Commit Graph
7 Commits
Author SHA1 Message Date
DeepFal 051eda6166 fix(ai): keep surrogate pairs whole when truncating model context (#4088)
* fix(ai): keep surrogate pairs whole when truncating model context

Head/tail truncation sliced strings by UTF-16 code unit, so a cut could
split an emoji into a lone surrogate. JSON.stringify then emits a
`\udXXX` escape that strict providers reject ("lone leading surrogate in
hex escape"), and because the same cut is reapplied every step, the chat
stays stuck on that 400 for every later request.

- Add textBoundaries helpers (sliceHead, sliceTail, safeSliceBounds) and
  use them at every truncation site on the model request path.
- Wrap provider models with a middleware that replaces any remaining lone
  surrogate with U+FFFD before the request body is serialized.

* fix(ai): keep own __proto__ keys when normalizing prompt values

toWellFormedValue cloned changed objects by plain assignment, so an own
__proto__ key (e.g. from JSON.parse) hit the inherited setter: the field
was dropped from the request and the clone's prototype changed. Clone
with the source prototype and define own properties instead.
2026-11-05 22:44:28 +08:00
陈大猫andyangfan e8c31a7800 fix(ai): keep Catty image attachments through routine context compression (#3765)
Typed request compression ran before every Catty request (routine pre-turn and every step) and replaced image/file parts with a text placeholder, so attached images never reached the model.

preserveAttachments now keeps image attachments on the routine pre-turn and step paths. Non-image files stay placeholders (chat providers reject e.g. text/plain; the model reads them via list_attachments/read_attachment). The explicit 413 retry still trades attachments for placeholders.

Verified in the real Electron app: image reaches the provider as image_url (fake OpenAI-compatible server and DeepSeek deepseek-flash, which described the test image correctly); text-file turns still succeed via attachment tools; request prefix stability across turns unchanged vs main (DeepSeek cache hits ~98% on follow-up turns).

Fixes #3764

Co-authored-by: yangfan <1413737839@qq.com>
2026-10-20 10:20:14 +08:00
陈大猫 df5c033dfa fix(ai): preserve Catty prompt cache and cache usage (#3714)
* fix(ai): preserve Catty prompt cache and provider usage

* fix(ai): gate explicit caching on the official Anthropic endpoint

* fix(ai): resolve provider defaults before enabling prompt caching
2026-10-18 12:20:18 +08:00
陈大猫 8cc8d938b8 fix(ai): preserve user text during request compression (#3303) 2026-09-07 10:30:35 +08:00
bincxz affe36814f fix(ai): address Catty context review findings 2026-07-17 12:26:36 +08:00
bincxz 29c4d12e82 feat(ai): harden Catty terminal context handling 2026-07-17 12:12:12 +08:00
陈大猫 517cbb6cee fix(ai): compress Catty requests only after 413 (#1327)
* fix(ai): compress Catty requests only after 413

* fix(ai): retry 413 after tool progress safely

* fix(ai): mark thrown 413 retries after tool progress

* fix(ai): preserve tool results in 413 retry
2026-06-09 13:11:42 +08:00