mirror of
https://github.com/different-ai/openwork
synced 2026-04-25 17:15:34 +02:00
2b91b4d777
* refactor(repo): move OpenWork apps into apps and ee layout Rebase the monorepo layout migration onto the latest dev changes so the moved app, desktop, share, and cloud surfaces keep working from their new paths. Carry the latest deeplink, token persistence, build, Vercel, and docs updates forward to avoid stale references and broken deploy tooling. * chore(repo): drop generated desktop artifacts Ignore the moved Tauri target and sidecar paths so local cargo checks do not pollute the branch. Remove the accidentally committed outputs from the repo while keeping the layout migration intact. * fix(release): drop built server cli artifact Stop tracking the locally built apps/server/cli binary so generated server outputs do not leak into commits. Also update the release workflow to check the published scoped package name for @openwork/server before deciding whether npm publish is needed. * fix(workspace): add stable CLI bin wrappers Point the server and router package bins at committed wrapper scripts so workspace installs can create shims before dist outputs exist. Keep the wrappers compatible with built binaries and source checkouts to avoid Vercel install warnings without changing runtime behavior.
146 lines
3.9 KiB
Markdown
146 lines
3.9 KiB
Markdown
# Notion Connection: Investigate Double-Restart Issue
|
|
|
|
**Branch:** `feat/notion-connection-fix`
|
|
**Priority:** P2
|
|
|
|
---
|
|
|
|
## Problem
|
|
|
|
From demo: "First time we connect Notion just didn't work. It says no write access... Second time we got CSRF error and third time it worked"
|
|
|
|
The connection flow requires multiple restart attempts:
|
|
1. Connect → "Write request denied"
|
|
2. Restart → CSRF error
|
|
3. Restart again → Finally works
|
|
|
|
---
|
|
|
|
## Hypothesis
|
|
|
|
The config needs to restart twice:
|
|
- Once to detect the MCP is configured
|
|
- Second time to actually activate the connection with valid tokens
|
|
|
|
### Code findings
|
|
- MCP auth modal: `packages/app/src/app/components/mcp-auth-modal.tsx:12`
|
|
- Auth flow calls OpenCode MCP endpoints: `mcp-auth-modal.tsx:155,180,311`
|
|
- Auth modal blocks when reload required: `mcp-auth-modal.tsx:139`
|
|
- Notion quick-connect flow: `packages/app/src/app/app.tsx:2177`
|
|
- Writes Notion MCP config: `app.tsx:2222`
|
|
- Marks reload required: `app.tsx:2246`
|
|
- Sets localStorage `openwork.notionStatus`: `app.tsx:2246`
|
|
- On startup, `openwork.notionStatus === "connecting"` triggers reload-required: `app.tsx:2817`
|
|
- MCP status refresh uses OpenCode `client.mcp.status`: `app.tsx:2294`
|
|
- File watcher emits reload-required for config changes: `packages/desktop/src-tauri/src/workspace/watch.rs:99-124`
|
|
- UI listener converts event to `markReloadRequired`: `packages/app/src/app/app.tsx:1868`
|
|
|
|
---
|
|
|
|
## Investigation Steps
|
|
|
|
### 1. Trace the OAuth flow
|
|
|
|
**Files to check:**
|
|
- `packages/app/src/app/components/mcp-auth-modal.tsx`
|
|
- Server-side MCP handling
|
|
|
|
Questions:
|
|
- When OAuth completes, where is the token saved?
|
|
- Does the app know the token is saved?
|
|
- Is there a race condition between token save and config reload?
|
|
|
|
Add checks around `openwork.notionStatus` transitions to confirm if status ever clears after the first reload.
|
|
|
|
### 2. Check what happens on first restart
|
|
|
|
Add logging:
|
|
```tsx
|
|
// In MCP status handling
|
|
console.log('[MCP] Notion config:', notionConfig);
|
|
console.log('[MCP] Notion status:', notionStatus);
|
|
console.log('[MCP] Has token:', !!notionToken);
|
|
```
|
|
|
|
### 3. Check CSRF error source
|
|
|
|
- Is this from Notion's OAuth?
|
|
- Is this from OpenWork server?
|
|
- Is there stale state from previous connection attempt?
|
|
|
|
Check if reload-required is firing twice (explicit `markReloadRequired("mcp")` + file watcher event) right after OAuth.
|
|
|
|
### 4. Check server restart logic
|
|
|
|
**Files:**
|
|
- `packages/desktop/src-tauri/src/openwork_server/spawn.rs`
|
|
- OpenCode server handling
|
|
|
|
Questions:
|
|
- Does restart fully clear MCP state?
|
|
- Is there caching that persists across restarts?
|
|
|
|
---
|
|
|
|
## Potential Fixes
|
|
|
|
### Option A: Auto-restart after OAuth
|
|
|
|
After OAuth token is saved, automatically trigger a server reload:
|
|
```tsx
|
|
const handleOAuthComplete = async () => {
|
|
await saveToken();
|
|
// Automatically reload to pick up new token
|
|
await reloadWorkspace();
|
|
};
|
|
```
|
|
|
|
### Option B: Hot-reload MCP connections
|
|
|
|
Instead of full restart, implement hot-reload for MCP:
|
|
```tsx
|
|
const refreshMcpConnection = async (name: string) => {
|
|
// Disconnect existing
|
|
await mcpDisconnect(name);
|
|
// Reconnect with new config
|
|
await mcpConnect(name);
|
|
};
|
|
```
|
|
|
|
### Option C: Fix the state mismatch
|
|
|
|
If the issue is stale state, ensure clean state on connection attempt:
|
|
```tsx
|
|
const connectNotion = async () => {
|
|
// Clear any cached state
|
|
clearMcpCache("notion");
|
|
// Proceed with fresh connection
|
|
await initiateOAuth("notion");
|
|
};
|
|
```
|
|
|
|
**Also consider:** clearing `openwork.notionStatus` after successful engine reload to avoid repeated reload prompts on startup.
|
|
|
|
---
|
|
|
|
## Reproduction Steps
|
|
|
|
1. Fresh install or clear Notion connection
|
|
2. Go to MCP settings
|
|
3. Click "Connect Notion"
|
|
4. Complete OAuth flow
|
|
5. Observe error
|
|
6. Click Reload
|
|
7. Check if working
|
|
8. If not, reload again
|
|
|
|
Track what state changes at each step.
|
|
|
|
---
|
|
|
|
## Success Criteria
|
|
|
|
- Notion connection works on first attempt after OAuth
|
|
- No need to restart multiple times
|
|
- Clear error messages if something goes wrong
|