBUG * Working on tool call display * working tool call ui! * deployng to workers.dev for sharing * Simplify chat UI to focus on main input with pastel gradient background - Hide sidebar (conversations, model picker, MCP servers) while preserving components - Hide navbar/title bar while preserving component - Remove title text and description above input field - Add pastel gradient background (pink-purple-indigo) - Expand conversation thread to full width - Create minimal, focused interface with only white input field visible * Add animated rotating conical gradient background - Create large 500% conical gradient with pastel color cycle - Position gradient off-screen (-200% top/left) for smooth rotation - Add 8s linear infinite rotation animation for fast testing - Use ::before pseudo-element to avoid repainting main content - Cycle through various pastel shades (pink, purple, blue, green, yellow, orange) - Performance-optimized: only transforms, no gradient repainting * Switch to hue-rotate animation for background gradient - Revert complex conical gradient approach - Return to original linear gradient (pink-100 → purple-50 → indigo-100) - Add hue-rotate filter animation cycling through 360 degrees - 8s duration for fast testing - Simpler implementation with continuous color shifting - Less control over saturation but smoother performance * Slow down hue-rotate animation to 60 seconds - Change animation duration from 8s to 60s for subtle color shifting - Creates gentle, barely noticeable background color transitions - Maintains visual interest without being distracting * Add random starting point for hue-rotate animation - Generate random negative animation-delay between 0 and -60 seconds - Starts animation at random point in the 60-second cycle - Each page load will show different initial colors - Uses useState with initializer function to set delay once on mount - CSS doesn't have random functions, so JavaScript handles the randomization * Use CSS custom property for cleaner random animation delay - Move animation-delay to CSS using custom property --random-delay - JavaScript still required to generate random value (CSS has no random functions) - Cleaner separation between styling (CSS) and logic (JavaScript) - Maintains same functionality with better organization * Add brain emoji model selector with fullscreen modal - Create ModelSelectionModal component with list-style interface - Add brain emoji (🧠) indicator below chat input, left-aligned - Show current model name in lowercase or red X if not configured - Fullscreen modal displays all available models with configuration status - Green checkmark for configured models, red alert for unconfigured - Integrate with existing API key management and model selection logic - Modal handles API key setup workflow when selecting unconfigured models - Clean separation between compact indicator and full selection interface * Isolate hue-rotate animation to background layer only - Create separate background layer using ::before pseudo-element - Move hue-rotate animation to background layer with z-index: -1 - Remove filter from main container to prevent inheritance - Use equivalent gradient colors in CSS (pink-100, purple-50, indigo-100) - Foreground elements (brain emoji, text) now maintain original colors - Background still animates smoothly without affecting content * Replace red X with grey 'none' text and add pointer cursor - Change unconfigured model indicator from red ✕ to grey 'none' text - Add explicit cursor-pointer class to brain emoji button - Maintains grey text color consistent with configured model names - Cleaner, less alarming visual for unconfigured state * Add MCP server indicator with plug emoji and modal - Create McpServerModal component based on existing McpServers component - Add plug emoji (🔌) indicator on right side, opposite to brain emoji - Show '0' when no servers connected, '1 (N)' when connected with N tools - Fullscreen modal with MCP server configuration UI - Support for single MCP server connection (foundation for multiple servers) - Integrate with existing MCP tools logic and useMcp hook - Clean layout with brain emoji on left, plug emoji on right * Move number to left of plug emoji - Change layout from '🔌 0' to '0 🔌' - Number now appears before the plug icon for better visual balance - Maintains same functionality and styling * Keep MCP modal in React tree to maintain connection - Replace conditional rendering with CSS visibility (hidden/block) - Remove early return null when modal is closed - McpConnection component stays mounted to preserve MCP server connection - Modal is always in DOM but visually hidden when isOpen is false - Prevents MCP disconnection when modal is dismissed * Fix MCP tools count update in plug icon indicator - Add onMcpToolsUpdate prop to ConversationThread interface - Pass setMcpTools function from ChatApp down to ConversationThread - Connect McpServerModal onToolsUpdate to parent state update - Now plug icon will correctly show '1 (N)' when N tools are available - Fixes issue where tool count remained at 0 despite successful connection * Add cursor pointer to X button and