v0.7.0

chore: bump to v0.7.0 — warm sync

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

v0.6.0

v0.6.0: Write pipeline with sqlite-vec migration (#7)

* feat: add vecstore module with sqlite-vec integration

Wrap sqlite-vec for vector search, replacing HNSW-based approach.
Provides init, insert, delete, search (with tombstone filtering),
and clear operations on a vec0 virtual table. Includes 5 unit tests.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* feat: integrate sqlite-vec into Store with transaction helpers

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* refactor: replace HNSW semantic lane with sqlite-vec in search

All search code paths now use store.search_vec() instead of
HnswIndex::search(). The hnsw module remains but is unused — deletion
is deferred to Task 5.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* refactor: replace HNSW rebuild with sqlite-vec inserts in indexer

- Remove HnswIndex import and HNSW rebuild steps (11-12)
- Insert vectors into vec0 table during chunk write loop
- Delete from vec0 when files are deleted or changed
- Clear vec0 on full rebuild
- Use store.next_vector_id() instead of scanning all vectors
- Add folder centroid computation and storage after indexing
- Add folder_centroids table migration and upsert/get methods in Store

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* refactor: remove hnsw_rs dependency, delete hnsw.rs — vectors now in sqlite-vec

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* feat: auto-migrate existing BLOB vectors to sqlite-vec on startup

Adds `migrate_vectors_to_vec0()` which copies BLOB vectors from
`chunks.vector` into the `chunks_vec` vec0 virtual table. Called from
`init()` after `init_vec_table()` so the virtual table is guaranteed
to exist. No-ops when vec0 is already populated or no BLOBs are present.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* feat: add tag registry with fuzzy resolution

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* feat: add link discovery module for auto-wikilinks

Scans note content for potential wikilink targets using exact filename
and alias matching. Supports case-insensitive search, word boundary
checking, existing wikilink skipping, and longest-match-first priority.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* feat: add folder placement module with type rules and semantic centroids

Three-strategy cascade: type-based rules (person/daily/workout + content
pattern detection) → semantic centroid matching against precomputed folder
embeddings → inbox fallback. 12 tests covering all strategies.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* feat: add write pipeline orchestrator with create, append, update, and move

Implements the writer module that ties together content analysis, tag
resolution, link discovery, folder placement, and atomic write+index.

- CreateNoteInput: 5-step pipeline (filename, tags, links, placement, write)
- AppendInput: append content with mtime conflict detection
- UpdateMetadataInput: frontmatter-only updates without re-chunking
- move_note: relocate files with store record updates
- All writes use temp+rename for atomicity with transaction rollback
- Pre-computes embeddings before holding DB lock
- Adds Store::resolve_file() for path/basename/#docid resolution
- Adds time crate for date formatting

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix: resolve clippy warnings in writer, links, and placement modules

* feat: add create, append, update_metadata, and move_note MCP write tools

Extends the MCP server with 4 write tools that expose the writer module
pipeline to Claude Code clients, completing the read-write tool surface.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* feat: add write CLI subcommands (create, append)

Adds `engraph write create` and `engraph write append` subcommands backed
by the writer module pipeline. Both support --content flag or stdin for
content input, with --json output mode.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* feat: add crash recovery — cleanup orphan .tmp files on startup

Scans the vault for leftover `.md.tmp` files on both `engraph index` and
`engraph serve` startup, removing any that survived a previous crash mid-write.
Logs the count if any are removed.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* test: add write pipeline integration tests

Three #[ignore] tests covering create_note searchability, append index
update, and mtime conflict detection. Run with:
  cargo test --test write_pipeline -- --ignored

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* chore: v0.6.0 — write pipeline, sqlite-vec migration, tombstone removal

Remove redundant tombstone writes from indexer (delete_vec handles it).
Replace tombstone loading in search with empty set. Fix clippy warning
in writer.rs. Apply cargo fmt across all modules. Bump version to 0.6.0.
Update CLAUDE.md with 19 modules, 190 tests, write pipeline docs, and
sqlite-vec architecture.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix: update stored mtime after rename to prevent false conflict detection

* feat: add archive/unarchive for soft-delete with index exclusion

- archive: moves note to 04-Archive/, adds archived frontmatter, removes from index
- unarchive: restores to original location (via archived_from), re-indexes
- indexer auto-excludes archive folder during walks
- MCP tools: archive, unarchive (13 total tools now)
- CLI: engraph write archive/unarchive

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* style: apply cargo fmt to archive/unarchive code

