SillyMemory 0.7.3 (experimental)

SillyMemory 0.7.3 — experimental pre-release

Published 2026-10-28 from public main commit
010bdaf3ecb399abee1a46a87c29369a7755966f.

Changes from 0.7.2

Update the transitive Zod lockfile resolution from 4.6.5 to 4.6.7, outside the
affected range of GHSA-mrhx-c7xc-m3w5.
The pinned LambdaDB SDK remains 0.9.0. The shipped browser bundle and extension
runtime are byte-identical to 0.7.2; Git installations use that bundle, so the
lockfile update does not change installed runtime behavior.

The README now links the existing 2:05 product walkthrough. It demonstrates
memory recall, source inspection, input token comparison, checkpoints and chat
branches with synthetic dialogue and scripted retrieval/replies. Its token
figures are host input estimates, not billing or typical savings.

Supported host: SillyTavern 1.19.0 at
06bde939fb1e9c4c8d8641d810f0a916b5bce127, with Chromium.
Retrieval defaults, credentials, storage formats and native reply/streaming
behavior are unchanged.

Install, update and rollback

Install the repository URL with an empty branch field to follow main.
Use Extensions → Manage extensions → SillyMemory → Update, then reload.
No npm installation, build step, storage migration or credential re-entry is
introduced by this patch. The update is available from public main.

Preserve native chats and backups before updating. For rollback, disable memory
and pin published v0.7.2 following the release procedure.
Code rollback does not undo remote writes or restore deleted data.

Validation and limits

See the acceptance record for measured checks and limits.
This release makes no new answer-quality, live-provider compatibility or
server-rollout claim. Other browsers, host versions, group/tool conversations
and concurrent multi-device writers remain outside the supported scope.

Main CI and
tag CI passed on Node.js 20.12.0 and 24.

SillyMemory 0.7.2 (experimental)

SillyMemory 0.7.2 — experimental pre-release

Prepared release boundary: 2026-10-26. Main promotion, tags and GitHub Release
publication remain separate actions.

Changes from 0.7.1

Update the LambdaDB JavaScript SDK from development version
0.8.1-dev.38730896042001 to stable 0.9.0, including the shipped browser bundle.
Retrieval still uses 200 rerank candidates and returns up to 50 passages.
Credential scope, storage formats, recent-history boundaries and native streaming
remain unchanged. No Python SDK is introduced.

Supported host: SillyTavern 1.19.0 at
06bde939fb1e9c4c8d8641d810f0a916b5bce127, with Chromium.

Install, update and rollback

Follow the installation guide.
Install the repository URL with an empty branch field to follow main.
Use Extensions → Manage extensions → SillyMemory → Update, then reload.
The SDK bundle ships in Git; no npm installation or compilation is required.
No storage migration or credential re-entry is introduced by this patch.

Preserve native chats and backups before updating. For rollback, disable memory
and pin published v0.7.1 following the release procedure.
Code rollback does not undo remote writes or restore deleted data.

Validation and limits

See the acceptance record for measured checks and limits.
This update makes no new answer-quality or server-rollout claim. Other browsers,
host versions, group/tool chats and concurrent multi-device writers remain outside
the supported scope. No live provider evaluation was added.

SillyMemory 0.7.1 (experimental)

Experimental release for SillyTavern 1.19.0 with Chromium.

Changes

• Frozen memory snapshots preserve the complete native JSONL transcript while indexing only the older-history partition used by ordinary synchronization. Recent messages stay in the transcript and are archived normally when later dialogue displaces them on a resumed writable path.
• Documentation and development tools focus on current behavior and final benchmark results. Retired experiment implementations and cohort-only fixtures have been removed; original evidence is preserved.
• The retained summary comparator resolves its maintained source files when recording the source digest.
• The isolated BEAM Python environment uses certifi==2026.10.19.

Install and update

In SillyTavern, choose Extensions → Install extension and enter:

https://github.com/lambdadb/SillyMemory

Leave the branch field empty to follow main. Existing installations can use Manage extensions → SillyMemory → Update, then reload. Updates follow the installed Git branch; automatic updates remain disabled. No npm install, compilation or server restart is required.