click-outside-to-dismiss - Add cursor-pointer class to modal X button - Add onClick handler to backdrop overlay to close modal - Add stopPropagation to modal content to prevent backdrop clicks - Now modal can be dismissed by clicking outside or using X button * Add cursor pointer and click-outside-to-dismiss to model selector modal - Add cursor-pointer class to X button in ModelSelectionModal - Add onClick handler to backdrop overlay to close modal - Add stopPropagation to modal content to prevent backdrop clicks - Consistent modal behavior between model selector and MCP server modals * Implement multiple MCP servers support - Completely rewrite McpServerModal to support multiple servers - Add McpServer interface with id, url, enabled, and name fields - Use localStorage 'mcpServers' array instead of single server URL - Add server list UI with status badges, toggle, and delete buttons - Add 'Add New Server' section at bottom with input and Add button - Support enabling/disabling individual servers with Power/PowerOff icons - Delete servers with Trash2 icon in top right of each server block - Aggregate tools from all enabled servers for parent component - Render multiple McpConnection components for enabled servers - Update plug indicator to show correct server count - Add debug information panel showing server and tool counts - Handle server names using hostname from URL - Each server gets unique connection state tracking by server ID * Update plug indicator to show enabled vs total counts - Change indicator format to show enabled/total for both servers and tools - Format: 'X (Y servers) A (B tools)' where: - X = enabled servers, Y = total servers - A = tools from enabled servers, B = total tools from all servers - Track tool counts per server in localStorage (mcpServerToolCounts) - Persist tool counts even when servers are disabled - Examples: - No servers: '0 (0 servers) 0 (0 tools)' - One disabled server with 2 tools: '0 (1 servers) 0 (2 tools)' - One enabled server with 2 tools: '1 (1 servers) 2 (2 tools)' - Clean up tool counts when servers are deleted * Update plug indicator format with comma separation - Change format from '0 (1 servers) 0 (2 tools)' to '0 (1) servers, 0 (2) tools' - Move words 'servers' and 'tools' outside brackets - Add comma separation between server and tool counts - More concise and readable format * Change plug indicator format to use forward slashes - Change from '0 (1) servers, 0 (2) tools' to '0/1 servers, 0/2 tools' - Use forward slash format showing enabled/total for both servers and tools - Cleaner, more compact visual representation - Examples: - No servers: '0/0 servers, 0/0 tools' - One disabled server with 2 tools: '0/1 servers, 0/2 tools' - One enabled server with 2 tools: '1/1 servers, 2/2 tools' * Soften modal background with radial gradient - Replace solid black 50% opacity with radial gradient - Center: 80% black (rgba(0,0,0,0.8)) - Corners: 70% black (rgba(0,0,0,0.7)) - Apply to both ModelSelectionModal and McpServerModal - Creates softer, more visually appealing backdrop - Remove bg-black bg-opacity-50 classes and use inline style * Replace gradient with uniform 80% black background - Remove radial gradient to fix banding issues - Use solid 80% black background (rgba(0,0,0,0.8)) - Apply to both ModelSelectionModal and McpServerModal - Clean, consistent backdrop without visual artifacts * Implement OAuth callback support for MCP servers - Add react-router-dom dependency for routing support - Create OAuthCallback component with animated background and loading spinner - Add routing to App.tsx with /oauth/callback route - Match styling with main app using animated-bg-container - Use onMcpAuthorization from use-mcp library for OAuth handling - Library already handles multiple server differentiation via state parameter - Each MCP server gets unique OAuth flow identified by state containing server URL - Clean up unused imports and variables in McpServerModal * Deploying to chat.use-mcp.dev * Add Qwen 3 32B model under Groq provider - Add qwen-3-32b model entry using qwen/qwen3-32b model ID - Uses same Groq provider as Llama 3.3, sharing API token - Model appears in selection modal between Llama and Claude - Clean name 'Qwen 3 32B' for user interface * Make gradient background fill entire content height - Change from h-dvh to min-h-screen on main container - Remove overflow-clip to allow content to expand - Update inner containers to use min-h-screen for proper height - Background gradient now extends with content rather than fixed to viewport - Allows for scrolling when content exceeds viewport height * Make gradient background repeat every 100vh - Add background-size: 100% 100vh to fix gradient height to viewport - Add background-repeat: repeat-y to tile gradient vertically - Gradient pattern now repeats down the page as content grows - Animation applies consistently to all repeated background tiles - Creates seamless repeating pattern while maintaining hue rotation * Doing some manual CSS hax for a change * Update AGENT.md with chat UI development guidelines - Add React Router and use-mcp to tech stack - Document background animation best