* feat: complete v0.6 spec coverage — content analysis, suggested_folder, incremental centroids, orphan cleanup, tag queries

- Gap 1: Add suggestion field to PlacementResult; add ticket ID detection
  (BRE-XXXX/DRIFT-XXX), meeting note detection, decision type_hint
- Gap 2: Inject suggested_folder frontmatter when semantic placement finds
  a below-threshold match during inbox fallback
- Gap 3: Incrementally update folder centroids after each note creation
  (weighted merge with existing centroid)
- Gap 4: Add verify_index_integrity() to clean orphan DB entries for files
  that no longer exist on disk; called on index and serve startup
- Gap 5: Add agent_created_tags(), low_usage_tags(), stale_tags() queries
  to store for tag hygiene tooling

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

v0.5.0 — MCP Server

What changed

engraph is now an MCP server. AI agents (Claude Code, etc.) connect via stdio and get direct access to your vault's intelligence — search, context bundles, people lookups, project overviews.

New command

engraph serve    # Start MCP stdio server

MCP Tools

Tool	What it does
search	Hybrid 3-lane search (semantic + FTS5 + graph)
read	Full note content with metadata and graph edges
list	Filter notes by folder and/or tags
vault_map	Vault structure overview — use this to orient in a new session
who	Person context: note + mentions + connections
project	Project context: note + children + tasks + team
context	Rich topic context with character budget trimming

Setup for Claude Code

1. Install: brew install devwhodevs/tap/engraph
2. Index your vault: engraph index ~/path/to/vault
3. Add to ~/.claude/settings.json:

{
  "mcpServers": {
    "engraph": {
      "command": "engraph",
      "args": ["serve"]
    }
  }
}

4. Restart Claude Code — engraph tools are now available

Stats

• 14 modules (was 13), serve.rs added
• 146 unit tests
• New deps: rmcp (official Rust MCP SDK), tokio, schemars

What's next (v0.6 — Write Pipeline)

• Create notes with auto-filing, tag resolution, link discovery
• Append to existing notes
• Write tools exposed via MCP

v0.4.0 — Context Engine

What changed

engraph v0.3 had hybrid search with a vault graph. v0.4 adds a context engine — six functions that assemble agent-ready context bundles from search results, graph connections, and vault metadata.

New commands

engraph context read <file_or_docid>              # Full note content with metadata + graph edges
engraph context list [--folder X] [--tags a,b]    # Filter notes by metadata
engraph context vault-map                          # Vault structure overview
engraph context who <name>                         # Person context bundle
engraph context project <name>                     # Project context bundle
engraph context topic <query> [--budget 32000]     # Rich topic context with budget trimming

All commands support --json for structured output.

Context engine highlights

• who "John Nelson" — returns the person's note, all vault mentions with snippets, wikilink connections
• project "Drift" — returns the project note, child notes, active tasks (unchecked checkboxes), team members, recent daily mentions
• topic "delivery date" --budget 8000 — runs hybrid search, reads full note content from disk, expands via graph, assembles within character budget with truncation markers
• vault-map — folder note counts, top tags by frequency, recently indexed files — everything an agent needs to orient in a new session

Internal improvements

• search_internal() extracted from run_search() — reusable 3-lane RRF search for programmatic consumers
• Five new Store query methods for filtering, aggregation, and listing
• ContextParams shared context struct for all functions

Upgrading from v0.3

No migration needed — context engine is read-only and works with existing databases. Run engraph index if you haven't recently.

Stats

• 13 modules (was 12), context.rs added (~1,050 lines)
• 144 unit tests (was 119)
• ~1,600 lines of new Rust code

What's next (v0.5 — MCP Server)

• MCP stdio transport so Claude Code can connect directly
• Context engine tools exposed as MCP tools
• HTTP daemon for REST API consumers

v0.3.0 — Vault Graph & Graph Search Agent

What changed

engraph v0.2 had two search lanes (semantic + keyword). v0.3 adds a vault graph and a graph search agent as the third lane — notes connected via wikilinks now surface in search results even when they don't match the query directly.