Existing 0.7.0 per-chat activation, saved connection scope, collection/index formats, retrieval defaults and native reply/streaming behavior are unchanged. When updating from 0.6.0, enable memory separately in each desired chat.

Installation guide · Update and rollback · Benchmark profiles

Validation and limits

The release passes 244 JavaScript tests and 32 Python tests, syntax, SDK and release checks. Candidate installation and native Update on the pinned host passed 22 assertions; the release preserves that candidate's browser runtime and version metadata. See the acceptance record for the exact measured commits and scope.

This remains experimental. Other host versions and browsers are unverified; group/tool conversations and concurrent multi-device writers are unsupported. Code rollback does not restore local or remote data. See current limits.

SillyMemory 0.7.0 (experimental)

SillyMemory 0.7.0 — experimental pre-release

Release boundary: 2026-10-22. This experimental release promotes the reviewed
0.7.0 code to public main at 9c31f0ac1b3fe405f5916df5b9f6930ae29cd571.

Supported host: SillyTavern 1.19.0 at
06bde939fb1e9c4c8d8641d810f0a916b5bce127, with Chromium.
Other host versions and browsers remain unverified.

Changes from 0.6.0

• Turn memory on or off directly in the chat panel. Each saved character chat has
  its own choice; extension settings manage the shared LambdaDB connection.
• Use native chat history, Save checkpoint, Create Branch and Back to parent chat
  as the ordinary workflow. Memory follows the selected path and its saved choice.
  Optional verified transcript/remote-memory saves live under Advanced → Frozen
  memory snapshots. Saving/resuming preserves activation and shows progress.
• Read status colors directly: green for ready/verified, orange for work in
  progress, yellow for warnings and red for errors. Ready and verified prompt
  delivery remain separate states.
• Expand Memory to preview selected source excerpts and inspect the full last-prompt
  memory. The panel shows estimated prepared input and memory tokens when the host
  exposes them. Optional dialogue-text comparison is local, excludes native
  system/character formatting and is not a full memory-off prompt or billing comparison.
• Reconnect from chat using the scoped saved key. Missing/rejected keys direct
  users to Connection settings. Reconnection ignores unsaved form edits and
  preserves per-chat choices; automatic reload/chat reconnect never starts an
  unprepared transport test.
• Complete initial Connect and prepare in one action, with temporary inference
  usage disclosed before connecting. Setup validates/saves the key, verifies test
  cleanup and leaves per-chat activation separate. Interrupted preparation exposes
  cleanup and Finish preparation.
• Refresh visible snapshot lists automatically on opening, chat changes and
  reconnection. List errors stay local; stale responses cannot replace a different
  story. Native input/confirmation popups handle rename/delete, including cancel,
  Escape and changed-chat protection.

Installation, update and activation

Install the repository URL with an empty branch field to follow public main.
Update through Extensions → Manage extensions → SillyMemory → Update,
then reload. Update follows the installed Git branch, not the newest tag/Release.
Automatic updates remain disabled; no build step or server restart is needed.

After updating from 0.6.0, turn memory on in each chat where you want it. The
old global enabled flag is not copied to every chat. Chats without a saved
sillymemoryEnabled preference start off. Existing explicit per-chat choices
from development builds remain saved in native chat metadata. Turning off keeps
stored memory but excludes that chat's automatic synchronization and retrieval.

Saved keys keep their account/endpoint/project scope and reconnect after reload.
Other same-page scripts can access browser storage; it is not a secret vault.
No new key type, server plugin, proxy or host-cookie forwarding is introduced.

Collection/index formats, ownership, Query1, managed embeddings/reranking, recent
history alignment, chunk packing, budgets and native reply/streaming formats are
unchanged from 0.6.0. No new remote migration is introduced. Earlier experimental
collections still require the 0.5.0 index configuration;
updating code does not repair indexes. Git installation/update checks do not prove
existing-collection compatibility or data migration.

Before changing versions, preserve native chats and backups. To roll back, disable
memory and pin published v0.6.0 using the release procedure.
Rollback does not undo remote writes/deletions or restore the previous global
activation choice. Resume normal branch updates only after leaving the pinned tag.