practices (hue-rotate, pseudo-elements) - Modal patterns for UX (click-outside, cursor-pointer, backdrop styling) - MCP server management (multiple servers, localStorage patterns, OAuth) - State persistence guidelines (localStorage vs sessionStorage) - UI indicator patterns (emoji-based, count formatting) - Component organization for stateful elements - Routing and OAuth callback implementation - Performance considerations for animations and aggregation * I am dum. CSS is smrt * BG css tweaks * Added a simple example server for the demo * feat: Add reasoning token support with expandable thinking blocks - Add reasoning field to AssistantMessage type - Create ReasoningBlock component with truncate/expand functionality - Add reasoning-capable models (Qwen QwQ, DeepSeek R1, Qwen 3 32B) - Implement reasoning event handling in stream processing - Enable reasoning format for Groq and Anthropic models - Style reasoning blocks with grey border and reduced font size - Display reasoning above assistant messages when available The reasoning blocks show truncated thinking tokens by default and expand on click to show full reasoning process from models like Qwen3-32b that support step-by-step reasoning output. * fix: Use extractReasoningMiddleware to properly parse thinking blocks - Import extractReasoningMiddleware and wrapLanguageModel from AI SDK - Wrap reasoning-capable models with extractReasoningMiddleware - Use 'think' as tagName to parse <think> XML blocks from models - Enable startWithReasoning option to handle models missing reasoning tags - Extract reasoning from result.reasoning after stream completes - Remove manual reasoning event handling and provider-specific config This fixes the issue where thinking blocks appeared as raw XML tags instead of parsed reasoning content. The middleware automatically extracts reasoning from <think> tags and exposes it via result.reasoning. * Added a simpe Hono server to test * fix: Remove thinking XML tags from main message content - Add regex-based cleaning to remove <think>...</think> tags from display text - Handle both closed and unclosed think tags in streaming content - Apply cleaning only for reasoning-capable models to avoid unnecessary processing - Update conversation content with cleaned response instead of raw response This ensures thinking tags are extracted for reasoning display but don't appear in the main message content, providing a clean user experience. * fix: Improve thinking tag removal with more aggressive cleaning - Add multiple regex passes to handle various think tag scenarios - Remove standalone <think> and </think> tags that might remain - Clean leading whitespace after tag removal - Add debug logging to track text cleaning process - Handle streaming cases where tags appear before closing tags arrive This provides more robust cleaning to ensure no think tags appear in the final message content during streaming scenarios. * fix: Add detailed debugging and use result.text for cleaned content - Add debug logging to track aiResponse vs result.text content - Use result.text from extractReasoningMiddleware for final content - This should provide the properly cleaned text without thinking tags - Comment out unused sidebar and edit functionality to fix TS errors The result.text should contain content that has been automatically cleaned by the extractReasoningMiddleware, while result.reasoning contains the extracted thinking content. * fix: Capture reasoning events during streaming instead of from result - Add back reasoning variable to track content during streaming - Handle reasoning events from extractReasoningMiddleware in stream processing - Build up reasoning content from individual reasoning textDelta events - Remove dependency on result.reasoning which wasn't being populated - Add reasoning to assistant message after stream completes The logs showed reasoning events were coming through correctly but we weren't capturing them. This fixes the issue by collecting reasoning content during streaming as the events arrive. * debug: Add more detailed logging for reasoning event handling - Add debug logs inside reasoning event handler to see if it's being triggered - Log reasoning content length as it's being collected - Add final reasoning content summary at end of stream - Log why reasoning isn't being added if conditions aren't met - Cast event to any to avoid potential TypeScript issues This will help identify why reasoning events aren't being captured despite appearing in the event stream logs. * fix: Clean think tags from reasoning content before display - Add cleaning logic to remove <think> and </think> tags from reasoning - Trim whitespace after tag removal for clean display - Add debug logging to show before/after cleaning - The middleware extracts reasoning but includes the original tags This ensures the reasoning blocks show clean thinking content without the XML tags that were being captured along with the reasoning text. * feat: Implement live reasoning streaming with timing display - Update ReasoningBlock to start expanded and show live reasoning content - Add timing tracking with start/end timestamps for reasoning duration - Auto-collapse reasoning blocks after completion with 500ms delay - Show timing summary (e.g. 