claude 1b8ae3c3e5 router/semantic: baseline scoring and contrast family analysis (slice 13) ...

ScoreLegacy runs the actual router cascade against the 136-example
corpus and produces per-route precision/recall/F1, confusion matrix,
false-action breakdown, and fast-path/residual/router-residual/pre-route
consumption counts. Pre-route consumed cases (command-prohibition grammar
matches at stage 0) are tracked separately — these never reach the
general cascade and should not be scored by the learned router.

ContrastFamilies splits the baseline report by transform tag (negation,
question, reported_speech, quotation, hypothetical, capability_question)
and reports per-family accuracy and false-action rate.

LegacyReport.String() renders the full baseline report with confusion
matrix and false-action case listing.

2026-09-07 02:08:51 +04:00

..

eval

router: update all Route callers for NormalizedInput + add invariant tests (slice 11)

2026-09-07 01:32:49 +04:00

semantic

router/semantic: baseline scoring and contrast family analysis (slice 13)

2026-09-07 02:08:51 +04:00

actioncandidate_test.go

router: update ActionCandidate tests for CapabilitySelection (slice 6b)

2026-09-06 21:02:17 +04:00

actioncandidate.go

router: ResolveActionCandidate reads CapabilitySelection first (slice 6b)

2026-09-06 21:02:13 +04:00

actmatcher_alias_test.go

a Russian act reaches a tool, and the seeds stop disagreeing (V-633)

2026-08-06 18:09:11 +04:00

acttarget_test.go

router: migrate refusesCommand and ActHasEntityTarget to CapabilitySelection (slice 6c)

2026-09-06 21:43:14 +04:00

acttarget.go

router: migrate refusesCommand and ActHasEntityTarget to CapabilitySelection (slice 6c)

2026-09-06 21:43:14 +04:00

agenda_test.go

router: update all Route callers for NormalizedInput + add invariant tests (slice 11)

2026-09-07 01:32:49 +04:00

agendaq_test.go

each agenda grammar is tested against its own example (V-552)

2026-08-05 21:40:49 +04:00

agendaq.go

the calendar answers his day, not any day (V-552)

2026-08-05 21:40:15 +04:00

barecapture_test.go

a missing slot asks, whatever the confidence (V-557)

2026-08-06 00:08:43 +04:00

barecapture.go

a missing slot asks, whatever the confidence (V-557)

2026-08-06 00:08:43 +04:00

boundary_test.go

router: update all Route callers for NormalizedInput + add invariant tests (slice 11)

2026-09-07 01:32:49 +04:00

calendar_test.go

router: refuse a plan question about a day that is not today

2026-08-01 14:07:43 +04:00

calendar.go

one weekday matcher, and a stem list stops answering for sredstvo (V-581)

2026-08-06 03:33:01 +04:00

capability_test.go

router: update all Route callers for NormalizedInput + add invariant tests (slice 11)

2026-09-07 01:32:49 +04:00

capability.go

router: introduce CapabilitySelection type and SelectCapability stage (slice 6b)

2026-09-06 21:01:59 +04:00

claim_test.go

router: update all Route callers for NormalizedInput + add invariant tests (slice 11)

2026-09-07 01:32:49 +04:00

claim.go

build a claim from a Decision, beside the existing path (V-565)

2026-08-06 00:54:55 +04:00

classifier.go

initial commit

2026-07-03 00:32:48 +02:00

commandframe_test.go

Add seven stage 0 frames and tighten three more (V-720)

2026-08-15 17:18:48 +04:00

commandframe.go

Add seven stage 0 frames and tighten three more (V-720)

2026-08-15 17:18:48 +04:00

complaint_test.go

sweep tail: the last three files pick a mechanism (V-528)

2026-08-04 19:10:00 +04:00

complaint.go

sweep tail: the last three files pick a mechanism (V-528)

2026-08-04 19:10:00 +04:00

currentversion_test.go

Add seven stage 0 frames and tighten three more (V-720)

2026-08-15 17:18:48 +04:00

currentversion.go

Add seven stage 0 frames and tighten three more (V-720)

2026-08-15 17:18:48 +04:00

dateparser_test.go

dates: a bare hour after a day word is an hour, not the current clock (V-551)

2026-08-05 21:17:21 +04:00

dateparser.go

The hour after "к" is read, and an hour nobody read is asked about (V-610)

2026-08-06 04:11:47 +04:00

decisiontrace.go

Wire the heads between stage 0 and the resident model (V-664)

2026-08-08 22:24:37 +04:00

dumpprompt_test.go

Dump the router prompt where the labeler can read it (V-661)

