Module 2 run: collapse the chain
==========================================================================
model      : claude-opus-5
max_tokens : 1024
limits     : 40 calls, 80000 tokens in total
anthropic  : 1.2.0
python     : 3.12.16
started    : 2026-11-20T00:11:42+00:00
tickets    : 10    queues: billing, auth, bug, general

--------------------------------------------------------------------------
chain, three calls per ticket
--------------------------------------------------------------------------
   1. OK   got 'billing'    key 'billing'
   2. OK   got 'billing'    key 'billing'
   3. MISS  got '`auth`'     key 'auth'
   4. OK   got 'bug'        key 'bug'
   5. OK   got 'auth'       key 'auth'
   6. OK   got 'bug'        key 'bug'
   7. OK   got 'billing'    key 'billing'
   8. OK   got 'bug'        key 'bug'
   9. OK   got 'billing'    key 'billing'
  10. MISS  got 'auth'       key 'general'

--------------------------------------------------------------------------
single, one call per ticket
--------------------------------------------------------------------------
   1. OK   got 'billing'    key 'billing'
   2. OK   got 'billing'    key 'billing'
   3. OK   got 'auth'       key 'auth'
   4. OK   got 'bug'        key 'bug'
   5. OK   got 'auth'       key 'auth'
   6. OK   got 'bug'        key 'bug'
   7. OK   got 'billing'    key 'billing'
   8. OK   got 'bug'        key 'bug'
   9. OK   got 'billing'    key 'billing'
  10. OK   got 'general'    key 'general'

--------------------------------------------------------------------------
result
--------------------------------------------------------------------------
  chain   8/10 matches   30 calls (3 per ticket)
  single  10/10 matches   10 calls (1 per ticket)

  scored again after stripping non-letters:
    chain   8/10 exact -> 9/10 normalised
    single  10/10 exact -> 10/10 normalised

  values returned that are not one of the four queue names:
    chain  1/10
    single 0/10

--------------------------------------------------------------------------
token usage, from the JSONL
--------------------------------------------------------------------------
  chain:summarize     10 calls  in    379  out   1753
  chain:candidates    10 calls  in    749  out   6969
  chain:pick          10 calls  in   5355  out    141
  single              10 calls  in   1659  out    195
  TOTAL               40 calls  in   8142  out   9058
  output split: thinking 3369 tokens, answer 5689 tokens
  stop_reason : end_turn 40
  empty answers: 0 of 40

finished   : 2026-11-20T00:14:28+00:00
wrote      : /home/runner/work/_temp/m2/run.jsonl
Next: python ch02-stack/ch2_course_check.py --part a
exit 0