'Thought for 0.4s') instead of content when collapsed - Add streaming cursor animation and real-time content updates - Update types to support reasoning timing and streaming state - Stream reasoning content live as tokens arrive during thinking phase This provides a much better UX where users can see the model thinking in real-time, then get a clean timing summary after completion. * fix: Create assistant message immediately when reasoning starts - Create assistant message as soon as first reasoning event arrives - This ensures reasoning block is visible from the beginning of thinking - Prevent duplicate message creation in text-delta handler - Move reasoning end time logic to text-delta start (when thinking finishes) This fixes the issue where reasoning blocks weren't appearing because the assistant message was only created after reasoning completed. * fix: Remove delay for instant reasoning block collapse - Remove 500ms setTimeout delay when reasoning finishes - Reasoning block now collapses immediately when response starts - Creates smoother transition from thinking to response phase - Eliminates jarring delay between reasoning end and collapse The reasoning block now collapses instantly as soon as the model starts generating the actual response, providing better UX flow. * feat: Add sliding window effect for streaming reasoning text - Use overflow:hidden + whitespace:nowrap to prevent line breaks - Apply justify-end flexbox to push content to the right - New tokens push older ones out to the left edge and hide them - Creates smooth sliding window effect showing only recent tokens - Only applies during streaming, keeps normal wrap after completion This creates a nice visual effect where you can see the latest reasoning tokens flowing in from right to left as the model thinks. * Add debug logging and Accept headers for MCP transport debugging - Added extensive debug logging to transport creation and connection flows - Added proper Accept header (application/json, text/event-stream) to transport options - Fixed scroll jittering during streaming with instant scroll behavior - Investigating HTTP transport 'Not connected' errors and SSE connection hanging * Add httpOnly option to disable SSE fallback for debugging - Added httpOnly option to UseMcpOptions to skip SSE fallback - Updated all HTTP transport failure paths to respect httpOnly flag - Temporarily enabled httpOnly in chat UI for streamable HTTP debugging - Updated AGENT.md to require frequent commits after changes * Replace httpOnly with transportType option and add UI controls - Replace httpOnly boolean with transportType ('auto'/'http'/'sse') option - Add transport type cycling button next to 'Add New Server' - Add blue transport badge next to connected server status - Save transport preference to localStorage - Update connection orchestration to handle sse-only and http-only modes - Remove explicit fallback messaging since orchestration now controls it * Fix tool results being mixed into thinking blocks - Add missing tool-result event handling in useStreamResponse - Tool results now appear as separate ToolResultMessage components - Prevents tool response content from being incorrectly added to reasoning content - Maintains proper message order: thinking → tool call → tool result → assistant response * Fix duplicate tool results and post-tool reasoning being hidden - Add deduplication logic for tool results to prevent duplicates - Detect when tool results have been processed and convert subsequent reasoning events to text content - This fixes the issue where the final assistant response was hidden in the thinking block - Tool results should now appear once and final response should appear as separate text * Fix message ordering and prevent content duplication in thinking/final response - Collect post-tool reasoning content separately instead of processing immediately - Process collected content as final assistant response after stream completes - This ensures proper message order: thinking → tool call → tool result → final response - Prevents duplication of thinking content in final response * Add intelligent content filtering to extract clean final answer - Extract only the final answer from post-tool reasoning content - Filter out thinking/reasoning patterns to prevent duplication - Use regex matching and line filtering to identify clean response content - Skip content that appears to be thinking rather than final answer - Limit response length to prevent long thinking content from appearing * Fix duplicate assistant messages by updating existing message - Check if there's already an assistant message with only reasoning content - Update existing assistant message with final content instead of creating new one - Add duplicate detection to prevent multiple messages with same content - Ensures only one assistant response appears after tool calls * Fix message ordering - create final response at end instead of updating existing - Remove logic that updates existing assistant message - Always create new final response