2026-08-08 18:44:47 +04:00

embedder_test.go

Embed a question and a stored note differently (Vikunja #371)

2026-07-31 11:38:07 +04:00

embedder.go

Record which embedder wrote the stored vectors and warn on a swap (#378)

2026-07-31 13:42:07 +04:00

embedderid_test.go

The tokenizer read every long word backwards (V-664)

2026-08-08 22:23:56 +04:00

factparser_corpus_test.go

the fact parser gets a corpus, and the LLM arm gets run (V-586)

2026-08-06 12:21:50 +04:00

fastpath_test.go

router: update all Route callers for NormalizedInput + add invariant tests (slice 11)

2026-09-07 01:32:49 +04:00

fastpath.go

router: add TryFastPath with stage-0 grammar evaluation (slice 10)

2026-09-07 01:24:25 +04:00

feedgrammar_test.go

router: update all Route callers for NormalizedInput + add invariant tests (slice 11)

2026-09-07 01:32:49 +04:00

feeds_test.go

rss: stop claiming "что нового" and stop re-noting the same items

2026-08-01 14:21:36 +04:00

feeds.go

feed questions ask the seeds, not a stem list (V-522)

2026-08-05 18:37:38 +04:00

fragment_test.go

Add seven stage 0 frames and tighten three more (V-720)

2026-08-15 17:18:48 +04:00

fragment.go

Add seven stage 0 frames and tighten three more (V-720)

2026-08-15 17:18:48 +04:00

grammar_bounds_test.go

grammar: bound ws, so phrasing stops when it is done (V-531)

2026-08-05 01:25:42 +04:00

habit_test.go

router: let a habit question outrank the day plan, and know the weekend

2026-08-01 14:06:05 +04:00

habit.go

one weekday matcher, and a stem list stops answering for sredstvo (V-581)

2026-08-06 03:33:01 +04:00

halfpast_test.go

half-past hours parse as the hour being entered (V-538)

2026-08-05 14:24:49 +04:00

halfpast.go

ordinal selection asks the lexicon, and declines a half hour (V-522)

2026-08-05 18:16:54 +04:00

heads.go

Make aggregate ONNX gates execute for real

2026-08-13 03:03:25 +04:00

help_test.go

Add seven stage 0 frames and tighten three more (V-720)

2026-08-15 17:18:48 +04:00

help.go

Add seven stage 0 frames and tighten three more (V-720)

2026-08-15 17:18:48 +04:00

hourunit_test.go

The hour unit has one home and it carries the dative plural (V-609)

2026-08-06 03:51:06 +04:00

implicitquery_test.go

Add seven stage 0 frames and tighten three more (V-720)

2026-08-15 17:18:48 +04:00

implicitquery.go

Add seven stage 0 frames and tighten three more (V-720)

2026-08-15 17:18:48 +04:00

intent.go

router: add CapabilitySelection field to Decision (slice 6b)

2026-09-06 21:02:09 +04:00

knowledge_test.go

Don't tell her she can't talk

2026-07-31 16:07:35 +04:00

knowledge.go

Stop asking for Russian in English on the last three prompts (V-404)

2026-08-02 10:01:58 +04:00

kpreposition_test.go

The hour after "к" is read, and an hour nobody read is asked about (V-610)

2026-08-06 04:11:47 +04:00

list_test.go

list: a named task list is not a grocery item (V-520)

2026-08-05 11:44:07 +04:00

list.go

list read-back asks the seeds; add and clear keep their tables (V-522)

2026-08-05 18:43:13 +04:00

llmrouter_test.go

router: update all Route callers for NormalizedInput + add invariant tests (slice 11)

2026-09-07 01:32:49 +04:00

llmrouter.go

Give the router prompt a destination to write (V-660)

2026-08-08 18:22:00 +04:00

matchtext_test.go

router: add NormalizeMatchText tests and property invariants (slice 9b)

2026-09-06 22:33:27 +04:00

matchtext.go

router: add MatchText to NormalizedInput and NormalizeMatchText function (slice 9a)

2026-09-06 22:32:51 +04:00

money_test.go

money and list: the dictionary matches the word (V-529)

2026-08-04 19:17:59 +04:00

money.go

money and list: the dictionary matches the word (V-529)

2026-08-04 19:17:59 +04:00

narrative_test.go

router: update all Route callers for NormalizedInput + add invariant tests (slice 11)

2026-09-07 01:32:49 +04:00

notecapture_test.go

Add seven stage 0 frames and tighten three more (V-720)

2026-08-15 17:18:48 +04:00

notecapture.go

Add seven stage 0 frames and tighten three more (V-720)

