refactor(core): simplify v0.23 lifecycle and vector surfaces

CHANGELOG.md

@@ -7,16 +7,23 @@
 - **#997**: Added the `bm hook` harness front door (SPEC-55). Lifecycle verbs
   (`session-start`, `pre-compact`) move the plugin hook logic into the package
   behind per-harness stdin adapters, with the session brief fenced as reference
-  data. Opt-in envelope capture (`captureEvents: true`, fail-closed) records
-  redacted lifecycle events into a local inbox WAL, projected deterministically
-  by `bm hook flush`; `bm hook status` shows the surface. `bm hook install` /
+  data. Default-on envelope capture records bounded lifecycle metadata into a
+  local inbox WAL, and `bm hook flush` archives it locally; `bm hook status`
+  shows the surface. `captureEvents: false` disables capture. `bm hook install` /
   `bm hook remove` wire the hooks into user-level harness config for standalone
   users with ownership-tagged, surgical merging. The Claude Code and Codex
   plugin hooks are now zero-logic PEP 723 uv scripts (`uv run --script`) that
   invoke `basic-memory hook` in-process; their dependency floor is bumped by
   release tooling, and `BM_BIN` overrides the uv-managed environment for
   development.
 
+### Maintenance
+
+- Removed the hook-specific redaction subsystem, its `detect-secrets`
+  dependency, and the retired lifecycle projector compatibility module.
+- Kept Milvus as a first-party optional vector backend while removing the
+  unused Python entry-point registry for separately packaged vector adapters.
+
 ## v0.22.1 (2026-06-12)
 
 Follow-up patch to v0.22.0. Fixes project and default-project resolution on

docs/semantic-search.md

@@ -106,7 +106,7 @@ All settings are fields on `BasicMemoryConfig` and can be set via environment va
 | Config Field | Env Var | Default | Description |
 |---|---|---|---|
 | `semantic_search_enabled` | `BASIC_MEMORY_SEMANTIC_SEARCH_ENABLED` | Auto (`true` when semantic deps are available) | Enable semantic search. Required before vector/hybrid modes work. |
-| `semantic_vector_index` | `BASIC_MEMORY_SEMANTIC_VECTOR_INDEX` | `"pgvector"` | Postgres vector storage adapter. `"pgvector"` is built in, `"milvus"` is available through the `basic-memory[milvus]` extra, and other names resolve through the `basic_memory.semantic_vector_indexes` Python entry-point group. SQLite always uses its built-in `sqlite-vec` adapter. |
+| `semantic_vector_index` | `BASIC_MEMORY_SEMANTIC_VECTOR_INDEX` | `"pgvector"` | Postgres vector storage adapter: `"pgvector"` or first-party `"milvus"` through the `basic-memory[milvus]` extra. SQLite always uses its built-in `sqlite-vec` adapter. |
 | `milvus_uri` | `BASIC_MEMORY_MILVUS_URI` | Unset | Milvus, Milvus Lite, or Zilliz Cloud connection URI. Required when `semantic_vector_index="milvus"`. |
 | `milvus_token` | `BASIC_MEMORY_MILVUS_TOKEN` | Unset | Optional Milvus or Zilliz Cloud authentication token. |
 | `milvus_timeout_seconds` | `BASIC_MEMORY_MILVUS_TIMEOUT_SECONDS` | `30.0` | Finite per-operation timeout for Milvus and Zilliz client calls. Increase it for unusually slow deployments. |
@@ -473,7 +473,7 @@ bm reindex -p my-project
 - **Dimension change**: After changing `semantic_embedding_dimensions`
 - **LiteLLM role change**: After changing `semantic_embedding_document_input_type` or `semantic_embedding_query_input_type`
 - **Literal prefix change**: After changing `semantic_embedding_document_prefix` or `semantic_embedding_query_prefix`
-- **Vector index change**: After completing the external-adapter cleanup procedure below and changing `semantic_vector_index`
+- **Vector index change**: After completing the vector-store cleanup procedure below and changing `semantic_vector_index`
 
 The reindex command shows progress with embedded/skipped/error counts:
 
@@ -538,7 +538,7 @@ The sqlite-vec extension is loaded per-connection. Vector tables are created laz
 
 The Alembic migration creates the dimension-independent chunks table. The embeddings table and HNSW index are deferred to runtime because they depend on the configured vector dimensions.
 