Acceptance and limits

See 0.7.0 acceptance for exact installation/update commits,
commands, retained artifacts and inherited runtime evidence. This preparation
adds no paid-provider or answer-quality experiments. Public-main fresh installation
passed 14 checks on the exact release commit, with zero page errors/provider
requests and disposable-profile removal. Main CI
passed on Node.js 20.12.0 and 24.

Group/tool conversations and concurrent multi-device writers remain unsupported.
Flattened Text Completion verification remains unavailable. Host context/output
space, the complete-query 8 KiB limit and recent-history limits still apply.
Previewed memory, prompt verification and token estimates do not establish answer
correctness, provider receipt, cost savings or new benchmark quality results.

SillyMemory 0.6.0 (experimental)

SillyMemory 0.6.0 — experimental pre-release

Prepared release boundary: 2026-10-21. Dated preparation, main promotion, version
tags and GitHub Release publication are separate steps. Confirm published versions
in GitHub Releases.

Supported host: SillyTavern 1.19.0, commit
06bde939fb1e9c4c8d8641d810f0a916b5bce127, with Chromium.
Other host versions and browsers remain unverified.

Changes from 0.5.0

• Connect, prepare and enable memory in guided steps. Prepare chat memory runs the
  existing synthetic connection test when needed; its temporary data must be
  cleaned up before proceeding. Connected credentials collapse automatically,
  while Advanced and checkpoint controls remain available.
• Open Memory above the native composer for operation progress, last-prompt
  verification and recovery actions. Ready means a supported chat can use memory;
  Last prompt verified describes the final host prompt check, not answer quality
  or provider receipt. Partial/unavailable verification remains explicit.
• Recover synchronization from the chat panel. Retry reply uses the native
  Regenerate action only for an unchanged pending normal/regenerate user turn,
  after rechecking chat/source, connection, readiness and host activity. It is an
  explicit new model request, never an automatic retry. Continue and swipe keep
  their native controls. Turning memory off restores native history behavior.
• See an applicable blocker before changing settings: unsupported chats and saved
  checkpoints take priority over Vector Storage. Built-in Vector Storage is not
  a dependency; SillyMemory connects directly to LambdaDB. Its existing conflict
  notice opens native settings without changing another extension's settings.
• Use the public benchmark reproduction profiles for
  bounded LongMemEval and BEAM evaluation. They are development tools, not a new
  installed retrieval mode or a new release-level quality guarantee.

Query1, managed OpenAI Small embeddings, Bayesian fusion, Jev 1.13.0 candidate 200 /
final 50 reranking, source validation, pair-aligned recent history, chunk packing,
4,096-token memory allocation/maximum and the host one-quarter-context cap keep
their 0.5.0 policies. The SDK and shipped bundle are unchanged. No query rewrite,
summary generation or extra inference phase is added to ordinary retrieval.

Installation, update and data

Follow the installation guide. Install the repository URL with an empty branch field to follow public main.
Use Extensions → Manage extensions → SillyMemory → Update, then reload.
The Update button follows the installed Git branch, not the newest tag/Release.
Automatic updates remain disabled; users do not need to build the extension.

This change introduces no new collection/index format or migration step relative
to 0.5.0. Its index requirements still apply to earlier experimental collections;
updating code does not repair their indexes. Keep original chats and backups.
Installation checks establish Git update and preference preservation, not live
existing-collection compatibility. See 0.5.0 storage requirements.

Credentials keep their account/endpoint/project scope. Browser storage remains
accessible to same-page extensions. No server plugin, host-credential forwarding,
new provider key or proxy is added. For rollback, disable memory, preserve native
chats, and pin the published v0.5.0 tag using the release procedure.
A code rollback does not undo remote writes or restore deleted data.

Acceptance and limits

See 0.6.0 acceptance for exact producer/installation/update
commits, commands, inherited runtime evidence and retained reports. Current
release preparation does not add paid-provider or answer-quality experiments.

Group/tool conversations, arbitrary host/browser compatibility and concurrent
multi-device writers remain unsupported. Flattened Text Completion prompts report
verification unavailable. Host context/output space, the 8 KiB complete-query
limit and recent-window limitations still apply. Existing benchmark scores and
historical managed-service acceptance retain their original protocols and scope.

