9 commits

Author	SHA1	Message	Date
CREDO23	32f6766cb6	fix(tokens): use canonical prompt_tokens_details path for cache fields ... LiteLLM normalizes every provider's cache fields onto usage.prompt_tokens_details (cached_tokens + cache_creation_tokens). The earlier fallback to usage.cache_read_input_tokens / usage.cache_creation_input_tokens was wrong: Anthropic-shaped fields only live there via a trailing setattr loop, and the canonical field name on the wrapper is cache_creation_tokens (not _input_tokens).	2026-05-20 09:55:39 +02:00
CREDO23	6090980c5e	obs(tokens): log prompt-cache read/write counts and hit ratio per LLM call	2026-05-20 09:51:44 +02:00
CREDO23	581bbfb5c1	perf(tokens): add per-call latency to capture log	2026-05-19 21:30:25 +02:00
DESKTOP-RTLN3BA\$punk	ae9d36d77f	feat: unified credits and its cost calculations	2026-05-02 14:34:23 -07:00
Anish Sarkar	9fc0976d5e	chore: ran linting	2026-04-14 21:52:26 +05:30
Anish Sarkar	f01ddf3f0a	feat: implement token usage recording in chat routes and enhance title generation handling	2026-04-14 20:56:07 +05:30
Anish Sarkar	f21bdc0668	feat: enhance LLM response handling and token usage tracking in chat services and UI components	2026-04-14 15:29:02 +05:30
Anish Sarkar	5af6005163	feat: improve token usage tracking and response handling in chat routes and services	2026-04-14 14:28:31 +05:30
Anish Sarkar	3cfe53fb7f	feat: implement token usage tracking for LLM calls with new accumulator and callback	2026-04-14 13:40:32 +05:30