-## Milvus and Pluggable Vector Indexes
+## Milvus Vector Index
 
 Postgres deployments can replace pgvector storage and nearest-neighbour lookup without
 replacing Basic Memory's SQL repositories or embedding providers. Milvus is the first-party
@@ -572,94 +572,30 @@ an existing collection uses another dimension or an incompatible schema, Basic M
 it and fails initialization instead of deleting shared vectors during a rolling deployment.
 Coordinate the exact collection migration, then run `bm reindex --embeddings`.
 
-Other provider names still resolve through the extension entry-point contract. A configured
-extension that is missing, duplicated, invalid, or returns an incompatible adapter fails
-explicitly at startup. Basic Memory does not silently fall back to pgvector, because doing so
-would split vectors across stores while appearing healthy.
-
 SQLite remains automatic in this version: local SQLite databases always select `sqlite-vec`, even
 if `semantic_vector_index` is set. The selector controls Postgres-backed runtimes only.
 
-### Extension Package Contract
-
-A separately distributed package registers one factory under the
-`basic_memory.semantic_vector_indexes` entry-point group:
-
-```toml
-[project.entry-points."basic_memory.semantic_vector_indexes"]
-qdrant = "acme_basic_memory_qdrant:create_index"
-```
-
-The factory receives an explicit scope and the validated Basic Memory configuration:
-
-```python
-from basic_memory.config import BasicMemoryConfig
-from basic_memory.repository.semantic_vector_index import (
-    SemanticVectorIndex,
-    VectorIndexScope,
-)
-
-
-def create_index(
-    *,
-    scope: VectorIndexScope,
-    app_config: BasicMemoryConfig,
-) -> SemanticVectorIndex:
-    ...
-```
-
-`VectorIndexScope` contains a stable, credential-free database namespace, project ID, embedding
-identity, and vector dimensions. Extensions must isolate physical storage by
-`scope.storage_key`, which contains only the stable database namespace and project ID.
-`embedding_identity` and `dimensions` describe the current vector schema for validation and
-initialization; adapters must not use those mutable fields to create a second unreachable
-project collection when the embedding configuration changes. Extensions own their client
-lifecycle, credentials, collection/index creation, vector persistence, and nearest-neighbour
-implementation.
-
-The returned `SemanticVectorIndex` has five asynchronous operations:
-
-- `initialize()` validates or creates backend storage.
-- `upsert(records)` idempotently writes vectors by `(entity_id, chunk_key)` for each record's
-  `source_hash` generation.
-- `delete(records)` removes stable keys only for each record's `source_hash` generation; stale or
-  missing records are successful no-ops.
-- `delete_entity(entity_id)` removes all vectors for one entity in the scope.
-- `search(query, limit)` returns stable keys with normalized cosine similarity in `[0, 1]`.
-
-The adapter never receives a SQLAlchemy session and never calls the embedding provider. Basic
-Memory owns chunking and embedding, while the extension owns vector storage and lookup.
-The built-in pgvector and sqlite-vec adapters additionally remove each pending SQL manifest row in
-the same database transaction as its vector so no newer generation can enter between those steps.
-Each `VectorRecord` and `VectorDeletion` carries the SHA-256 source generation that produced its
-value. Basic Memory holds the matching SQL manifest locked across extension adapter I/O and the
-ready-state transition or deletion (`FOR UPDATE` on Postgres and a conditional write lock on
-SQLite), so an older overlapping sync cannot overwrite or remove a newer generation under the same
-stable adapter key.
-
-Adapters may also implement the separate `SemanticVectorIndexReconciler` capability. After a
-vector reindex, Basic Memory passes it the complete set of current ready keys so the adapter can
-delete scoped external orphans. Keeping reconciliation separate preserves the narrow required
-storage protocol while allowing external stores to reclaim records left by interrupted deletes or
-ready-state commits.
+Milvus implements Basic Memory's internal `SemanticVectorIndex` storage boundary. Basic Memory
+continues to own chunking, embeddings, and the SQL manifest; Milvus owns only vector persistence,
+nearest-neighbour lookup, and scoped orphan cleanup.
 