SillyMemory 0.5.0 (experimental)

SillyMemory 0.5.0 — experimental pre-release

Prepared release boundary: 2026-10-20. Dated notes do not establish publication;
confirm published versions in GitHub Releases.
This boundary captures product changes merged through PR #88 at
cbd27ab470e67af105b2bd56fd3f642c6702cf17, with 0.5.0 installation/host acceptance
merged in PR #90. PR #89's separate BEAM protocol work is not a release prerequisite.
Main promotion, tagging and GitHub Release publication require maintainer authorization.

Supported host: SillyTavern 1.19.0, commit
06bde939fb1e9c4c8d8641d810f0a916b5bce127, with Chromium.
Other host versions and browsers remain unverified.

Changes from 0.4.0

• Query1 searches only the eligible generation anchor in one role-labelled request.
  Recent dialogue stays in generation. Normal input retains at least five eligible
  messages, and a completed reply can retain six to keep an adjacent pair together.
  Continue uses the message being extended; checkpoints still synchronize the full
  completed transcript.
• Bayesian vector/English-text search uses managed OpenAI Small with vector k200,
  native Jev 1.13.0 candidate 200 and final 50, replacing k100/candidate 100. The
  initial and maximum memory allocation remain 4,096 tokens, with the host
  one-quarter-context cap and explicit saved budgets respected.
• Valid chunks are admitted in retrieval order using their final assembled or
  separate size. Selected sources are presented chronologically in one labelled,
  quoted user-context archive. Source identities, original text and recent native
  roles remain intact; archive framing and shared guidance count toward the budget.
• Reranking receives source role, verified host-time context and original text.
  Embedding and English full-text search still use original text only. Unknown
  dates remain unknown; host send times do not establish event dates, and UTC alone
  does not establish a speaker's local date. BEAM batch reference dates are
  evaluation metadata, never product send times.
• Natural guidance supports personalized answers and distinguishes event/role
  matches without a blanket follow-up-question obligation. Native reply formats
  and streaming remain unchanged. Memory preparation failures still stop generation;
  disabling memory restores ordinary native history.
• The shipped SDK is pinned to 0.8.1-dev.38730896042001, the tested candidate-200
  build. No user build step or additional embedding-provider key is needed.

Large embeddings, enriched retrievalText, 8,192-token memory and generated summaries
remain unadopted. There is no extra query-rewrite or generation model phase.
See the product decision,
current defaults and
BEAM evaluation for the separate measured comparisons.
Their historical scores are not a new 0.5.0 answer-quality result or a guarantee.

Installation, update and storage compatibility

Public installs should continue following main through
Extensions → Install extension with https://github.com/lambdadb/SillyMemory
and an empty branch field. Before main promotion, reviewers can select develop
using a disposable host/profile and a fresh test chat, checking the exact installed
commit against the acceptance record. The measured preparation branch was removed
after PR #90 merged; a new branch installation is a new measurement.
Do not install a second copy in an existing profile. The Update button follows
its installed Git branch, not the newest tag or GitHub Release; reload after updating.
Automatic updates remain disabled.

Older experimental collections are not automatically migrated. The role and
conversation-time rerank fields require new keyword indexes. A code update does
not add these indexes to an existing remote collection. Start with a fresh owned
story collection; older collections require explicit owned deletion and
resynchronization before this policy can be used. Preserve original chats and
backups first, and review the affected story branches/checkpoints before any
collection deletion. Do not delete unrelated collections or treat installation
checks as proof of existing-data compatibility. See
the index policy and
storage controls.

Verified browser credentials remain scoped to the native account and exact
endpoint/project. Reload verifies the saved key before restoring the prior enabled
state. Forget key removes it and disables memory. Browser storage is not an isolated
secret vault. No server plugin, SillyTavern credential forwarding or proxy is added.

For rollback, disable memory and preserve native chats first. Pin a known-good
published tag using the release procedure.
Code rollback does not restore remote/local data or reverse collection deletion.

Acceptance and remaining limits

See 0.5.0 acceptance for exact commits, installation/update,
local host/emulator results and retained evidence. These checks are distinct from
historical live LambdaDB acceptance and generated-answer evaluations.

