savings never seems to change

[ITFinesse]

What happened?

running: cce savings

always reports the same savings constantly.

Bee · 12 queries

⛁ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ 88% tokens saved

Input savings 1.9M tokens $27.78
Output savings 4.8k tokens $0.36
──────────────────────────────────────────
Total saved 1.9M tokens $28.15
~154.8k tokens / query ~$2.35 / query

Breakdown:
retrieval 84% ▰▰▰▰▰▰▰▰▰▰ 1.8M $26.76 · 12 calls
chunk compression 3% ▰▱▱▱▱▱▱▱▱▱ 68.5k $1.03 · 12 calls
output compression* <1% ▰▱▱▱▱▱▱▱▱▱ 4.8k $0.36 · 12 calls

• estimated. output compression assumes a 500-token avg reply; progressive disclosure compares against full payload dump.
  Output compression levels seen: max=12
  Cost estimate based on Opus pricing (input $15.0/1M, output $75.0/1M)

after a agent chat

Bee · 12 queries

⛁ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ 88% tokens saved

Input savings 1.9M tokens $27.78
Output savings 4.8k tokens $0.36
──────────────────────────────────────────
Total saved 1.9M tokens $28.15
~154.8k tokens / query ~$2.35 / query

Breakdown:
retrieval 84% ▰▰▰▰▰▰▰▰▰▰ 1.8M $26.76 · 12 calls
chunk compression 3% ▰▱▱▱▱▱▱▱▱▱ 68.5k $1.03 · 12 calls
output compression* <1% ▰▱▱▱▱▱▱▱▱▱ 4.8k $0.36 · 12 calls

• estimated. output compression assumes a 500-token avg reply; progressive disclosure compares against full payload dump.
  Output compression levels seen: max=12
  Cost estimate based on Opus pricing (input $15.0/1M, output $75.0/1M)

What did you expect?

savings to update

Steps to reproduce

cce savings after chats

Relevant logs or error output

Python version

na

OS

na

CCE version

latest

[rajkumarsakthivel]

Thanks for reporting this. Two things going on here:

Why savings weren't updating: Savings only increment when the agent calls context_search or expand_chunk (the CCE MCP tools). If the agent uses Read/Grep/Write directly instead of CCE tools, no new savings get recorded. This is the most likely reason the numbers stayed at 12 queries.

To verify: check if your CLAUDE.md (or equivalent instruction file) has the CCE instructions telling the agent to use context_search. Running cce init again will re-add them if missing.

What we fixed (PR #110):

1. cce savings now shows a freshness hint like "last query 5m ago" or "last query 2d ago" next to the project name, so you can immediately tell whether new queries happened during your latest chat
2. Query count is now derived from both memory.db and stats.json (whichever is higher), so the display stays accurate even if one source gets out of sync

Before the fix:

  Bee · 12 queries

After:

  Bee · 12 queries · last query 2d ago

If you see "last query 2d ago" after a fresh chat, that confirms the agent isn't calling context_search. Re-running cce init should fix the instruction file.