Sergey Kozyrenko f012dce6e4 fix(cache): let the partial answer through when the model reports truncation ...

Adapters return the response together with the truncation error so the caller
can keep what arrived. The cache wrapper dropped it, so anyone generating
through the cache got a bare error and lost the partial answer.

The wrapper now forwards both. A failed generation is still not cached.

Verified by mutation: restoring the discard reddens the new test.

PAGI-132

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

2026-08-23 15:44:25 +07:00

..

inmemory

fix: multiple linter issues and logic in tests

2025-06-23 13:43:06 +03:00

cache_test.go

feat: enhance streaming functionality with done chunk handling

2025-06-23 21:09:44 +03:00

cache.go

fix(cache): let the partial answer through when the model reports truncation

2026-08-23 15:44:25 +07:00

doc.go

llms: add caching functionality for Models (#564)

2024-03-19 21:10:35 -07:00

mocks_test.go

feat: enhance streaming functionality with done chunk handling

2025-06-23 21:09:44 +03:00

truncation_test.go

fix(cache): let the partial answer through when the model reports truncation

2026-08-23 15:44:25 +07:00