Chat Completion final delivery is checked; other completion formats may report
verification unavailable. Sufficient host context/output space is still required.
Queries above 8 KiB fail explicitly; long ambiguous exchanges can outgrow the recent
window. Correct retrieval does not guarantee correct answers. No arbitrary-provider,
multilingual, group-chat or multi-device concurrent-writer support is established.
A wider candidate pool can increase reranking latency and cost; this preparation
makes no new service-latency, billing or benchmark claim.

SillyMemory 0.4.0 (experimental)

SillyMemory 0.4.0 — experimental pre-release

Long-term memory for SillyTavern. Powered by LambdaDB.

Prepared release boundary: 2026-10-13. Confirm publication through the
GitHub Release.
Dated notes alone do not promote main or publish a release; those actions require
maintainer authorization. Supported host: SillyTavern 1.19.0, commit
06bde939fb1e9c4c8d8641d810f0a916b5bce127; other versions remain unverified.

User-visible changes

• Preserve at least five recent eligible messages in native order and roles,
  keeping an adjacent input/reply pair together. Normal input retains five;
  a completed reply can retain six. Archive displaced completed pairs together
  as separate message chunks, not a delayed five-message write queue.
• Search the two latest eligible messages in one role-labelled request. Bayesian
  vector/English-text fusion uses managed Jev reranking, vector k=100, at most
  100 candidates and 50 returned chunks. The initial memory budget is 4,096 tokens,
  capped at one quarter of the host context. An explicit saved budget is respected.
• Assemble touching/overlapping chunks from the same source without repeating
  provenance headers. Available host-message timestamps accompany excerpts and
  are explicitly distinct from dates mentioned in the story. No LLM preprocessing
  or additional retrieval request is added.
• Keep native replies, model/API formats and streaming. Natural guidance asks for
  clarification of ambiguous intent and acknowledges missing/conflicting evidence.
  There is no forced JSON response or accumulated clarification state.
• Save a verified key for this native account and exact endpoint/project in a
  dedicated browser record. Reload verifies it before restoring the previous
  memory-enabled state. Forget key removes it and disables memory. Other code on
  the same page can access browser storage; this is not an isolated secret vault.
• Use story collections and isolated chat branches by default, including explicit
  named checkpoints and inherited committed history. The old opt-in/conversion
  path is removed; there is no existing-user data migration contract.
• Keep preparation failures visible even after background synchronization. A
  failed generation preserves the submitted input and directs retry to Regenerate.

Installation and operation

After authorized promotion, install https://github.com/lambdadb/SillyMemory
from Extensions → Install extension, leaving the branch field empty to follow
main. Existing branch-based installations use Manage extensions → SillyMemory
→ Update, then reload. No npm install, build, server plugin, proxy setting or
host restart is required. LambdaDB must accept the browser origin via direct CORS.

A LambdaDB project key with document, collection and branch permissions is required.
Use Save and connect, run the synthetic connection test, prepare chat memory,
then enable memory for a saved character chat. Managed embeddings and reranking
still incur service costs; no separate embedding-provider key is needed.
Disable memory to use ordinary native history. See the README,
saved credentials and checkpoints.

Evidence and limits

See installation readiness and
bounded native-reply validation. The latter uses
real managed LambdaDB and Sol, including streaming and reconnection; it is a narrow
English UX acceptance run, not an official benchmark score or quality guarantee.
Transient search 503s required explicit regeneration retries; their backend cause
is unknown. A failed preparation stops generation rather than silently continuing.

Final Chat Completion delivery is verified against prepared memory/recent turns;
text-completion delivery remains visibly unverified. Long ambiguous conversations
can outgrow the recent window. Queries above 8 KiB fail explicitly. Small host
contexts may truncate even recent turns; keep sufficient context and output space.
No multi-device concurrent-writer, arbitrary-provider or multilingual quality
support is established. Rollback changes code, not remote/local data; see
release procedures.

Publication verification

Published as an experimental pre-release on 2026-10-13. Main and annotated tag
v0.4.0 both identify efd5df8d5cdecdc95714c3b1f51314c60ed7ae6e. The promotion preserves the
reviewed develop head as the merge commit's second parent; its tree is unchanged.

