* fix(ai): cap Catty agent request payload to prevent HTTP 413
Long-running chats accumulated full terminal tool outputs in SDK history
while token-based compaction only triggered near the model context window,
so nginx gateways could reject oversized JSON bodies before the model saw them.
Add a byte-budget pass that compresses verbose output, tail-preserving
truncation, and a safe sliding window before each Catty agent turn.
Fixes#1323
Co-authored-by: Cursor <cursoragent@cursor.com>
* fix(ai): compaction was using unfiltered sdkMessages, fix didAdjust and emergency loop
* fix(ai): compact and retry oversized Catty requests
* fix(ai): preserve current input while guarding 413 retries
* fix(ai): avoid false 413 detection and fit oversized current input
* fix(ai): pair replayed tool results chronologically
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
Symptom: when an AI request is proxied through nginx (or any gateway)
and the request body exceeds client_max_body_size, the proxy returns a
413 HTML error page. The Vercel AI SDK then fails to parse the HTML
as a chat completion and surfaces a cryptic Zod validation error like
"Expected 'id' to be a string." through the UI — users have no idea
what's wrong.
Root cause: classifyError only did light sanitization and returned the
raw SDK message. It also string-coerced the error before inspection, so
the structured statusCode / responseBody fields that APICallError
attaches were thrown away.
Fix: classifyError now accepts `unknown` and inspects the full error
shape. Adds explicit branches for:
- HTTP 413 (from statusCode, cause.statusCode, or message text) →
"Request too large — exceeded proxy size limit. Try shorter
message, fewer attachments, or raise client_max_body_size."
- HTTP 502/503/504 → retryable upstream-gateway message
- HTML response body (starts with <!DOCTYPE/<html> or contains such
tags anywhere) → "Server returned HTML error page, likely a proxy
intercept."
- Zod/schema parse shapes ("Expected 'X' to be …", "Invalid JSON
response", "Type validation failed") → "Response could not be
parsed; proxy may have replaced/truncated the body."
In every classified case the raw SDK text is still appended ("Raw: …")
so users can report the underlying error verbatim.
useAIChatStreaming.ts callers now pass the raw error to classifyError
instead of `.message`, so the new structured branches actually fire.
Also wired infrastructure/ai/*.test.ts into the npm test glob.
Closes#765
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat: add skip TLS verification option for AI providers
Self-hosted AI endpoints (vLLM, text-generation-webui, etc.) often use
self-signed TLS certificates which Node.js rejects by default, causing
502 Bad Gateway errors. Add a per-provider "Skip TLS certificate
verification" checkbox that sets rejectUnauthorized=false on both
streaming and non-streaming requests.
Ref #294
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: surface real error message instead of generic 502 Bad Gateway
- Pass the actual bridge error message in statusText so Vercel AI SDK
shows the real cause (e.g. "HTTP is only allowed for localhost",
"URL host is not in the allowed list", TLS errors)
- Show real error details for 5xx provider errors instead of generic
"The AI provider returned a server error" message
Previously all connection-level errors were masked as "Bad Gateway"
making it impossible for users to diagnose configuration issues.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: pass server error body details through to the user
- Read HTTP error response body before resolving (was resolving before
body was read, losing the error detail)
- Parse OpenAI-compatible JSON error format to extract error.message
- Return error Response with body+statusText for non-2xx instead of
empty stream, so Vercel AI SDK shows the real server error
- Now users see e.g. "502 model not loaded" instead of just "Bad Gateway"
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: widen link modifier key dropdown to prevent text wrapping
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* Revert "fix: widen link modifier key dropdown to prevent text wrapping"
This reverts commit 1f75686391.
* fix: apply skipTLSVerify to model listing requests
ModelSelector.aiFetch() didn't pass providerId, so the provider-level
skipTLSVerify was not applied when refreshing/listing models. Add
skipTLSVerify as a direct parameter alongside the provider lookup.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: keep error detail in Response body, not statusText
statusText only accepts single-line Latin-1 — multiline or non-ASCII
error messages from self-hosted gateways would throw TypeError before
the AI SDK could read them. Move detailed error to body instead.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: return JSON error body for AI SDK compatibility, fix FetchBridge type
- Wrap error responses in OpenAI-compatible JSON format so Vercel AI
SDK's failedResponseHandler extracts the message correctly instead
of showing a blank error
- Update FetchBridge type to match the expanded aiFetch parameter list
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: add ASCII statusText fallback for non-OpenAI SDK providers
Anthropic/Google SDKs fall back to Response.statusText when they can't
parse the error body. Add safe ASCII statusText alongside the JSON body.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Security: API keys no longer transit IPC as plaintext; renderer sends
providerId and main process decrypts via safeStorage. Add ReDoS
protection for user-supplied blocklist regex. Sanitize error messages
to strip file paths and sensitive URLs before displaying in chat.
- Bug fixes: approval timeout now notifies user in chat instead of
silently aborting. statusText cleared consistently across all 7 code
paths. useAIState persistence race condition fixed with mountedRef
guard, storage sync validated with type checks.
- Accessibility: InlineApprovalCard uses role="alertdialog" with focus
on approve button. ChatInput menus have proper ARIA roles, labels,
and aria-expanded. ThinkingBlock toggle has aria-expanded/controls.
Model selector submenu supports keyboard navigation.
- UX: switching agents preserves old session and creates new one.
Approval card buttons disabled immediately after click.
- Performance: text-delta streaming batched via requestAnimationFrame.
- Refactor: extract useAIChatStreaming, useToolApproval, and
useConversationExport hooks from AIChatSidePanel (1514 → 751 lines).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add errorInfo field to ChatMessage with type classification
(network/auth/timeout/provider/agent/unknown) and retryable flag
- Create errorClassifier.ts to map raw error strings into user-friendly
structured messages with actionable hints
- Replace inline "**Error:**" text appending with dedicated error messages
rendered as styled error cards with AlertCircle icon
- Ensure streaming indicator is cleared immediately on error in ACP and
external agent flows (not just in finally block)
- Mark previous message executionStatus as failed only when it was running
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>