New features

Vault graph

The indexer now builds a graph from your vault's wikilinks and people mentions:

• Wikilink edges — bidirectional, extracted from [[...]] patterns during indexing
• Mention edges — auto-detected when a file mentions a person from your People folder (supports aliases from frontmatter)
• Stored in SQLite edges table with ON DELETE CASCADE

Graph search agent (3rd RRF lane)

Search now runs three lanes in parallel:

1. Semantic — HNSW vector similarity
2. Keyword — FTS5 BM25 exact match
3. Graph — expands top results by following wikilinks 1-2 hops

The graph agent finds related notes that semantic and keyword search miss. Decay scoring (0.8x for 1-hop, 0.5x for 2-hop) ensures directly relevant notes rank higher. Relevance filtering prevents tangential results.

engraph graph CLI

engraph graph show <file_or_docid>   # Show a note's connections
engraph graph stats                   # Vault graph overview

Example output:

BRE-2579 - Order Sync Metafields.md (#831e2a)

Outgoing wikilinks (6):
  → John Nelson.md (#bf84f0)
  → BRE-1728 - Order Sync Delivery Date.md (#05ac0c)
  ...

Mentioned by (4):
  ← 2026-03-24.md (#da8b4a)
  ...

Status output

Now includes edge counts:

Edges:      623 (546 wikilinks, 77 mentions)

Upgrading from v0.2

Databases are automatically migrated (new edges table created on first open). Run a rebuild to populate the graph:

engraph index --rebuild /path/to/vault

For people detection, add your People folder to ~/.engraph/vault.toml:

[structure.folders]
people = "People"

Stats

• 12 modules (was 11), graph.rs added
• 119 unit tests (was 91)
• ~1,400 lines of new Rust code
• Live vault test: 84 files, 623 edges (546 wikilinks, 77 mentions), 89% connected

What's next (v0.4 — Server & Context)

• MCP server (stdio transport)
• HTTP daemon
• Context engine (context(), who(), project(), token-budgeted bundles)

v0.2.0 — Hybrid Search & Vault Intelligence Foundation

What changed

engraph v0.1 was pure semantic search — embed everything, find similar vectors. It worked for conceptual queries ("notes about architecture patterns") but missed exact matches ("BRE-2579", "John Nelson", "2026-03-20").

v0.2 adds a second search lane (FTS5 keyword search) and fuses both with Reciprocal Rank Fusion. It also lays the groundwork for vault-aware intelligence with structure detection, docids, and a pluggable model layer.

New features

Hybrid search (Semantic + FTS5)

Search now runs two lanes in parallel:

• Semantic — HNSW vector search for conceptual similarity
• Keyword — SQLite FTS5 with BM25 ranking for exact matches

Results are merged via Reciprocal Rank Fusion (RRF). Items found by both lanes rank highest.

Smart chunking

Replaces the old ## -only splitter with a break-point scoring algorithm. Finds optimal split points at headings (score 50–100), code fences (80), thematic breaks (60), and blank lines (20). Code fence protection prevents splitting inside code blocks. 15% overlap for context continuity.

Docids

Every indexed file gets a deterministic 6-char hex ID (#ab12cd) shown in search results for quick reference.

Vault profiles

engraph init auto-detects your vault:

• Structure: PARA, Folders, or Flat
• Type: Obsidian, Logseq, or Plain markdown
• Stats: file count, wikilinks, frontmatter usage, unique tags, folder depth

Writes vault.toml for future configuration.

Pluggable model layer

ModelBackend trait enables future model swapping without changing consumer code. Model registry ships with known-good models. Manage via engraph models list and engraph models info.

--explain flag

See exactly how each result was ranked:

04-Archive/TLDR - Monday standup.md
  RRF: 0.0323
    semantic: rank #3, raw 0.48, +0.0159
    fts: rank #1, raw 5.20, +0.0164

New commands

engraph init [path]           # Auto-detect vault structure, write vault.toml
engraph configure             # Interactive setup (placeholder)
engraph models list           # List available embedding models
engraph models info <name>    # Show model details
engraph search --explain      # Per-lane RRF score breakdown

Upgrading from v0.1

Existing databases are automatically migrated (new docid column + FTS5 table). However, to fully populate the new features, run:

engraph index --rebuild /path/to/vault

Breaking changes

• Search output now includes #docid after file paths
• JSON output adds "docid" field to results
• Empty JSON results return [] instead of "No results found."

Stats

• 11 modules (was 7), 5 new: docid, fts, fusion, model, profile
• 91 unit tests (was 44)
• ~2,400 lines of new Rust

What's next (v0.3 — Context & Graph)

• Vault graph (wikilink edges, tag edges, people detection)
• Graph search agent (link-following from results)
• Context Engine with token-budgeted bundles

v0.1.6

ci: auto-update Homebrew formula on release

v0.1.5

fix: suppress noisy ort/hnsw_rs logs in production, clean up clippy

v0.1.4

fix: build native binaries instead of cross-compiling (ort-sys limita…

v0.1.2

style: cargo fmt