message at end of conversation - This ensures proper order: thinking → tool call → tool result → final response - Still prevents duplicates by checking existing content * Make thinking block collapse immediately when tool call starts - End reasoning phase as soon as first tool-call event is detected - Set reasoningEndTime and stop reasoning streaming immediately - Creates linear flow: thinking expands → tool call (thinking collapses) → tool result → final response - More aggressive/responsive UX that doesn't wait for final response to collapse thinking * Fix reasoning/content separation and eliminate empty message divs - Remove extractReasoningMiddleware as it expects <think> tags but API sends reasoning fields directly - Ensure reasoning and content tokens create separate assistant messages - Filter empty assistant messages at conversation level to prevent empty divs - Add deduplication for tool results to prevent duplicates - Only render content div when message has actual content * Clean up debug logging in useStreamResponse.ts - Remove all debugLog calls and debug infrastructure - Simplify error logging by removing verbose prefixes - Fix unused options parameter to eliminate warnings - Clean up formatting and extra blank lines * Add targeted debug logging for message operations - Add debugMessages helper function that logs with [Messages] prefix - Log when new messages are added to conversation (not updates) - Log reasoning assistant messages, tool calls, tool results, and content messages - Log stream processing completion and function completion - Debug logging enabled with localStorage.setItem('USE_MCP_DEBUG', 'true') * Enhance debug logging to show complete conversation state - Show entire messages array after each new message addition - Include message index, role, content/reasoning status, and tool info - Add final conversation state with content/reasoning lengths - Provides complete visibility into data structure evolution * Simplify debug logging to show raw JSON messages - Replace complex message summaries with JSON.stringify(conv.messages) - Provides direct, unfiltered view of message data structure - Easier to inspect actual data and debug issues * Remove all duplicate detection logic - Remove existingToolCall and existingToolResult checks - Add every API event as a new message regardless of duplicates - Provides complete unfiltered log of all API data after middleware - Useful for debugging and understanding raw data flow * Branching off * fixing prettier * lfg manual style * wip, changing the logic in useStreamResponse * wip, at least prettier likes it * Big sigh, finally getting control of my markup * Starting to realise how dumb the vibecoded UI was oops * Ok turns out I can be dumb too * Handling reasoning then messages at least * tool responses are easy now too * visual tweaks * Only updating one conversation at a time * Ok _much_ better timing and just me understanding this stuff * Almost ready * fin * Default transport type to HTTP instead of auto * Move transport toggle to left and make it only affect new servers - Add newServerTransportType state for next server to be added - Move toggle to left side directly after 'Add New Server' label - Add transportType to McpServer interface to store per-server transport - Update connections to use server-specific transport type - Change toggle styling to rounded-full (pill style) * Round corners more on blue connected server transport type badge Change from rounded-md to rounded-lg for better visual consistency * Tighten markdown spacing in chat messages - Add custom prose-tight CSS class to reduce excessive spacing - Reduce line height from 1.75 to 1.5 for better readability - Decrease heading margins (1rem top, 0.5rem bottom) - Reduce paragraph and list item spacing - Apply tighter spacing to both user and assistant messages * removing dupe server * Forcing parsed reasoning on groq * Added inline in-conversation errors * Add `preventAutoAuth` (to chat example) (#18) * feat: add preventAutoAuth option to prevent automatic authentication popups - Add preventAutoAuth option to UseMcpOptions to control automatic popup behavior - Introduce pending_auth state for when auth is required but auto-popup is prevented - Enhance authentication flow to check for existing tokens before triggering auth - Add prepareAuthorizationUrl() helper method in BrowserOAuthClientProvider - Update authenticate() method to handle pending_auth state transitions - Improve chat UI components to handle new authentication states gracefully - Add clear call-to-action UI with primary popup button and fallback new tab link - Maintain backward compatibility - existing code continues to work unchanged This prevents browser popup blockers from interfering with authentication flows when users return to pages with MCP servers, providing better UX and control. * Fixing couple of things --------- Co-authored-by: Glen Maddern <glen@glenmaddern.com> * Chat example cooked and ready * I aint got time for this typescript nonsense rn * fixing pnpm lock * prettier --------- Co-authored-by: Glen Maddern <glen@glenmaddern.com>
1 parent c189d5a commit 26965cb