• Main CI and tag CI passed on Node.js 20.12.0 and 24.
• Actual SillyTavern 1.19.0/Chromium default Git URL installation of public main
  passed 14 assertions, including exact commit/version and preference/ownership
  preservation after reload, with zero page errors or provider calls. Temporary
  profiles were removed. This is installation acceptance, not a new answer-quality run.
• Original report, screenshot, measured producer and publication receipts are
  retained locally in artifacts/archive/release-040-publication-v1/evidence.tar.gz;
  SHA-256 f235a8a4f7e5456891e7fd2c13db128b3f4e923e97be0e6b751d3be944baf5bf. Member bytes were read back and verified. This local-only archive
  is not downloadable from a fresh clone or this Release.

SillyMemory 0.3.0 (experimental)

SillyMemory 0.3.0 — experimental pre-release

Long-term memory for SillyTavern. Powered by LambdaDB.

Release boundary: 2026-10-04. Confirm publication through the
GitHub Release.
Install from https://github.com/lambdadb/SillyMemory, leaving the branch field
empty to follow main; no build step is required.
Supported host: SillyTavern 1.19.0, commit
06bde939fb1e9c4c8d8641d810f0a916b5bce127. Other host versions are unverified.

What changes

• Optional versioned story memory shares committed history across native chat
  branches. Each chat path has a writable LambdaDB branch in its story collection.
  Unchanged inherited documents do not require another embedding submission.
• Named checkpoints save the native transcript, hidden swipes and chat metadata
  together with a retained memory branch. Resume creates a new chat/path; it does
  not overwrite the source. The manager lists ready/pending/invalid states and
  supports retry, resume, rename and individually scoped deletion.
• Interrupted preparation/resume can be retried. Changed checkpoint content,
  duplicate identities, missing branches and mismatched snapshots are rejected.
• Ordered commit confirmation can reduce repeated reads while waiting to fork.
  It keeps conservative full checks when write history is unknown or ambiguous,
  and always verifies the complete checkpoint contents before readiness.

Retrieval, chunking, memory budget defaults and recent-message protection are
unchanged from 0.2.0. This release adds history reuse and checkpoint lifecycle
behavior; it does not claim improved answer accuracy or consistently lower latency.

Update and opt in

Back up SillyTavern data first. Use Manage
extensions → SillyMemory → Update, then reload. Enter the LambdaDB key again
and enable memory when ready. The key remains session-only. Direct browser CORS
and LambdaDB managed embeddings remain in use; no separate embedding-provider key,
server plugin, proxy setting or host restart is required.

Updating alone leaves existing per-chat memory in use. As in 0.2.0, a new client
session can resubmit unchanged documents in that layout; the zero-upsert reuse
checks apply to opted-in versioned branches. To enable reuse for a
selected story, connect, prepare memory, select a saved character chat and click
Use versioned memory for this story. Confirm and enable memory. Initial opt-in
indexes older local history into a new story collection, consuming managed
embedding/storage usage. Old per-chat data is preserved for explicit cleanup;
there is no automatic bulk migration or deletion.

For that story, enter an optional name and click Save / finish checkpoint.
Memory stays disabled afterward. Use Refresh story checkpoints and Resume
to create a new path, then enable memory. A pending save can use Finish / retry.
Commit waiting may take minutes; an upsert acknowledgement alone does not mean a
snapshot contains the write. See versioned memory,
checkpoint recovery and integrity and
commit confirmation.

Cleanup, backup and rollback

Delete this chat's remote memory removes its branch in an opted-in story;
other paths/checkpoints remain. Delete all owned remote memory removes owned
story families, checkpoint branches and old per-chat/shared collections. Native
chat files remain, but their old checkpoints can no longer resume without their
remote branches. Native chat deletion and extension uninstall do not automatically
remove remote memory. Removing a branch is not proof of physical snapshot erasure.

A checkpoint includes chat transcript/swipes/metadata, not character cards,
World Info, files, model configuration or external state. Native host files and
the installation's owner identity are required; LambdaDB alone is not a complete
backup or cross-device setup. Keep a native host-data backup.

Disable memory before code rollback. Version 0.2.0 cannot manage smstory_*
collections or checkpoint controls and uses its former per-chat layout. Use
0.3.0 cleanup before a permanent downgrade if remote removal is wanted; do not
delete remote memory merely to switch versions. Pinning v0.2.0 changes code,
not transcripts, metadata or remote writes. The upgrade test checks code rollback
after test-data cleanup, not reversal of live versioned metadata.