 ### SQL Manifest and Failure Recovery
 
 `search_vector_chunks` remains the authoritative manifest even when vectors live in an external
 store. Each row records the selected `vector_index`, embedding identity, stable chunk key, and an
 `embedding_status` of `pending` or `ready`.
 
-Writes and deletes commit `pending` before calling the adapter. A successful adapter operation then
+Writes and deletes commit `pending` before calling the vector index. A successful operation then
 makes the manifest row ready or removes it. Vector writes are generation checked inside built-in
-adapter transactions; extension writes retain the manifest lock across adapter I/O. If the external
+SQL adapter transactions; Milvus writes retain the manifest lock across client I/O. If the external
 operation fails, the pending row is not searchable and the next sync safely retries the idempotent
 operation. Adapter search results are hydrated only through current, ready manifest rows, so stale
 or orphaned external matches fail closed.
 
-Basic Memory deliberately refuses to mutate manifest rows owned by a different external adapter.
-Before switching `semantic_vector_index`, keep the old adapter configured and use that extension's
-project-scope administrative cleanup to remove the old vectors. Remove the corresponding
+Basic Memory deliberately refuses to mutate manifest rows owned by a different vector index.
+Before switching `semantic_vector_index`, keep the old index configured and remove its
+project-scoped vectors. Remove the corresponding
 `search_vector_chunks` manifest rows only after the external cleanup succeeds. Then switch the
 configured adapter and run `bm reindex --embeddings` to populate the new store. If configuration
-was switched too early, restore the old adapter first; the ownership check will continue to fail
+was switched too early, restore the old index first; the ownership check will continue to fail
 closed until cleanup is completed.

plugins/claude-code/CHANGELOG.md

@@ -98,17 +98,20 @@ Memory's durable graph**, rather than a memory layer of its own. See
   by release tooling) and the script invokes
   `basic-memory hook <event> --harness claude` in-process with the hook JSON
   on stdin. `BM_BIN` overrides the uv-managed environment for development.
-  The brief/checkpoint logic lives in the released package; opt-in
-  `captureEvents: true` additionally records redacted event envelopes to a
-  local inbox. uv is the required prerequisite; the first run fetches from
-  PyPI, later runs use uv's cache.
+  The brief/checkpoint logic lives in the released package; lifecycle envelopes
+  containing bounded metadata are captured by default and can be disabled with
+  `captureEvents: false`. uv is the required prerequisite; the first run fetches
+  from PyPI, later runs use uv's cache.
 - **SessionStart hook now nudges toward `/basic-memory:bm-setup` on first run** — when
   no `basicMemory` config block is present in either settings file. The nudge
   survives a failed/empty task query (so a brand-new user with no project yet still
   sees it), and stops once setup writes the config. (Phase 3)
 
 ### Removed (clean break)
 
+- Hook-specific secret scanning and the `redactKeys` / `redactPaths` settings.
+  Lifecycle envelopes now stay small by construction instead of carrying a
+  general-purpose redaction subsystem.
 - The six bundled skills (`placement`, `knowledge-capture`, `knowledge-organize`,
   `continue-conversation`, `research`, `edit-note`). Equivalent, framework-agnostic
   workflows live in the top-level [`skills/`](../../skills) package

