Commit Graph

  • 09f1696fce Merge the eval label and kill script fixes overnight/eval-label-kill kami 2026-07-31 14:32:45 +04:00
  • 80f7322294 Don't fail when docker confirms nothing is running kami 2026-07-31 14:32:45 +04:00
  • fa5aebfbe4 Merge the delivery boundary fixes overnight/delivery-boundary kami 2026-07-31 14:30:54 +04:00
  • 59cec63da1 List the columns in the table rebuild kami 2026-07-31 14:30:54 +04:00
  • 02e8786695 Stop the containers instead of claiming success (#380) kami 2026-07-31 14:30:51 +04:00
  • 0272dc9d89 Record a suppressed care nudge instead of dropping it silently (#370) kami 2026-07-31 14:27:47 +04:00
  • 2ad7635501 Merge the address-form eval check overnight/address-check kami 2026-07-31 14:27:16 +04:00
  • 9949b309b1 Don't let a time word blind the third-person check kami 2026-07-31 14:27:08 +04:00
  • a788ca3915 Label eval runs with the model the server actually loaded (#379) kami 2026-07-31 14:25:28 +04:00
  • 62d47d28ac Add an eval check for formal and third-person address (#384) kami 2026-07-31 14:25:23 +04:00
  • e9ff2c4912 Never send the full nudge body off-box (#368) kami 2026-07-31 14:23:54 +04:00
  • 3dbf67f8f9 Drop the city time-zone table overnight/system-replies-pr kami 2026-07-31 14:14:18 +04:00
  • 84ba217892 Say so when the day asked about is out of reach kami 2026-07-31 14:05:21 +04:00
  • f179ae2fde Merge the system reply fixes kami 2026-07-31 14:03:42 +04:00
  • d00929ac0b Answer the day the user asked about and the city he named (#388) kami 2026-07-31 14:02:57 +04:00
  • b6f47fbeb6 Merge the clock and calendar routing rule overnight/clock-intent-pr kami 2026-07-31 13:53:35 +04:00
  • 2e9b9ec1cf Warn separately when a row has no text to re-embed overnight/embedder-backfill-pr kami 2026-07-31 13:52:56 +04:00
  • bfb57c3148 Give the router prompt a rule for clock and calendar questions (#374) kami 2026-07-31 13:52:00 +04:00
  • d1f6f6355f Merge the vector backfill kami 2026-07-31 13:51:25 +04:00
  • 92ecb691de Re-embed stored notes and facts after an embedder swap (#378) kami 2026-07-31 13:50:43 +04:00
  • 4282f6b9a9 Warn about vectors written before the marker existed overnight/embedder-marker-pr kami 2026-07-31 13:44:29 +04:00
  • 7bb9f9be06 Merge the embedder marker kami 2026-07-31 13:42:40 +04:00
  • 1e47eaca5a Record which embedder wrote the stored vectors and warn on a swap (#378) kami 2026-07-31 13:42:07 +04:00
  • 892330eb84 Merge the note recall fix overnight/note-recall-pr kami 2026-07-31 13:33:14 +04:00
  • 9a3bcd7c46 Merge the thinking-off measurement overnight/thinking-off-pr kami 2026-07-31 13:31:31 +04:00
  • 98ee701e03 Let a note win a recall, not only a fact (#373) kami 2026-07-31 13:30:38 +04:00
  • 04c1088088 Measure thinking off on routing properly — it does not win (#376) kami 2026-07-31 13:30:22 +04:00
  • 07c191d8b8 Merge dialogue session persistence overnight/dialogue-persist-pr kami 2026-07-31 13:21:03 +04:00
  • c668310b3e Persist the dialogue session so a restart keeps the conversation kami 2026-07-31 13:20:30 +04:00
  • 1bd2acdc2a Do not exempt Russian words that are both noun and verb overnight/persona-2p-pr kami 2026-07-31 12:55:45 +04:00
  • 15e5dd8eaa Merge the second-person gender check kami 2026-07-31 12:54:48 +04:00
  • 10cf6f525c Check that nudges do not address the owner in the feminine kami 2026-07-31 12:54:18 +04:00
  • e2210f6844 Merge the clarify-expiry notice overnight/clarify-expiry-pr kami 2026-07-31 12:52:36 +04:00
  • 214a4032cf Tell him when an expired clarify question is dropped kami 2026-07-31 12:51:47 +04:00