Evidence and limits

See the 0.3.0 validation record for installation acceptance
and exact producer boundaries. Historical lifecycle/recovery and 1,000-message
reuse evidence remains linked in the feature documents. No new answer-quality
evaluation is included; the historical results remain
qualified by their original datasets and producers.

One active writer is required. Concurrent devices, other extensions editing the
same native files, and out-of-band remote writes are unsupported. Local-file and
remote-branch operations are not a distributed transaction. Long-duration
snapshot/expiry correctness is assumed from LambdaDB, not newly measured here.
A retained checkpoint branch must still exist and remain unchanged for resume.

Licensed under AGPL-3.0-only. Experimental status is not a stable-support
commitment.

SillyMemory 0.2.0 (experimental)

SillyMemory 0.2.0 (experimental)

Long-term memory for SillyTavern. Powered by LambdaDB.

Release date: 2026-10-03. Install from https://github.com/lambdadb/SillyMemory
through Extensions → Install extension, leaving the branch/tag field empty
to follow main. See the installation guide.

Changes

• Memory requests go directly from the browser to LambdaDB over HTTPS/CORS.
  No host proxy configuration or restart is required. Cookies and host CSRF
  headers are omitted, and the API key remains session-only.
• Each character chat and native branch gets its own owned collection. Chat
  renames retain identity; branches and duplicates index independently.
  Current-chat deletion preserves other chats; all-owned cleanup also discovers
  this installation's tagged collections after browser bookkeeping loss.
• Long-message chunks prefer paragraph, sentence and word boundaries within
  800 Unicode code points, with exact source offsets and no overlap. New layout
  IDs allow reindexing without accepting obsolete chunks as current evidence.
• Recalled passages keep user/assistant roles, speaker names and literal source
  text. Retrieval considers the latest anchor and a separate contextual turn,
  including assistant context and explicit Continue generation.
• Identical selected excerpts share one body with all selected source positions,
  leaving room for distinct context. Defaults remain 12 recent messages and an
  adjustable 800-token memory budget. Vector retrieval remains the default;
  the evaluated hybrid candidate was rejected.
• Memory in the last prompt distinguishes prepared memory from memory
  verified in the final host prompt. Stop on missing context defaults on and
  cancels generation before the model request when verifiable context is lost.
  Adjust context/output settings and retry, or disable the option to proceed
  with a warning. The submitted user message remains in the chat.
• Final-prompt verification requires complete message/line boundaries, so a
  missing short turn such as OK cannot be verified by BOOK A TRIP. Known
  speaker prefixes and line-separated injections remain supported.
• A pending final-prompt check cannot be overwritten by another generation.
  If another extension prevents its final event, reload before trying again.

Update and remote data

Use Manage extensions → SillyMemory →
Update, reload, enter the LambdaDB key again and re-enable memory. Ownership,
recent-message settings and memory budget remain intact. Existing installations
receive the default missing-context stop behavior. No separate embedding-provider
key or integration is required: embeddings remain managed by LambdaDB.

The first enabled sync rebuilds each opened chat into a separate collection
from its local SillyTavern history. It does not move or delete the old 0.1.0
shared collection. That collection's cleanup pointer is retained. New indexing
consumes managed embedding usage and storage; old and new copies may coexist.
Boundary-aware layout changes also reindex affected chunks in existing development
chat collections. Allow synchronization to finish before evaluating retrieval.

Delete this chat’s remote memory removes only that chat's collection.
Delete all owned remote memory removes all this installation's chat/branch
collections and its previous shared memory. It is not a legacy-only cleanup
button. Local chats remain and can rebuild memory after preparation and enabling.
Back up SillyTavern data before updates or cleanup. Native chat deletion and
extension uninstall do not automatically remove remote collections; use the
appropriate memory deletion control first when remote removal is intended.

Rollback

Disable memory first if an update causes problems. Advanced users may pin the
published v0.1.0 tag following RELEASING.md. Code rollback
does not undo remote writes or chat metadata changes. Version 0.1.0 does not
understand the new per-chat collections; use 0.2.0 cleanup before a permanent
downgrade/uninstall if those collections should be removed. Do not delete remote
memory merely to switch code versions. The installation test checks code rollback
after remote cleanup, not reversal of a live data migration.

