Symmetric counterpart to file_upload: source is in the vault,
target is on the local filesystem.
path (source) accepts a short link or vault-relative path;
resolution goes through path_resolver.resolve_to_absolute.
Short-link ambiguity is surfaced as error="ambiguous" with the
candidate list — the download is **not** executed in that case.
local_path (target) is a plain filesystem path. **Optional** —
when empty, the file lands in a session-scoped temp dir (lazy,
auto-cleaned on process exit; each call gets its own subdirectory
so concurrent agents don't trample each other). overwrite=True
is the default since most download flows are intentional refreshes.
BREAKING CHANGE: renamed step from file_download to download
and updated method signature to accept local_path and overwrite parameters.
refactor(list): switch from file_store index to direct filesystem walk
Reads directly from the filesystem (Path.iterdir / Path.rglob),
**not** the file_store index. The store may lag behind disk during
indexing or after rapid mutations; for the most current view,
the on-disk walk is the source of truth.
BREAKING CHANGE: renamed step from file_list to list and removed
dependency on file_store.graph for enumeration.
feat(property): restrict property operations to markdown files only
Property steps now operate only on .md files; non-markdown targets
get error="not markdown" and the call is **not** executed.
BREAKING CHANGE: property steps (read, update, delete) now validate
file extension and reject non-markdown files.
refactor(stat): rename file_stat to stat
BREAKING CHANGE: step name changed from file_stat to stat.
refactor(upload): rename file_upload to upload
BREAKING CHANGE: step name changed from file_upload to upload.
feat(traverse): support multiple seeds in graph traversal
Allow path parameter to accept either a single seed (str) or
a list of seeds for batch traversal operations.
Add support for `paths=None` in `get_nodes` method to return all real
nodes as the graph's "scan everything" entry point. When `paths` is
explicitly passed as an empty list, returns empty list. Virtual
placeholders are filtered out in both cases.
BREAKING CHANGE: The `get_nodes` method signature changed from
requiring a list of paths to accepting an optional list or None.
refactor(memory): move runtime_response imports to steps package
Move runtime_response imports from memory package to steps package to
improve code organization and reduce circular dependencies.
refactor(config): update obsidian configuration table structure
Update the obsidian configuration markdown table to improve clarity
and fix incorrect section headings. Change 'stat/list' to 'file' and
'return specific tag info' to 'tags' section with proper commands.
chore(memory): remove deprecated toolkit modules
Remove deprecated agent_toolkit.py, file_toolkit.py, and
graph_toolkit.py modules as functionality has been moved to steps
package.
feat(steps): add new CRUD file operations
Add new steps for file operations including file_download and
file_list operations with proper path resolution and filtering
capabilities.
- Rename local_file_graph.py to nx_file_graph.py and update class name to NxFileGraph
- Add proper import error handling for networkx dependency in NxFileGraph
- Create new LocalFileGraph implementation using pure Python dict-based storage
- Update base_file_graph to include abstract methods for node/link operations
- Modify upsert_nodes to handle multiple nodes and edge management
- Implement proper edge bookkeeping with inverse and pending edge tracking
- Add rebuild_links functionality to reconstruct graph relationships
- Update logging to show real vs virtual nodes and edge counts
- Maintain backward compatibility with existing graph persistence mechanisms
- Rename local_file_graph.py to nx_file_graph.py and update class name to NxFileGraph
- Add proper import error handling for networkx dependency in NxFileGraph
- Create new LocalFileGraph implementation using pure Python dict-based storage
- Update base_file_graph to include abstract methods for node/link operations
- Modify upsert_nodes to handle multiple nodes and edge management
- Implement proper edge bookkeeping with inverse and pending edge tracking
- Add rebuild_links functionality to reconstruct graph relationships
- Update logging to show real vs virtual nodes and edge counts
- Maintain backward compatibility with existing graph persistence mechanisms
BREAKING CHANGE: Removed the iter_nodes abstract method from BaseFileGraph
and its implementations in LocalFileGraph and Neo4jFileGraph as it was
no longer being used in the codebase.
Remove path and line numbers from chunk ID calculation to improve
cache stability. The chunk identity now uses only the content hash
instead of including path::start::end coordinates.
This change removes the documentation about chunk identity format
and updates the hash_text call to use only the text content.
Add mistletoe as a project dependency for enhanced markdown parsing
capabilities. Refactor the LinkedFileParser to use a proper AST-based
approach with MdNode tree structure, replacing the previous flat token
processing method. The new implementation provides better handling of
markdown elements including tables, code fences, lists, and headings,
with improved chunking logic that maintains document structure in
generated content segments.
The changes include:
- Add mistletoe dependency to pyproject.toml
- Implement proper AST node representation with MdNode class
- Create recursive chunking algorithm with TOC preservation
- Add support for frontmatter extraction with FileFrontMatter schema
- Optimize leaf node splitting with proper boundary detection
- Include part numbering for split content pieces
- Replace legacy line-based chunking with AST tree approach that builds
a complete document skeleton with content inlined under relevant
sections
- Add new chunking parameters: chunk_chars (default 2000) and embed_toc
(default True) to control content size and TOC inclusion
- Implement recursive chunking algorithm that respects structural
boundaries (code lines, table rows, list items) and prevents splits
inside blocks
- Introduce part markers [Part X/N] for oversized leaf blocks that
require splitting
- Add CLI tool for inspecting parsed chunks and edges with options for
preview and configuration
- Refactor edge extraction to use FileEdge.from_text instead of
parse_wikilinks for consistency
BREAKING CHANGE: Chunk format changes significantly with full TOC
skeleton wrapping content, affecting embedding models expecting
breadcrumb prefixes.
Remove SqliteFileStore backend and simplify the base file store
interface. The file store now focuses on in-memory + JSONL persistence
for the (file → chunks) graph with reduced complexity.
BREAKING CHANGE: SqliteFileStore is removed, only LocalFileStore remains
available.
docs(protocol): add typed edge link protocol documentation
Add comprehensive documentation for the link protocol supporting
typed edges in body text. This includes specification for three
legal inline forms (bare wikilink, line-level Dataview,
inline-bracketed Dataview), predicate syntax rules, and the
machine-managed Relations section convention for organizing
discovered edges.
fix(memory): update path reference from vault_root to working_dir
Change the memory_create operation's path anchoring from
vault_root to working_dir to maintain consistency with the
current working directory configuration.
refactor(components): remove edge_extractor module and simplify parsing
Remove the edge_extractor component module entirely and
inline edge extraction logic directly into LinkedFileParser
using parse_wikilinks utility. This simplifies the architecture
by eliminating the separate edge extraction component and
delegating edge discovery to the maintainer's enrichment operations.
feat(parser): update parse method signature and simplify edge extraction
Modify LinkedFileParser to return (FileNode, list[FileChunk])
tuple instead of ParsedFile, remove dependency on BaseEdgeExtractor,
and implement direct wikilink parsing from body text only.
```