  • dc70a5a7ab Show clarify_max_attempts in the deployed config overnight/clarify-rework kami 2026-07-31 12:34:28 +04:00
  • 06aded6ab0 Merge commit 'd2be98e' into overnight-jul31 kami 2026-07-31 12:34:00 +04:00
  • d2be98ee2a Say out loud when she gives up instead of dropping the request kami 2026-07-31 12:31:55 +04:00
  • 62d320f93a Let her ask three times, and let a restated answer win kami 2026-07-31 12:31:46 +04:00
  • 796e6af3cf Merge commit '74a7088' into overnight-jul31 overnight/phrasing kami 2026-07-31 12:24:20 +04:00
  • 74a70880a8 Write up the phrasing eval: 0/15 to 13/15, and what the number hides kami 2026-07-31 12:23:54 +04:00
  • a40bc559d5 Fix the nudge phrasing prompt: stop teaching the model to echo the example kami 2026-07-31 12:23:03 +04:00
  • 1db0fcfcd0 Merge commit '4ca68d2' into overnight-jul31 overnight/bakeoff kami 2026-07-31 12:07:38 +04:00
  • 4ca68d2f3f Bake off LFM2.5 against Qwen3.5-0.8B on the RU routing fixture kami 2026-07-31 12:07:06 +04:00
  • 5a1d465db5 Merge commit '4383844' into overnight-jul31 overnight/recall-margin kami 2026-07-31 11:52:21 +04:00
  • 43838445ab Write up the margin gate results kami 2026-07-31 11:51:21 +04:00
  • 11831c6ace Gate recall on the margin over the runner-up, not just the score kami 2026-07-31 11:50:39 +04:00
  • 93c1a41d4a Route with the resident model by default overnight/router-on kami 2026-07-31 11:47:57 +04:00
  • bce5ed210c Merge branch 'worktree-agent-a76ce40c73601d90d' into overnight-jul31 overnight/slot-extract kami 2026-07-31 11:44:32 +04:00
  • c31f0d1001 Extract slots for LLM router decisions too kami 2026-07-31 11:44:06 +04:00
  • 34521c30b8 Merge branch 'worktree-agent-ad5da57e47b822152' into overnight-jul31 overnight/embedder-e5 kami 2026-07-31 11:39:35 +04:00
  • 1d48755d12 Record the recall numbers after the embedder swap kami 2026-07-31 11:38:18 +04:00
  • f6d5a2a7a4 Swap the embedder to multilingual-e5-small (Vikunja #371, #372) kami 2026-07-31 11:38:18 +04:00
  • 751c2a705f Embed a question and a stored note differently (Vikunja #371) kami 2026-07-31 11:38:07 +04:00
  • 5d5b0cfd49 Merge branch 'worktree-agent-a0ab2a9b7439296e3' into overnight-jul31 overnight/router-refusal kami 2026-07-31 11:37:53 +04:00
  • bd16ca69e5 Let the LLM router answer "unknown" when it cannot route kami 2026-07-31 11:35:37 +04:00
  • 0ed386eca6 Re-measure the router on a quiet box and record the numbers overnight/eval-rerun kami 2026-07-31 11:34:57 +04:00
  • 1c4eab2107 Merge commit '94eb92f' into overnight-jul31 overnight/eval-rerun-base overnight/eval-harnesses kami 2026-07-31 10:11:56 +04:00
  • 0914e0a3d5 Merge commit '4ba9a6f' into overnight-jul31 kami 2026-07-31 10:11:29 +04:00
  • c860808528 Make make test actually gate on gofmt and vet overnight/fmt-gate kami 2026-07-31 10:10:27 +04:00
  • f7442c3aea Run gofmt over the seven files that had drifted kami 2026-07-31 10:09:32 +04:00
  • 75b067ac51 Merge the accepted-routine fix and drop reminder_id from accept overnight/routines-fire kami 2026-07-31 10:07:37 +04:00
  • 424d1b3446 Fire accepted routines every interval, not once (Vikunja #366) kami 2026-07-31 02:45:30 +04:00
  • c47886c2bc Merge branch 'worktree-agent-ab5b5c61a32cac4fe' into overnight-jul31 overnight/router-prompt kami 2026-07-31 02:42:53 +04:00
  • f6236da760 Collapse the duplicate away-detail and panic tests overnight/away-leak kami 2026-07-31 02:42:21 +04:00
  • 54dc43516b Add accepted-routine timestamps to the store (Vikunja #366) kami 2026-07-31 02:41:49 +04:00
  • fa51a48958 Merge branch 'worktree-agent-afe3f2ec18b2b8497' into overnight-jul31 kami 2026-07-31 02:40:10 +04:00
