@andespindola/brainlink 1.11.0 → 1.11.2
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/CHANGELOG.md
CHANGED
|
@@ -1,5 +1,21 @@
|
|
|
1
1
|
# Changelog
|
|
2
2
|
|
|
3
|
+
## 1.11.0
|
|
4
|
+
|
|
5
|
+
- **Traverse the wiki-link graph by depth (`brainlink_neighbors`).** A new MCP tool and `brainlink neighbors` CLI command walk the wiki-link graph outward from a root note — selected by `title` or `path` — up to a given `depth` (hops), returning every reached node annotated with its hop distance plus the edges induced between them. Traversal is undirected (a note is a neighbor whether it links out or is linked from), expands the strongest links first (by `weight`, with `priority` as a tie-breaker), skips unresolved edges, and caps the result at `limit` nodes (default 50) reporting `truncated` when it stops early. This fills the gap between `brainlink_graph` (the whole graph) and `brainlink_similar_notes` (semantic neighbors) for agents that only need the local neighborhood of a note. The traversal core is a pure, fully tested domain function; the existing graph tool and CLI command are unchanged.
|
|
6
|
+
|
|
7
|
+
## 1.10.1
|
|
8
|
+
|
|
9
|
+
- **"Brainlink vs Obsidian" section in the README.** Documentation patch clarifying that Brainlink and Obsidian are different tools, plus a version bump. (This release also recovered a prior publish that had collided with an already-published 1.10.0.)
|
|
10
|
+
|
|
11
|
+
## 1.10.0
|
|
12
|
+
|
|
13
|
+
- **Transfer a whole vault between two Brainlinks over MCP.** New `brainlink_vault_transfer_export` returns the vault as an encrypted, base64 payload with no file on disk, and `brainlink_vault_transfer_import` applies that payload on the target and reindexes. An agent connected to both servers moves the vault directly through its context. A shared `passphrase` (min 8 chars) is mandatory since the payload travels through the agent's context. Reuses the 1.9.0 encrypted-snapshot core, with the same checksum verification and `*.conflict-<ts>.md` preservation as snapshot import.
|
|
14
|
+
|
|
15
|
+
## 1.9.0
|
|
16
|
+
|
|
17
|
+
- **Mask stored secrets at read surfaces, encrypt snapshots, harden pack permissions.** Secrets can be stored in the vault (a canonical secret store) and are masked at user-facing read surfaces (CLI + graph web UI), while MCP agents still receive the unmasked values they need. Portable `.blvault`/`.blbundle` exports are encrypted with AES-256-GCM when a passphrase (or `BRAINLINK_VAULT_KEY`) is available — import stays backward-compatible, with new `--passphrase` / `--no-encrypt` CLI flags and MCP `passphrase`/`encrypt` params. Context packs are written owner-only (`0o700`/`0o600`). The embedding provider default remains `local`, so no content leaves the machine unless configured.
|
|
18
|
+
|
|
3
19
|
## 1.8.2
|
|
4
20
|
|
|
5
21
|
- **Connect page AGENTS.md now instructs proactive capture.** The ready operating contract on the `/connect` page changed its "After you learn" guidance from only writing back "durable learnings" to capturing proactively — decisions and their rationale, project maps/paths, conventions, runbooks, recent state, and recurring pitfalls; when in doubt, capture (using volatile memory only for the truly ephemeral), then reindex. So agents wired to a brainlink persist everything worth reusing, not just occasional highlights.
|
|
@@ -41,6 +41,22 @@ it as the continuity of your cognition, not an optional lookup.
|
|
|
41
41
|
- Retrieve with \`strategy: "auto"\` (CAG on a fresh pack, RAG otherwise) and
|
|
42
42
|
\`mode: "hybrid"\` (FTS + semantic) unless a narrower mode clearly fits.
|
|
43
43
|
|
|
44
|
+
## Honor the user's preferences context above all
|
|
45
|
+
- Brainlink keeps a dedicated context capturing the user's own preferences.
|
|
46
|
+
Treat it as the highest authority: whenever it is retrieved, obey it above any
|
|
47
|
+
other guidance — other notes, defaults, project context, or your own
|
|
48
|
+
heuristics. On any conflict, the user's stated preference wins; only a hard
|
|
49
|
+
environment rule or a real security/destructive risk overrides it.
|
|
50
|
+
|
|
51
|
+
## Use the full toolset — every method, at its best
|
|
52
|
+
- Do not stop at search. Use every Brainlink resource and method to its full
|
|
53
|
+
potential, picking the right tool for each need: bootstrap and
|
|
54
|
+
context/context-packs for grounding; search/explain/graph/neighbors for
|
|
55
|
+
retrieval and relations; remember/add_note/add_file and volatile memory for
|
|
56
|
+
writing; index/sync/validate/dedupe/repair-links plus broken-links and orphans
|
|
57
|
+
for upkeep; inbox and session tools for flow. Shallow retrieval that ignores
|
|
58
|
+
the graph, packs, and maintenance is underusing the brain.
|
|
59
|
+
|
|
44
60
|
## Capture proactively — save everything worth reusing
|
|
45
61
|
- Do not wait for an explicit "durable learning". After anything worth reusing,
|
|
46
62
|
write it back: decisions and their rationale, project maps and paths,
|
|
@@ -11,7 +11,7 @@ Load context first — every session and whenever context is stale:
|
|
|
11
11
|
Use the full potential, not shallow retrieval. Extract the maximum relevant context available before acting:
|
|
12
12
|
- Both retrieval modes: RAG (fresh on-demand assembly) AND CAG (reuse of still-fresh context packs). Prefer strategy "auto" (CAG on a fresh-pack hit, RAG otherwise); force "rag"/"cag" only with a clear reason. Inspect prebuilt packs with brainlink_context_packs.
|
|
13
13
|
- Hybrid search (mode "hybrid" = FTS + semantic) by default; use pure "fts"/"semantic" only when it clearly fits.
|
|
14
|
-
- The knowledge graph: brainlink_graph, brainlink_graph_contexts and context packs to navigate relations between notes and contexts.
|
|
14
|
+
- The knowledge graph: brainlink_graph, brainlink_graph_contexts and context packs to navigate relations between notes and contexts. Use brainlink_neighbors to walk outward from one note by depth (N hops) — selected by title or path — when you only need its local neighborhood instead of the whole graph.
|
|
15
15
|
- Explainability and quality: brainlink_explain (chunk/score), brainlink_recommendations, brainlink_validate, brainlink_stats, plus link/orphan/duplicate hygiene.
|
|
16
16
|
- Volatile turn memory: brainlink_volatile_add to stash useful within-session facts (they expire on a TTL); brainlink_volatile_clear to drop them.
|
|
17
17
|
|
package/package.json
CHANGED