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

60 files changed

+9026
-13
lines changed

examples/chat-ui/.gitignore

Lines changed: 30 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,30 @@
1+
# Logs
2+
logs
3+
*.log
4+
npm-debug.log*
5+
yarn-debug.log*
6+
yarn-error.log*
7+
pnpm-debug.log*
8+
lerna-debug.log*
9+
10+
node_modules
11+
dist
12+
dist-ssr
13+
*.local
14+
15+
# Editor directories and files
16+
.vscode/*
17+
!.vscode/extensions.json
18+
.idea
19+
.DS_Store
20+
*.suo
21+
*.ntvs*
22+
*.njsproj
23+
*.sln
24+
*.sw?
25+
26+
.wrangler
27+
28+
# Playwright test results
29+
test-results/
30+
playwright-report/

examples/chat-ui/.prettierrc

Lines changed: 7 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,7 @@
1+
{
2+
"trailingComma": "es5",
3+
"tabWidth": 2,
4+
"semi": false,
5+
"singleQuote": true,
6+
"printWidth": 140
7+
}

examples/chat-ui/AGENT.md

Lines changed: 73 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,73 @@
1+
# AI Chat Template - Development Guide
2+
3+
## Commands
4+
- **Dev server**: `pnpm dev`
5+
- **Build**: `pnpm build` (runs TypeScript compilation then Vite build)
6+
- **Lint**: `pnpm lint` (ESLint)
7+
- **Deploy**: `pnpm deploy` (builds and deploys with Wrangler)
8+
- **Test**: `pnpm test` (Playwright E2E tests)
9+
- **Test UI**: `pnpm test:ui` (Playwright test runner with UI)
10+
- **Test headed**: `pnpm test:headed` (Run tests in visible browser)
11+
12+
## Code Style
13+
- **Formatting**: Prettier with 2-space tabs, single quotes, no semicolons, 140 char line width
14+
- **Imports**: Use `.tsx`/`.ts` extensions in imports, group by external/internal
15+
- **Components**: React FC with explicit typing, PascalCase names
16+
- **Hooks**: Custom hooks start with `use`, camelCase
17+
- **Types**: Define interfaces in `src/types/index.ts`, use `type` for unions
18+
- **Files**: Use PascalCase for components, camelCase for hooks/utilities
19+
- **State**: Use proper TypeScript typing for all state variables
20+
- **Error handling**: Use try/catch blocks with proper error propagation
21+
- **Database**: IndexedDB with typed interfaces, async/await pattern
22+
- **Styling**: Tailwind CSS classes, responsive design patterns
23+
24+
## Tech Stack
25+
React 19, TypeScript, Vite, Tailwind CSS, Hono API, Cloudflare Workers, IndexedDB, React Router, use-mcp
26+
27+
## Chat UI Specific Guidelines
28+
29+
### Background Animation
30+
- Use CSS `hue-rotate` filter for color cycling (more performant than gradient animation)
31+
- Isolate animations to `::before` pseudo-elements with `z-index: -1` to prevent inheritance
32+
- Use `background-size: 100% 100vh` and `background-repeat: repeat-y` for repeating patterns
33+
- Initialize with random animation delay using CSS custom properties
34+
35+
### Modal Patterns
36+
- Use `rgba(0,0,0,0.8)` for modal backdrops (avoid gradients that cause banding)
37+
- Implement click-outside-to-dismiss with `onClick` on backdrop and `stopPropagation` on content
38+
- Add `cursor-pointer` to all clickable elements including close buttons
39+
- Use `document.body.style.overflow = 'hidden'` to prevent background scrolling
40+
41+
### MCP Server Management
42+
- Store multiple servers in localStorage as JSON array (`mcpServers`)
43+
- Track tool counts separately (`mcpServerToolCounts`) for disabled servers
44+
- Use server-specific IDs for state management
45+
- Each MCP server gets unique `useMcp` hook instance for proper isolation
46+
- OAuth state parameter automatically handles multiple server differentiation
47+
48+
### State Persistence
49+
- Use localStorage for user preferences (model selection, server configurations)
50+
- Use sessionStorage for temporary state (single server URL in legacy components)
51+
- Clear related data when removing servers or models
52+
53+
### UI Indicators
54+
- Use emoji-based indicators (🧠 for models, 🔌 for MCP servers)
55+
- Format counts as `enabled/total` (e.g., "1/2 servers, 3/5 tools")
56+
- Place model selector on left, MCP servers on right for balance
57+
- Show "none" instead of red symbols for cleaner unconfigured states
58+
59+
### Component Organization
60+
- Keep disabled components in React tree but hidden with CSS (`{false && ...}`)
61+
- This prevents MCP connections from being destroyed when modals close
62+
- Use conditional rendering sparingly, prefer CSS visibility for stateful components
63+
64+
### Routing and OAuth
65+
- Use React Router for OAuth callback routes (`/oauth/callback`)
66+
- OAuth callback should match main app styling with loading indicators
67+
- use-mcp library handles multiple server OAuth flows automatically via state parameters
68+
69+
### Performance Considerations
70+
- Avoid animating gradients directly (causes repainting)
71+
- Use transform and filter animations (hardware accelerated)
72+
- Aggregate tools from multiple sources efficiently
73+
- Minimize localStorage reads in render loops