Evidence and remaining limits

The release validation record separates Git URL/UI upgrade,
real managed-memory integration, and retained emulator evidence. These checks
establish installation and lifecycle behavior, not answer-quality improvement.

The historical 42-question English evaluation
scored 24/42 with SillyMemory at 32K, 14/42 with plain 32K and 29/42 with full
history at 128K. Median answer-input tokens were 3,282, 31,403 and 104,933.5,
respectively. This supports a token/accuracy tradeoff on that consumed split,
not equal full-history quality. It predates the collection/chunking changes.
The earlier 64-answer synthetic run scored 30/32
memory-on and 4/32 off with provisional grading; its two Korean quotation misses
remain documented. Neither result is a fresh evaluation of this release tree.
Boundary-aware chunking improved source-boundary preservation in a small pilot;
it did not demonstrate an answer-accuracy gain. No hybrid-search benefit is claimed.

Only SillyTavern 1.19.0 is validated. LambdaDB must allow the browser page
origin; this version needs no SillyTavern proxy setting or restart. See the
direct-CORS validation for the new transport boundary.
API keys stay in browser memory and must be re-entered after reload. Use one
active tab per browser profile. Group chats, files/tools and other prompt-rewriting
memory extensions remain outside support. Cross-device setup and long-duration
operation are not newly validated. Browser journal loss can leave obsolete remote
records; local-source validation rejects them, but complete cleanup can require
recreating the owned collection. Ownership recovery requires the installation's
owner identity; a remote collection alone is not portable chat state.

Licensed under AGPL-3.0-only. This remains an experimental pre-release,
not a stable-support commitment.

SillyMemory 0.1.0 (experimental)

SillyMemory 0.1.0 (experimental)

Long-term memory for SillyTavern. Powered by LambdaDB.

SillyMemory keeps recent messages intact and retrieves relevant passages from
older conversation under a memory token budget. This first experimental release
supports individual character chats, including synchronization after generation,
edits, swipes, deletion and reload, with separate character/chat/branch scopes.
The settings panel shows injected memory and synchronization progress, and offers
controls to disable memory or delete the installation's owned remote memory.

Install and connect

1. Use SillyTavern 1.19.0. Other host versions are unverified.
2. In Extensions → Install extension, enter
   https://github.com/lambdadb/SillyMemory and leave the branch/tag field empty
   to follow main. No extension build step or server plugin is needed.
3. Set enableCorsProxy: true in SillyTavern's config.yaml and restart the host.
4. Open the SillyMemory settings, enter your LambdaDB endpoint, project and API
   key, and run the synthetic upsert/query/delete test. After it passes, create
   the memory collection and enable memory in a character chat.

See the installation guide
for requirements and detailed setup. Keys are held only in browser memory;
reload requires key re-entry and re-enabling memory. Source passages and search
queries are sent through the host proxy to LambdaDB and its managed embedding
provider. LambdaDB and embedding usage incur costs.

Limits and data cleanup

Use one active extension tab per account/browser profile. Group chats, files,
tool messages, Data Bank, World Info and other prompt-rewriting memory extensions
are outside this release's scope. Disable built-in Vector Storage chat
vectorization when using SillyMemory. Cross-device concurrent editing and
long-duration reliability are unverified.

Deleting or renaming chats, clearing browser storage, or uninstalling the
extension does not clean up every remote scope. Use Delete all owned remote
memory and clear any pending test collection before uninstalling. Local chats
remain intact. Remote deletion does not establish erasure from provider backups.

Validation

Recorded checks include 68 unit tests on Node 20/24, 17 real Git URL installation
and candidate-update assertions, and 188 real-host/emulator checks including two
controlled host-crash recoveries. A held-out synthetic evaluation retrieved the
target and returned its expected code in 12/12 memory-on cases, with 11/12 strict
code-only answers. These are bounded synthetic results, not a general recall or
availability guarantee. Installation checks cover unreleased commits, not an
upgrade between published versions.

See the validation record
for exact source identities and remaining checks. Updates are manual through
SillyTavern's extension manager. Disable memory first if a problem occurs.

Licensed under AGPL-3.0-only.