plugins/claude-code/README.md

@@ -91,11 +91,11 @@ local checkout. Two disclosures:
 - **Network fetch on first run.** uv downloads `basic-memory` from PyPI at a
   pinned minimum version (bumped by release tooling); later runs use uv's
   cache.
-- **Event capture is opt-in and off by default.** Setting `captureEvents: true`
-  (the JSON boolean — strings never enable it) records redacted lifecycle-event
-  envelopes to a local inbox under your Basic Memory home. Inspect with
-  `basic-memory hook status`, archive locally with `basic-memory hook flush`.
-  The lifecycle trace never becomes a graph note.
+- **Event capture is on by default.** Setting `captureEvents: false` disables
+  lifecycle-event envelopes. Other values fail closed. Envelopes contain bounded
+  lifecycle metadata and live in a local inbox under your Basic Memory home.
+  Inspect with `basic-memory hook status`, archive locally with
+  `basic-memory hook flush`. The lifecycle trace never becomes a graph note.
 
 Every failure path exits 0 — the hooks stay invisible rather than disrupt a
 session.
@@ -151,9 +151,7 @@ settings (or select it via `/config`).
 | `preCompactCapture` | `extractive` | How checkpoints are produced |
 | `sessionProfile` | `general` | `coding` makes checkpoints schema-backed `coding_session` notes with required Git identity |
 | `repository` | _(none)_ | User-confirmed stable repository identifier (`owner/name`); required for the `coding` profile |
-| `captureEvents` | `false` | Opt-in: record redacted lifecycle-event envelopes to the local inbox (see `basic-memory hook status` / `flush`). Only the JSON boolean `true` enables it. |
-| `redactKeys` | `[]` | Additional payload keys to redact before an event enters the local inbox |
-| `redactPaths` | `[]` | Additional paths to redact from working-directory and path-bearing capture content |
+| `captureEvents` | `true` | Record bounded lifecycle-event envelopes to the local inbox (see `basic-memory hook status` / `flush`). Set the JSON boolean `false` to disable it; other values fail closed. |
 
 The plugin seeds schemas for notes the Claude integration writes directly:
 `decision`, `task`, and the session type relevant to the selected profile. A

plugins/claude-code/hooks/pre_compact.py

@@ -4,7 +4,7 @@
 # dependencies = ["basic-memory>=0.22.1"]
 # ///
 """PreCompact hook — the entire hook. All logic (settings resolution, the
-extractive checkpoint note, opt-in envelope capture) lives in the released
+extractive checkpoint note, lifecycle-envelope capture) lives in the released
 basic-memory package behind `basic-memory hook pre-compact`; this script is
 only the launcher, and uv resolves the dependency floor declared above.
 scripts/update_versions.py bumps that floor at release time.

plugins/claude-code/hooks/session_start.py

@@ -4,7 +4,7 @@
 # dependencies = ["basic-memory>=0.22.1"]
 # ///
 """SessionStart hook — the entire hook. All logic (settings resolution, the
-context brief, opt-in envelope capture) lives in the released basic-memory
+context brief, lifecycle-envelope capture) lives in the released basic-memory
 package behind `basic-memory hook session-start`; this script is only the
 launcher, and uv resolves the dependency floor declared above.
 scripts/update_versions.py bumps that floor at release time.

plugins/claude-code/settings.example.json

@@ -9,9 +9,7 @@
     "recallPrompt": "You have Basic Memory available for this project. Before answering recall questions (\"what did we decide\", \"where did we leave off\"), search the graph first — prefer structured filters (search_notes with type/status). When the user makes a material decision, capture it as a note with type: decision. Cite permalinks when referencing prior work.",
     "preCompactCapture": "extractive",
     "sessionProfile": "general",
-    "captureEvents": false,
-    "redactKeys": [],
-    "redactPaths": [],
+    "captureEvents": true,
     "placementConventions": null,
     "teamProjects": {
       "my-team/notes": { "promoteFolder": "shared" }

plugins/claude-code/skills/bm-setup/SKILL.md

@@ -111,15 +111,12 @@ Ask only what you can't infer. Cover:
    task schemas, so I can find them precisely later — okay?" (See "Seed the
    schemas" below.)
 
