A plugin can send a request it has already rendered, such as the
response template function. resolve_send_inputs snapshotted whatever
request it was handed, so those versions held rendered values, secrets
included, where the user had templates, and a restore would have
written the literals back.
snapshot_request_for_send now takes a request id and reads the stored
row inside its transaction. A request with no id or no stored row gets
no version, as a failed snapshot already did. gRPC and WebSocket connect
go through the same method, so all three capture the same thing.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Restore told open editors to re-read the request before its change event
had reached the store, so header and parameter rows could keep showing
the old content. The restore RPC now returns the restored request, and
the frontend puts it in the store before calling wasUpdatedExternally,
the same way deleteModel applies its change locally once the RPC
resolves.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The comparison query kept its last answer as placeholder data across
version changes, so switching to a response with no version, or to one
whose comparison was still loading, could leave the Request Changed
dropdown offering another response's diff and restore. Placeholder data
now carries over only within the same version, and the hook returns a
comparison only when it is for the version asked about.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Conflicts were all at registration points; both sides are kept.
Main now routes every write through one connection (#642): ClientDb only
reads and writes go through `with_tx` on a WriteDb. Versioning follows:
snapshot, restore, prune and the cascade delete move to WriteDb, while the
lookups and the "has it changed" check stay on ClientDb. Callers take a
write transaction the way their neighbours do:
- The HTTP send snapshots in its own transaction from
`resolve_send_inputs`, and a failure there (including not getting the
writer) is still logged and swallowed, so the response just has no
version.
- gRPC and WebSocket connect snapshot inside the transaction that upserts
the connection.
- The snapshot and restore RPCs, on the desktop and in the browser host,
each run in a transaction.
- The request and workspace cascades live on main's WriteDb deletes.
The Request Changed dropdown moves into the response pane's
`statusLineRight` slot, since main moved the status line into the shared
pane view.
The plugin runtime's copy of the model bindings picks up `versionId`, and
`gen_rpc.ts` is regenerated.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
CVE-2026-111497 is an out-of-bounds write in the Brotli decoder. It
reaches Yaak through the bundled Node (`yaaknode`, used by plugins) and
through the pure-Rust `brotli-decompressor` that yaak-http uses to decode
`br` responses.
- Bump the plugin runtime Node from 24.11.1 to 24.22.2 and update the
vendoring script's SHA256 pins from the signed SHASUMS256.txt.
- Move yaak-http from `brotli = "8"` to `brotli = "9"`, locking brotli
9.1.1 and brotli-decompressor 6.2.0.
Tauri's pinned git rev still pulls brotli 8.0.1 and brotli-decompressor
5.0.0 for its embedded-asset compression; that needs a Tauri bump.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The shared tree treats every node with a children list as a folder and
collapses an empty one before the left arrow moves to its parent. Left
out of the collapsed map, a leaf path never read as collapsed, so the
left arrow kept trying to collapse it and never moved up.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every domain and path node starts collapsed, so paths stay hidden until
a host is expanded. The nodes the user expands are kept by id in local
storage, so they stay open across launches; hosts seen later start
collapsed. The Domains folder, selection and counts are as before. The
shared tree is unchanged: the proxy hands it its own collapsed atom.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Send a flow's write event inside its transaction, so a clear's event,
sent after its own commit, always comes after it.
- Record a saved request only on a flow that still exists, so a save
that finishes after a clear does not bring the flow back.
- Count the total again from the database a second after new flows,
so a flow both in a page and in the events around it settles to the
right count.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Replay live changes over a page of older flows, so a slow flow that
ends while the page loads is not lost from the window.
- Read list_flows' page and total from one snapshot, so a flow written
between them is not counted twice.
- Commit a clear's flow deletion only after its bodies' deletion, so a
failure leaves the flows to clear again rather than orphaned bodies.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The app held every flow in memory and redid all of its work for each new
one: the sidebar mounted every domain and path node (its tree scrolled
inside a container that never bounded it), every tree row looked up its
atoms by scanning every atom the family had made, and the layout, detail
pane and list rows re-rendered per flow. At 3,000 flows each new one cost
about 90 ms of main thread, more than the traffic left time for.
The app now holds a window of the newest 2,000 flows, applies new ones in
batches, and loads older ones in pages from the database on scrolling to
the top, through a new list_flows RPC that pages in SQL. The database
keeps every flow.
Also adds clear_flows: an RPC, POST /flows/clear on the control API, and
a Clear button in the header that Cmd+K opens, confirming first.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Rebuild a bare-path plain-HTTP request's URL from Host, answer 400
"no host in request" when there is none, and refuse with 508 a rebuilt
request that is routed back into the proxy, which would otherwise loop
without end.
- Drop Host when forwarding to an HTTP/2 upstream. Google's servers reset
every stream that carries Host alongside :authority.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Use an object with a null prototype so inherited property names are not
mistaken for existing fields and __proto__ remains a literal JSON key.
Cover __proto__, constructor, and toString with regression cases.