Execute the init command. This method: 1. Resolves and validates the target path 2. Validates the view name 3. Creates the directory if it doesn't ex

Save a change to the repository. The change is serialized and written to the `.atomic/changes/` directory using a content-addressed two-level directo

Save learnings to the agent's context file. If the file doesn't exist, creates it with the learnings section. If the file exists but has no learnings

Sign a message

Convert a [`ProvenanceGraph`] into a [`TraceRecord`] suitable for appending to `.agent-trace/traces.jsonl`. `root` is the repository root used for co

Iterate over untracked files.

Extract and store KG data for a vault entry in one step.

Builder: set view filter.

Set a change filter for view-aware output.

Builder: set files to diff.

Builder: set the initial view name.

Open a file for writing, creating it if necessary.

Write this header to a writer. Writes exactly 64 bytes. # Errors Returns an I/O error if the write fails.

Add a single file to tracking. This is the low-level function that actually modifies the database. It does NOT check if the file exists on disk or is

Append a patch proposal node (change recorded). Returns the new node's ID.

Run `atomic <args>` inside `dir`.

Canonicalize a JSON value into its RFC-8785 string form.

Get a reference to the underlying database.

Pure decision core for identity resolution with a default fallback. Split from [`resolve_identity_with_default_fallback`] so the fallback logic can b

Compute the diff between two sequences of lines. This is the main entry point for diffing. It takes two slices of [`Line`]s and returns a [`DiffResul

Diff two snapshots to produce a `TurnChanges`.

Execute a query plan against a repository.

Get file content as it was BEFORE a specific change was applied. This method retrieves the content of a file at the state immediately prior to a chan

Get the inode for a tracked path. # Arguments `txn` - A transaction (read or write) `path` - The normalized path string

Returns this trunk's branches in **file order** — top-of-file first. Iteration over `TRUNK_BRANCHES` alone returns branches in `(change_id, branch_id

Get the number of entries (files + directories).

Lift an authored memory (frontmatter + body) into a canonical node, attest it (hash + sign) with the given identity, and return the node. Mirrors [`li

Normalize a bare id, a `memory/<id>` path, an `<id>.md` name, or a full `memory/<id>.md` path all to the canonical vault path `memory/<id>.md`. This

Output the content of an alive graph to a span buffer. This function traverses the graph in SCC order (as computed by [`compute_order`](crate::output

Read the contents of a file into a buffer. The buffer is **appended to**, not cleared.

Record a directory creation. Duplicate directory paths are automatically deduplicated. # Arguments `path` - The path of the directory that was crea

Initialize a repo with a single recorded file `file.txt` = "v1\n".

Set the columns with full configuration. Use this when you need to customize alignment or width constraints. # Arguments `columns` - The column con

Sets the content range.

Walk the directory tree and build a snapshot of all files. Skips directories matching any of the ignore patterns. Only includes regular files (not di

Attempt to merge a conflict group using token-level semantics. # Algorithm 1. Extract the two competing vertices (only 2-way supported). 2. Retrieve

Get the neighborhood subgraph around a node.

Deflate changed vault files from disk into redb. Scans the vault working copy, detects changes, and updates the database. Returns the list of paths t

Verify an attested node's content hash and signature against a public key.

Builder: set the AI-assisted flag.

Builder: set the AI provider.

Builder: set the diff algorithm.

Builder method to set column

Builder method to set exported flag

Set the new line number.

Mark that this change has a provenance section. Automatically sets the `HAS_PROVENANCE` flag.

Write the optional UNHASHED section. This section is NOT included in the content hash. It's intended for data that shouldn't affect the change's iden

Add a detected change.

Add a change hash that this graph explains.

Add a file item.

Adds a new file with content, automatically tokenizing into lines.

Add files to the accumulated files_touched list (deduplicating).

Add a built graph_op to the result.

Add a zombie span (simplified API).

Align the view pointer without materializing the working copy. This persists the new view name to `.atomic/current_view` so that subsequent `status()

Append raw bytes to the contents buffer. Returns the starting position of the appended content, which can be used to reference this content in hunks.

Calculate which changes need to be pulled from the remote. Compares remote and local changelists to determine which remote changes are missing locall

Get the number of children for a span. This includes both regular and extra children.

Collect files from the working copy. Scans the working copy to find all files present on disk. This is the complement to `collect_tracked_files` for

Convert a `Position<Option<Hash>>` to a [`CompactPosition`]. # Errors Returns [`FormatError::HashNotFound`](super::super::error::FormatError::HashNo

Convert a remote error to a CLI error. Maps the various remote error types to appropriate CLI error types with user-friendly messages and suggestions

Set the total cost in USD.

Creates a file creation operation. # Arguments `trunk_id` - The trunk ID for the new file `path` - The file path `encoding` - Optional text encoding