-6. **Lifecycle-event capture.** "Should I also keep a local, redacted trail of
-   SessionStart and PreCompact events for diagnostics?" Default to **off**.
-   Explain that the normal session brief and PreCompact checkpoint work either
-   way; enabling this adds envelopes to a local inbox until `bm hook flush`
-   archives them locally. It never creates knowledge-graph notes. Only the JSON
-   boolean `true` enables capture.
-   - If enabled, optionally ask for repo-specific `redactKeys` (additional payload
-     keys) and `redactPaths` (working directories or path-bearing content). The
-     built-in redaction floor still applies when these lists are empty.
+6. **Lifecycle-event capture.** "The plugin keeps a local trail of SessionStart
+   and PreCompact events for diagnostics by default. Keep that on?" Default to
+   **on**. Explain that the normal session brief and PreCompact checkpoint work
+   either way; capture adds bounded lifecycle metadata to a local inbox until
+   `bm hook flush` archives it locally. It never creates knowledge-graph notes.
+   Persist the JSON boolean `false` only when the user opts out.
    - Capture stays local and personal. It never writes directly to team projects.
 
 7. **How active should I be? (output style)** "Want me to proactively capture —
@@ -210,9 +207,7 @@ Build the `basicMemory` block from the interview:
     "preCompactCapture": "extractive",
     "sessionProfile": "coding",
     "repository": "owner/name",
-    "captureEvents": false,
-    "redactKeys": [],
-    "redactPaths": [],
+    "captureEvents": true,
     "placementConventions": "<learned or suggested summary, or null>",
     "teamProjects": {}
   },
@@ -229,9 +224,7 @@ Only include `outputStyle` if the user opted in. Ask whether this is a **team
 default** (write/merge into `.claude/settings.json`, suggest committing it) or
 **personal** (`.claude/settings.local.json`). **Merge** into any existing file —
 read it, add/replace only the keys above, preserve everything else. Use compact,
-valid JSON. Always persist `captureEvents` as a JSON boolean. Empty `redactKeys`
-and `redactPaths` lists may be omitted; when present, they must be JSON arrays of
-strings.
+valid JSON. Always persist `captureEvents` as a JSON boolean.
 
 Writing the `basicMemory` block is also what stops the SessionStart hook's first-run
 nudge — the config's presence is the signal that setup has run.
@@ -264,8 +257,8 @@ ref before closing — don't let the next session's brief come up empty.
 Confirm what you did in a few lines: the project mapping, the session profile
 (and confirmed repository for a coding setup), which schemas were seeded vs.
 already present, whether placement was learned or suggested, the smoke-test
-result, whether lifecycle-event capture is enabled, any extra redaction controls,
-the shared hook inbox/flush state, and whether the output style is on.
+result, whether lifecycle-event capture is enabled, the shared hook inbox/flush
+state, and whether the output style is on.
 
 Then handle activation based on the output style:
 - **Output style enabled** → it's fixed at session start, so the full capture

plugins/claude-code/skills/bm-status/SKILL.md

@@ -26,7 +26,7 @@ This is a quick diagnostic — gather the facts and lay them out; don't over-inv
      (default `sessions`), `rememberFolder` (default `bm-remember`),
      `preCompactCapture` mode (default `extractive`), `sessionProfile` (default
      `general`), `repository` (coding profile only), `captureEvents` (default
-     `false`), `redactKeys`, and `redactPaths`.
+     `true`).
    - From the **root** settings object (not `basicMemory`): whether `outputStyle` is
      `basic-memory` — i.e. whether the capture reflexes are on.
 
@@ -73,8 +73,6 @@ you couldn't determine, rather than failing the whole report):
 - Session profile:   <general | coding>
 - Repository:        <owner/name or none>
 - Event capture:     <enabled | disabled>
-- Redact keys:       <configured count or none>
-- Redact paths:      <configured count or none>
 - Shared hook inbox: <path or unavailable>
 - Shared pending envelopes: <count or unavailable>
 - Shared archived envelopes: <count or unavailable>