  • 5fd25d7ad7 Test the away-channel minimal body and the panicking sink (#368, #369) kami 2026-07-31 02:39:38 +04:00
  • c22fc352fc Merge branch 'worktree-agent-a1610b8c5376eadd6' into overnight-jul31 kami 2026-07-31 02:37:30 +04:00
  • 215aa331c5 Recover from a panicking sink so the attempt is always closed (#369) kami 2026-07-31 02:35:54 +04:00
  • 859bbf750f Never send a nudge body off-box when the summary is empty (#368) kami 2026-07-31 02:34:54 +04:00
  • 8a174c1c70 Score the recall fixture and write up what it shows overnight/recall-eval kami 2026-07-31 02:34:28 +04:00
  • ee3e6a9eaf Wire the snooze read into the Gatherer and honour it for reminders (#364) overnight/snooze-works kami 2026-07-31 02:32:24 +04:00
  • 8acb8a97c6 Read the recorded snooze outcomes back out of the nudges table (#364) kami 2026-07-31 02:29:51 +04:00
  • 1f55207b58 Merge the snooze read path and wiring kami 2026-07-31 02:33:41 +04:00
  • 4db109346a Ignore the .claude directory kami 2026-07-31 02:32:44 +04:00
  • 2f00593411 Wire the snooze read into the Gatherer and honour it for reminders (#364) kami 2026-07-31 02:32:24 +04:00
  • 0b3b8d0a9e Test the clarify round-trip end to end at the daemon level overnight/clarify-wiring kami 2026-07-31 02:29:04 +04:00
  • fe0e654ab1 Ask the question, then act on the answer kami 2026-07-31 02:26:11 +04:00
  • a54ebac0cb Work out which slot is missing and phrase one short question kami 2026-07-31 02:22:18 +04:00
  • 784d688b44 Merge the clarify wiring kami 2026-07-31 02:30:59 +04:00
  • 4ba9a6f422 Add a deterministic scorer for nudge phrasing (Vikunja #323) kami 2026-07-31 02:30:52 +04:00
  • 32687b3712 Read the recorded snooze outcomes back out of the nudges table (#364) kami 2026-07-31 02:29:51 +04:00
  • db3e706bdc Merge the delivery routing and durability tests kami 2026-07-31 02:29:05 +04:00
  • 2f4257e194 Test the clarify round-trip end to end at the daemon level kami 2026-07-31 02:29:04 +04:00
  • 7d8b0af99d Test voice fallthrough and delivery durability overnight/delivery-tests kami 2026-07-31 02:27:33 +04:00
  • cfd38d53cf Ask the question, then act on the answer kami 2026-07-31 02:26:11 +04:00
  • a2835bbdf6 Cover every cell of the delivery routing table kami 2026-07-31 02:24:43 +04:00
  • c262c1e4c0 Merge the routine accept path and store gaps kami 2026-07-31 02:24:07 +04:00
  • a33ad82178 Let the /routines page accept a proposal, gated at step-up (#46) overnight/routine-accept kami 2026-07-31 02:23:28 +04:00
  • af35ec3629 Work out which slot is missing and phrase one short question kami 2026-07-31 02:22:18 +04:00
  • 9145b83100 Add the clarify data layer: a parked question with one missing slot kami 2026-07-31 02:20:28 +04:00
  • 43470abc57 Add a held-out note-recall harness (fixture + scorer) kami 2026-07-31 02:18:58 +04:00
  • 94eb92fb15 Label LLM eval runs with the model llama-server has loaded kami 2026-07-31 02:17:20 +04:00
  • 707c3e5040 Ignore deps and models as symlinks, not just directories kami 2026-07-31 02:17:15 +04:00
  • ace9fbbc06 Merge the llm_router flag and the kill-maven fix kami 2026-07-31 02:17:00 +04:00
  • 3884db33e9 Give proposed routines a status filter and pin down the dedup rule (#46) kami 2026-07-31 02:16:39 +04:00
  • dfb8d26b62 Fix kill-maven.sh so it actually kills llama-server overnight/llm-router-flag kami 2026-07-31 02:15:30 +04:00
  • bf99fd4192 Add a voice.llm_router flag, default off kami 2026-07-31 02:15:30 +04:00
  • 6d9aa83b6e Merge the proactive rule and gate tests kami 2026-07-31 02:15:15 +04:00
  • f75072175d Merge the pending-question data layer kami 2026-07-31 02:14:33 +04:00