APEX/BrowserUse_and_ComputerUse_skills
5
0
Fork 0
Code Issues 6 Pull requests 1 Projects 1 Releases Packages Wiki Activity Actions

fix: context counter shows cached token count in status bar

Browse source 
Anthropic prompt caching splits input into cache_read_input_tokens,
cache_creation_input_tokens, and non-cached input_tokens. The context
counter only read input_tokens (non-cached portion), showing ~3 tokens
instead of the real ~18K total. Now includes cached portions for
Anthropic native provider only — other providers (OpenAI, OpenRouter,
Codex) already include cached tokens in their prompt_tokens field.

Before: 3/200K | 0%
After: 17.7K/200K | 9%
This commit is contained in:
0xbyt4 2026-03-17 04:39:11 +03:00
parent cfa87e77a9
commit 8d0a96a8bf
2 changed files with 124 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

9
run_agent.py
View file

@ -5256,6 +5256,15 @@ class AIAgent:
if hasattr(response, 'usage') and response.usage:
if self.api_mode in ("codex_responses", "anthropic_messages"):
prompt_tokens = getattr(response.usage, 'input_tokens', 0) or 0
if self.api_mode == "anthropic_messages":
# Anthropic splits input into cache_read + cache_creation
# + non-cached input_tokens. Without adding the cached
# portions, the context bar shows only the tiny non-cached
# portion (e.g. 3 tokens) instead of the real total (~18K).
# Other providers (OpenAI/Codex) already include cached
# tokens in their input_tokens/prompt_tokens field.
prompt_tokens += getattr(response.usage, 'cache_read_input_tokens', 0) or 0
prompt_tokens += getattr(response.usage, 'cache_creation_input_tokens', 0) or 0
completion_tokens = getattr(response.usage, 'output_tokens', 0) or 0
total_tokens = (
getattr(response.usage, 'total_tokens', None)