examples/chat-ui/README.md

Lines changed: 30 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,30 @@
1+
# AI chat template
2+
3+
An unofficial template for ⚛️ React ⨉ ⚡️ Vite ⨉ ⛅️ Cloudflare Workers AI.
4+
5+
Full-stack AI chat application using Workers for the APIs (using the Cloudflare [vite-plugin](https://www.npmjs.com/package/@cloudflare/vite-plugin)) and Vite for the React application (hosted using [Workers Static Assets](https://developers.cloudflare.com/workers/static-assets/)). Provides chat functionality with [Workers AI](https://developers.cloudflare.com/workers-ai/), stores conversations in the browser's [IndexedDB](https://developer.mozilla.org/en-US/docs/Web/API/IndexedDB_API), and uses [ai-sdk](https://sdk.vercel.ai/docs/introduction), [tailwindcss](https://tailwindcss.com/) and [workers-ai-provider](https://github.com/cloudflare/workers-ai-provider).
6+
7+
## Get started
8+
9+
Create the project using [create-cloudflare](https://www.npmjs.com/package/create-cloudflare):
10+
11+
```sh
12+
npm create cloudflare@latest -- --template thomasgauvin/ai-chat-template
13+
```
14+
15+
Run the project and deploy it:
16+
17+
```sh
18+
cd <project-name>
19+
npm install
20+
npm run dev
21+
```
22+
23+
```
24+
npm run deploy
25+
```
26+
27+
## What's next?
28+
29+
- Change the name of the package (in `package.json`)
30+
- Change the name of the worker (in `wrangler.jsonc`)

