Commit History

Revert to working prompt, remove broken fallback
a63a522

Patryk Studzinski commited on

Improve infill prompt and add fallback for non-JSON model outputs
0dfa1e4

Patryk Studzinski commited on

Fix: Improve infill prompt to prevent model from copying template
2de7977

Patryk Studzinski commited on

Fix: Remove unsupported use_xformers_attention parameter
9153886

Patryk Studzinski commited on

Fix: Use direct model.generate() with proper KV caching instead of pipeline
eaa2e37

Patryk Studzinski commited on

Add KV caching and batch processing optimizations for 5-10x speedup
ab2e415

Patryk Studzinski commited on

Improve Polish grammar in infill prompt + remove debug logs
14fc89e

Patryk Studzinski commited on

Fix: Handle double-escaped JSON in infill parser + add debug logging
6cc98f9

Patryk Studzinski commited on

Fix: Handle function-call style
093fabc

Patryk Studzinski commited on

adding infill
5fabfb8

Patryk Studzinski commited on

Fix gemma chat template fallback
42e3538

Patryk Studzinski commited on

pre-downloading-all-models-at-startup
cf748a3

Patryk Studzinski commited on

model-lazy-loading
b50a781

Patryk Studzinski commited on

first-imrpvement-commit
a7fd202

Patryk Studzinski commited on

Fix: correct parameter name model_name_or_path
a1c0774

Patryk Studzinski commited on

adding main.py
6348ce6

Patryk Studzinski commited on

using a placeholder auth
ff33042

Patryk Studzinski commited on

cleanup after split to separate mcp service
3297dba

Patryk Studzinski commited on

feat: add VERSION file
87a12c6

Patryk Studzinski commited on

add get method
5c2acfd

Patryk Studzinski commited on

adding-github-files-to-spaces
9a9ec03

Patryk Studzinski commited on