Maurice f901982903 fix(llm-parse): LLM_MAX_OUTPUT_TOKENS sets the answer cap, a cut-off answer says so ...

The OpenAI-compatible client always asked for 4096 tokens, and a document
whose bookings needed more came back as "no reservations" with nothing that
pointed at the cap. LLM_MAX_OUTPUT_TOKENS now sets the cap for every
provider; unset keeps 4096 and 8192 so a small local model's context still
fits. A finish_reason of "length" is reported like Anthropic's max_tokens
stop, naming the limit and the variable that raises it.

Closes #2677

2026-10-18 13:21:50 +02:00

..

e2e

feat(tours): a paddle tour's obstacle checklist and check bench on the server

2026-10-17 00:29:18 +00:00

fixtures

test(db): boot a v4.3.3 install with real rows on this release

2026-10-08 13:02:57 +02:00

helpers

feat(tours): a paddle tour's obstacle checklist and check bench on the server

2026-10-17 00:29:18 +00:00

integration

test(auth): read the disabled MFA state through the ORM, formatted

2026-10-18 12:11:41 +02:00

unit

fix(llm-parse): LLM_MAX_OUTPUT_TOKENS sets the answer cap, a cut-off answer says so

2026-10-18 13:21:50 +02:00

websocket

style: format server and shared with prettier

2026-10-10 21:23:37 +02:00

build-schema-snapshot.ts

refactor(db): open the database through a DatabaseLifecycle provider

2026-10-08 18:38:28 +02:00

global-setup.ts

refactor(db): open the database through a DatabaseLifecycle provider

2026-10-08 18:38:28 +02:00

setup.console-noise.ts

style: format server and shared with prettier

2026-10-10 21:23:37 +02:00

setup.ts

style: format server and shared with prettier

2026-10-10 21:23:37 +02:00