2026-08-15 17:18:48 +04:00

numwords_test.go

half-past hours parse as the hour being entered (V-538)

2026-08-05 14:24:49 +04:00

numwords.go

The hour unit has one home and it carries the dative plural (V-609)

2026-08-06 03:51:06 +04:00

onnxembedder.go

Make aggregate ONNX gates execute for real

2026-08-13 03:03:25 +04:00

onnxruntime_test.go

Make aggregate ONNX gates execute for real

2026-08-13 03:03:25 +04:00

onnxruntime.go

Make aggregate ONNX gates execute for real

2026-08-13 03:03:25 +04:00

possession_test.go

router: update all Route callers for NormalizedInput + add invariant tests (slice 11)

2026-09-07 01:32:49 +04:00

possession.go

Harden semantic boundaries and repair dialogue state

2026-08-13 03:00:31 +04:00

praxis_test.go

Add seven stage 0 frames and tighten three more (V-720)

2026-08-15 17:18:48 +04:00

praxis.go

router: set ResolvedBy at each function selection point

2026-09-06 13:50:03 +04:00

question_test.go

Add seven stage 0 frames and tighten three more (V-720)

2026-08-15 17:18:48 +04:00

question.go

Add seven stage 0 frames and tighten three more (V-720)

2026-08-15 17:18:48 +04:00

reminderreport_test.go

Add seven stage 0 frames and tighten three more (V-720)

2026-08-15 17:18:48 +04:00

reminderreport.go

Add seven stage 0 frames and tighten three more (V-720)

2026-08-15 17:18:48 +04:00

remindersubject_test.go

a bare "напомни" asks instead of failing to parse (V-548)

2026-08-05 19:36:29 +04:00

remindersubject.go

a bare "напомни" asks instead of failing to parse (V-548)

2026-08-05 19:36:29 +04:00

rollforward_test.go

a clock already past rolls to its next occurrence (V-544)

2026-08-05 15:14:29 +04:00

router_test.go

router: update all Route callers for NormalizedInput + add invariant tests (slice 11)

2026-09-07 01:32:49 +04:00

router.go

router: Router.Route consumes NormalizedInput directly (slice 11)

2026-09-07 01:32:45 +04:00

singletoken_test.go

Add seven stage 0 frames and tighten three more (V-720)

2026-08-15 17:18:48 +04:00

singletoken.go

Add seven stage 0 frames and tighten three more (V-720)

2026-08-15 17:18:48 +04:00

slots_ru_test.go

Preserve context across conversation intents (V-542)

2026-08-13 02:14:46 +04:00

slots.go

router: set ResolvedBy at each function selection point

2026-09-06 13:50:03 +04:00

source.go

router: add MatchText to NormalizedInput and NormalizeMatchText function (slice 9a)

2026-09-06 22:32:51 +04:00

stage0.go

router: set ResolvedBy at each function selection point

2026-09-06 13:50:03 +04:00

stagezero.go

Add seven stage 0 frames and tighten three more (V-720)

2026-08-15 17:18:48 +04:00

task_phrases.go

Address PR review comments on 50, 52, 53, 54, 59, 61

2026-08-01 12:50:47 +04:00

task_phrases.json

list: a named task list is not a grocery item (V-520)

2026-08-05 11:44:07 +04:00

task_test.go

router: answer the task-list ask at stage 0 too, and mirror it in the fixture (V-467)

2026-08-04 04:29:37 +04:00

task.go

router: answer the task-list ask at stage 0 too, and mirror it in the fixture (V-467)

2026-08-04 04:29:37 +04:00

taskstatus_test.go

Add seven stage 0 frames and tighten three more (V-720)

2026-08-15 17:18:48 +04:00

taskstatus.go

router: set ResolvedBy at each function selection point

2026-09-06 13:50:03 +04:00

timementions_test.go

The hour after "к" is read, and an hour nobody read is asked about (V-610)

2026-08-06 04:11:47 +04:00

timementions.go

The hour after "к" is read, and an hour nobody read is asked about (V-610)

2026-08-06 04:11:47 +04:00

url_test.go

Read a web page when he names one, and watch a few on a timer (#259)

2026-08-01 03:40:21 +04:00

url.go

Read a web page when he names one, and watch a few on a timer (#259)

2026-08-01 03:40:21 +04:00

weekday_test.go

one weekday matcher, and a stem list stops answering for sredstvo (V-581)

2026-08-06 03:33:01 +04:00

worldquery_test.go

Add seven stage 0 frames and tighten three more (V-720)

2026-08-15 17:18:48 +04:00

worldquery.go

Add seven stage 0 frames and tighten three more (V-720)

2026-08-15 17:18:48 +04:00