docs: rewrite README, add CHANGELOG, CONTRIBUTING, examples

- README rewritten as landing page: value prop, architecture,
  examples, comparison table, roadmap
- Added CHANGELOG.md with full release history
- Added CONTRIBUTING.md
- Added examples/claude-code-setup.md for MCP integration
- Removed internal planning docs from tracking (live in vault)
- Updated Cargo.toml with repository, keywords, categories
- Updated repo description and topics on GitHub

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

Author: devwhodevs

.gitignore

@@ -3,3 +3,11 @@
 *.swo
 .DS_Store
 .worktrees/
+
+# Internal planning docs (live in Obsidian vault, not in repo)
+docs/specs/
+docs/superpowers/
+
+# Positioning and visual asset TODOs (not part of the distributed package)
+PROJECT_POSITIONING.md
+VISUAL_ASSETS_TODO.md

CHANGELOG.md

@@ -0,0 +1,76 @@
+# Changelog
+
+## [0.7.0] - 2026-03-25
+
+### Added
+- **File watcher** — `engraph serve` now watches the vault for changes and re-indexes automatically (2s debounce)
+- **Placement correction learning** — detects when users move notes from suggested folders, updates centroids
+- **Fuzzy link matching** — sliding window Levenshtein matching (0.92 threshold) during note creation
+- **First-name matching** — matches "Steve" to `[[Steve Barbera]]` for People folder notes (suggestion-only)
+- `created_by` column and filter — track note origin, filter with `engraph context list --created-by`
+- `placement_corrections` table for observability
+- `link_skiplist` table schema (reserved for future use)
+
+### Changed
+- Centroid updates use true online mean (was EMA 0.9/0.1)
+- Indexer refactored: `index_file`, `remove_file`, `rename_file` extracted as public functions
+- Bulk indexing uses batched transactions for performance
+- `run_index_shared` variant accepts external store/embedder references
+
+### Fixed
+- Content hash consistency between `diff_vault` and `index_file` (BOM handling)
+
+## [0.6.0] - 2026-03-25
+
+### Added
+- **Write pipeline** — create, append, update_metadata, move, archive, unarchive notes
+- **sqlite-vec** replaces HNSW for vector search (single SQLite database)
+- **Tag registry** with fuzzy Levenshtein resolution
+- **Link discovery** — exact basename and alias matching during note creation
+- **Folder placement** — type rules, semantic centroids, inbox fallback
+- **Archive/unarchive** — soft delete with metadata preservation
+- 6 new MCP write tools (13 total)
+
+### Changed
+- All vectors stored in SQLite vec0 virtual table (was HNSW + separate files)
+- Atomic writes via temp file + rename for crash safety
+- Mtime-based conflict detection for concurrent edits
+
+## [0.5.0] - 2026-03-24
+
+### Added
+- **MCP server** — `engraph serve` starts stdio MCP server via rmcp SDK
+- 7 read-only MCP tools: search, read, list, vault_map, who, project, context
+
+## [0.4.0] - 2026-03-24
+
+### Added
+- **Context engine** — 6 functions: read, list, vault_map, who, project, topic
+- Token-budgeted context bundles for AI agents
+- Person and project context assembly from graph + search
+
+## [0.3.0] - 2026-03-24
+
+### Added
+- **Vault graph** — bidirectional wikilink + mention edges built during indexing
+- **Graph search agent** — 3rd RRF lane with 1-2 hop expansion
+- People detection from configured People folder
+
+## [0.2.0] - 2026-03-24
+
+### Added
+- **Hybrid search** — semantic (embeddings) + keyword (FTS5 BM25) fused via RRF
+- Smart chunking with break-point scoring algorithm
+- Docid system (6-char hex file IDs)
+- Vault profiles with auto-detection (`engraph init`)
+- Pluggable model layer (`ModelBackend` trait)
+- `--explain` flag for per-lane score breakdown
+
+## [0.1.0] - 2026-03-19
+
+### Added
+- Initial release
+- ONNX embedding model (all-MiniLM-L6-v2, 384-dim)
+- SQLite metadata storage
+- Incremental indexing
+- `.gitignore`-aware vault walking

CONTRIBUTING.md

@@ -0,0 +1,53 @@
+# Contributing to engraph
+
+Thanks for your interest in contributing to engraph.
+
+## Getting started
+
+```bash
+git clone https://github.com/devwhodevs/engraph
+cd engraph
+cargo test --lib    # 225 tests, no network required
+```
+
+## Before you start
+
+**Open an issue first.** For anything beyond a typo fix, please open an issue to discuss what you'd like to change. This avoids wasted effort if the change doesn't fit the project direction.
+
+## Development workflow
+
+1. Fork the repo and create a branch from `main`
+2. Write tests for any new functionality
+3. Run the full check suite:
+   ```bash
+   cargo fmt --check
+   cargo clippy -- -D warnings
+   cargo test --lib
+   ```
+4. Open a pull request with a clear description of what and why
+
+## Architecture
+
+The codebase is 20 Rust modules behind a lib crate. See `CLAUDE.md` for detailed architecture documentation — it's designed for AI-assisted development but serves as a thorough codebase guide for human contributors too.
+
+Key modules:
+- `store.rs` — SQLite persistence (all tables, queries, migrations)
+- `indexer.rs` — vault walking, chunking, embedding, index updates
+- `serve.rs` — MCP server with 13 tools
+- `watcher.rs` — file change detection and real-time re-indexing
+- `search.rs` — 3-lane hybrid search orchestration
+
+## What makes a good contribution
+
+- Bug fixes with tests
+- Performance improvements with benchmarks
+- New MCP tools that expose existing functionality
+- Documentation improvements
+- Platform support (Windows, other architectures)
+
+## Code style
+
+- Run `cargo fmt` before committing
+- No clippy warnings (`cargo clippy -- -D warnings`)
+- Prefer small, focused PRs over large changes
+- Tests are expected for new functionality

Cargo.toml

@@ -2,8 +2,13 @@
 name = "engraph"
 version = "0.7.0"
 edition = "2024"
-description = "Local semantic search for Obsidian vaults"
+description = "Local knowledge graph for AI agents. Hybrid search + MCP server for Obsidian vaults."
 license = "MIT"
+repository = "https://github.com/devwhodevs/engraph"
+homepage = "https://github.com/devwhodevs/engraph"
+readme = "README.md"
+keywords = ["obsidian", "semantic-search", "mcp", "rag", "knowledge-graph"]
+categories = ["command-line-utilities", "database", "text-processing"]
 
 [dependencies]
 clap = { version = "4", features = ["derive"] }