examples/chat-ui/api/index.ts

Lines changed: 118 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,118 @@
1+
import type { LanguageModelV1StreamPart } from 'ai'
2+
import { streamText, extractReasoningMiddleware, wrapLanguageModel } from 'ai'
3+
import { createWorkersAI } from 'workers-ai-provider'
4+
import { Hono } from 'hono'
5+
6+
interface Env {
7+
ASSETS: Fetcher
8+
AI: Ai
9+
}
10+
11+
type message = {
12+
role: 'system' | 'user' | 'assistant' | 'data'
13+
content: string
14+
}
15+
16+
const app = new Hono<{ Bindings: Env }>()
17+
18+
// Handle the /api/chat endpoint
19+
app.post('/api/chat', async (c) => {
20+
try {
21+
const { messages, reasoning }: { messages: message[]; reasoning: boolean } = await c.req.json()
22+
23+
const workersai = createWorkersAI({ binding: c.env.AI })
24+
25+
// Choose model based on reasoning preference
26+
const model = reasoning
27+
? wrapLanguageModel({
28+
model: workersai('@cf/deepseek-ai/deepseek-r1-distill-qwen-32b'),
29+
middleware: [
30+
extractReasoningMiddleware({ tagName: 'think' }),
31+
//custom middleware to inject <think> tag at the beginning of a reasoning if it is missing
32+
{
33+
wrapGenerate: async ({ doGenerate }) => {
34+
const result = await doGenerate()
35+
36+
if (!result.text?.includes('<think>')) {
37+
result.text = `<think>${result.text}`
38+
}
39+
40+
return result
41+
},
42+
wrapStream: async ({ doStream }) => {
43+
const { stream, ...rest } = await doStream()
44+
45+
let generatedText = ''
46+
const transformStream = new TransformStream<LanguageModelV1StreamPart, LanguageModelV1StreamPart>({
47+
transform(chunk, controller) {
48+
//we are manually adding the <think> tag because some times, distills of reasoning models omit it
49+
if (chunk.type === 'text-delta') {
50+
if (!generatedText.includes('<think>')) {
51+
generatedText += '<think>'
52+
controller.enqueue({
53+
type: 'text-delta',
54+
textDelta: '<think>',
55+
})
56+
}
57+
generatedText += chunk.textDelta
58+
}
59+
60+
controller.enqueue(chunk)
61+
},
62+
})
63+
64+
return {
65+
stream: stream.pipeThrough(transformStream),
66+
...rest,
67+
}
68+
},
69+
},
70+
],
71+
})
72+
: workersai('@cf/meta/llama-3.3-70b-instruct-fp8-fast')
73+
74+
const systemPrompt: message = {
75+
role: 'system',
76+
content: `
77+
- Do not wrap your responses in html tags.
78+
- Do not apply any formatting to your responses.
79+
- You are an expert conversational chatbot. Your objective is to be as helpful as possible.
80+
- You must keep your responses relevant to the user's prompt.
81+
- You must respond with a maximum of 512 tokens (300 words).
82+
- You must respond clearly and concisely, and explain your logic if required.
83+
- You must not provide any personal information.
84+
- Do not respond with your own personal opinions, and avoid topics unrelated to the user's prompt.
85+
${
86+
messages.length <= 1 &&
87+
`- Important REMINDER: You MUST provide a 5 word title at the END of your response using <chat-title> </chat-title> tags.
88+
If you do not do this, this session will error.
89+
For example, <chat-title>Hello and Welcome</chat-title> Hi, how can I help you today?
90+
`
91+
}
92+
`,
93+
}
94+
95+
const text = await streamText({
96+
model,
97+
messages: [systemPrompt, ...messages],
98+
maxTokens: 2048,
99+
maxRetries: 3,
100+
})
101+
102+
return text.toDataStreamResponse({
103+
sendReasoning: true,
104+
})
105+
} catch (error) {
106+
return c.json({ error: `Chat completion failed. ${(error as Error)?.message}` }, 500)
107+
}
108+
})
109+
110+
// Handle static assets and fallback routes
111+
app.all('*', async (c) => {
112+
if (c.env.ASSETS) {
113+
return c.env.ASSETS.fetch(c.req.raw)
114+
}
115+
return c.text('Not found', 404)
116+
})
117+
118+
export default app

examples/chat-ui/e2e/build.spec.ts

Lines changed: 45 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,45 @@
1+
import { test, expect } from '@playwright/test'
2+
import { exec } from 'child_process'
3+
import { promisify } from 'util'
4+
5+
const execAsync = promisify(exec)
6+
7+
interface ExecError extends Error {
8+
stdout?: string
9+
stderr?: string
10+
}
11+
12+
test.describe('Build Tests', () => {
13+
test('should build without type errors', async () => {
14+
try {
15+
const { stderr } = await execAsync('pnpm build', {
16+
cwd: process.cwd(),
17+
timeout: 60000,
18+
})
19+
20+
// Check for TypeScript compilation errors
21+
expect(stderr).not.toContain('error TS')
22+
expect(stderr).not.toContain('Type error')
23+
} catch (error) {
24+
const execError = error as ExecError
25+
console.error('Build failed:', execError.stdout, execError.stderr)
26+
throw new Error(`Build failed: ${execError.message}`)
27+
}
28+
})
29+
30+
test('should lint without errors', async () => {
31+
try {
32+
const { stderr } = await execAsync('pnpm lint', {
33+
cwd: process.cwd(),
34+
timeout: 30000,
35+
})
36+
37+
// ESLint should pass without errors
38+
expect(stderr).not.toContain('error')
39+
} catch (error) {
40+
const execError = error as ExecError
41+
console.error('Lint failed:', execError.stdout, execError.stderr)
42+
throw new Error(`Lint failed: ${execError.message}`)
43+
}
44+
})
45+
})

0 commit comments

Comments
 (0)