Skip to content

v1.12.0

Choose a tag to compare

View all tags
@github-actions github-actions released this 11 Oct 04:11
· 90 commits to main since this release
v1.12.0
4df7c73
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Install

curl -fsSL https://raw.githubusercontent.com/JeremiahM37/grimoire/main/install.sh | sh   # Linux, macOS
brew install JeremiahM37/tap/grimoire                                                   # Homebrew
scoop bucket add jeremiahm37 https://github.com/JeremiahM37/scoop-bucket && scoop install grimoire   # Windows
docker run -p 9111:9111 -v grimoire-vault:/vault ghcr.io/jeremiahm37/grimoire:1.12.0
go install github.com/JeremiahM37/grimoire/go/cmd/grimoire@latest    # the module lives in go/, so @latest resolves the go/v1.12.0 tag

Windows PowerShell one-liner: irm https://raw.githubusercontent.com/JeremiahM37/grimoire/main/install.ps1 | iex.
Archives (binary beside web/ and plugins/: extract and run), .deb, .rpm and checksums.txt are attached below.

Changelog

  • 146767e Add attest: resolve evidence references against the store
  • ab18a55 Add calibrate package: learned agent reliability and retention ordering
  • ff57a62 Add conflict: a deterministic score for the sides of a dispute
  • bcefda0 Add distill package: bake-in selection and deterministic example export
  • 0b4d281 Add grimoire memory distill: read-only export of bake-in candidates
  • 3e0b9d8 Add judge-validity harness, triples, offline tests and live results
  • 36a1ceb Add phrase families to the injection check
  • fc6cb88 Add pixel signatures and an opt-in captioner to memimage
  • 266abbf Add second fresh set with a fixed dev and held-out split
  • 79aec52 Add shared live-call budget and cap BEAM live runs with it
  • 25992e5 Add structured skill specs, predicates and the offline skill gate
  • bb4e73a Add task-benefit study harness with offline tests and live results
  • 63acdc1 Add the sealed visibility and its bullet fields
  • 6b83956 Add the sealed-fact key store and the tombstone log
  • 62b8742 Add visual similarity lookup and automatic captions to image memory
  • 8d91a11 Authenticate tombstones under the session key
  • e563786 Benchmarks: make zip strict so the harness lints clean
  • a6f9e57 Canonicalise text into the readings a model would act on
  • bbca6cc Context overlap: stem terms so a paraphrased prompt still meets the stored fact
  • 619ec25 Disputes: score suggestions with learned per-agent reliability
  • c3fd5c9 Document calibration: measured simulation, one-event limit, open questions
  • 5856154 Document injection defence measurements and open gaps
  • 2172f25 Document recall scale diagnosis and the 80% target result
  • 2237059 Document sealed facts, tombstones and their limits
  • 95e6e68 Document the parametric memory recipe and its limits
  • 89d20b3 Document the phrase families and their measured limits
  • b27a74e Document the skill gate and its limits
  • 39fcd18 Document visual similarity and opt-in captions
  • 1d2d06e Document what provenance verifies, what it scores, and what is open
  • 365d9bb Export shell-outcome helpers and record a gate verdict on candidates
  • f92c610 Gate skills before accept and list applicable skills
  • cde69c5 Graph: drag, wheel and pinch work again; the layer sigma listens on stays attached
  • ab2d90c Hook correction events into memory writes and expose agent reliability
  • 32add26 Index picture fingerprints and register caption settings
  • f1a961d Judge every reading of a line in the injection check
  • ec0662f Limits doc: record what the depth round changed and what it did not
  • 11adfdc Measure injection recall and benign false positives offline
  • 2ac0cd6 Merge branch 'feat/md-attest' into feat/memory-depth
  • e1e2d56 Merge branch 'feat/md-calibrate' into feat/memory-depth
  • ad82c99 Merge branch 'feat/md-codegraph' into feat/memory-depth
  • 770de3c Merge branch 'feat/md-distill' into feat/memory-depth
  • 1b1c9de Merge branch 'feat/md-evalkit' into feat/memory-depth
  • 815d793 Merge branch 'feat/md-image' into feat/memory-depth
  • a604e0b Merge branch 'feat/md-inject' into feat/memory-depth
  • edbfab2 Merge branch 'feat/md-recall' into feat/memory-depth
  • ce44059 Merge branch 'feat/md-seal' into feat/memory-depth
  • af4964e Merge branch 'feat/md-seal' into feat/memory-depth
  • 9e215b7 Merge branch 'feat/md-skills' into feat/memory-depth
  • 4df7c73 Merge pull request #5 from JeremiahM37/feat/memory-depth
  • de62955 Narrow injection patterns to context and add non-English core phrases
  • 1a43b97 Note list: raise the default cap to 1000
  • 2275cce Note list: written notes come before what a connector pulled in, so a mailbox sync cannot crowd them out of the cap
  • 385d568 Record BEAM Haiku slice, evalkit call ledger and status doc
  • f96636a Record whether evidence resolved, and cap unverified evidence at inferred
  • 675aee3 Refuse model rewrites of memory notes holding pulled entries
  • d72dc41 Resolve code-graph call edges by type (Go) and scope (Python, TS/JS)
  • 2987ad8 Score exact runs of a query's name words in recall
  • 99138cd Score held-out halves, a fresh set and a larger benign set
  • ba3c1e8 Score second fresh set in dev and held-out halves
  • 3d8183d Seal and attest: fix the findings from the security review
  • 688a139 Seal facts on write, reveal them to their owner, tombstone forgets
  • a90398c Sealed facts: stop consolidate, cascade forget and note writes from bypassing them
  • be31967 Show call confidence in the callers API, CLI and code-graph docs
  • e0e25a4 Strip tombstoned bullets from cloud-folder pulls at write time
  • 26c5a4f Suggest a winner in the disputes listing, with its breakdown
  • 0b8d03c Test sync routes end to end for forgotten facts
  • 59719e6 Tests: use escape sequences for format characters and drop a self-division
  • 984f3da Verify evidence when an agent writes a fact
Assets 13
Loading