Embed a single vault entry using the provided embedding function. Chunks the content by heading/paragraph, computes embeddings via `embed_fn`, and st

Delete an unstarted backlog intent. This is intentionally conservative: only backlog intents with no linked goals can be deleted. Started work should

Get the current view name.

Compute the set of file paths visible on a view. Visibility includes the view's own changes AND all changes inherited through the parent chain. A dr

Sets whether to use code-aware tokenization.

Sets whether to use code-aware tokenization. When enabled, the tokenizer recognizes multi-character operators, string literals, numeric literals, and

Builder: set the dry-run flag.

Builder: set the `--force` flag.

Builder: set the output format.

Builder: set the starting sequence number.

Builder: set full hash display.

Set the maximum file size for diffing. Files larger than this are treated as binary. # Arguments `size` - Maximum size in bytes, or `None` for no l

Set the modification time.

Builder: set the name-only flag.

Builder: set the no-color flag.

Builder: set whether to hide untracked files.

Builder: set the remote name or URL.

Builder: set the remote state.

Set the tagged flag. Tagged changes represent named snapshots in the history. # Arguments `tagged` - Whether this change is tagged # Example ```r

Builder: set tags-only filter.

Builder: set the timeout in seconds.

Add a warning.

Add another `PhaseTokens` into `self` (used for running totals).

Add a credit to the statistics.

Adds a branch operation with its ID.

Add a directory item.

Associate an inode update with a graph_op. When the change is applied locally, this update will be used to maintain the tree/inode database tables.

Adds a leaf operation with its ID.

Adds a new line to a file and returns its branch ID. # Arguments `trunk_id` - The file to add the line to `after` - The branch to insert after (None

Adds a line to this file.

Record a missing up-context position.

Add a single path to tracking.

Add a pending edge to be created.

Create statistics for an added file. # Arguments `path` - The file path `lines` - Number of lines in the new file

Aggregate tool usage from condensed entries.

Create options that include everything.

Set whether to allow conflicts.

Append content to the buffer and return the position range. # Arguments `data` - Content bytes to append # Returns A tuple of (start_position, end

Append content to the contents blob. Returns the starting position of the appended content.

Persist deferred operations atomically. The caller holds pristine's write transaction, which serializes journal writers across processes.

Append an LLM response node — the agent's final answer for the turn. Created at turn end from the agent's closing message: the stop payload's respons

Truncate bodies so their combined length fits `budget_chars`. Start with an even allocation, then give unused capacity from short bodies to higher-ran

Independent oracle: base with the given (index → new line) replacements.

Fold messages and parts into [`TurnData`]. The current turn starts at the last user message; reasoning, response, and step statistics are scoped to t

Assert content after a specific change matches.

Run `atomic <args>` inside `repo_dir` with `HOME` pointed at `home_dir` so the identity store resolves to a test-owned location.

Attach a Bearer JWT auth header to the remote config. `identity_override` — if provided, this identity name is used directly, bypassing URL-based inf

Lift+attest a stored memory with the given keypair.

A minimal, fully attested Memory that conforms — the base for mutation.

Build inputs for a minimal, liftable intent and attest them into a node.

The attested-node sidecar file for an intent.

The attested-node sidecar file for a memory.

Build a compact text representation of KG nodes and edges for the LLM prompt. Groups nodes by kind and renders edges as relationship sentences. Hard

Build an identity filter from the command arguments.

Build the summarization prompt from the condensed transcript.

Build the HTTP remote configuration. `identity_hint` — explicit identity name to use (from `--identity` or `RemoteEntry.identity`). When `None`, fall

Build the HTTP remote configuration. Creates an `HttpRemoteConfig` with the timeout and security settings specified by the user.

Build the HTTP remote configuration. `identity_hint` — explicit identity name to use (from `--identity` or `RemoteEntry.identity`). When `None`, fall

Build a tree structure from flat paths. Organizes items into a hierarchical structure by parent-child relationships. # Arguments `items` - Flat lis

Iterate over only the change operations (non-Equal).

Resolve `urn:atomic:change:<hash>` evidence URNs against the change store at `repo_root`. One violation per URN that does not resolve. With no reposit

Chunk text by heading boundaries, then by paragraph, respecting a maximum token budget per chunk. Strategy: 1. Split on `#` headings (any level) 2. I

Split content into chunks and return chunking statistics. This is the same as [`chunk_content`] but also returns [`ChunkingStats`] for logging and pr

Classify an imported commit as normal, merge, or squash.

Set whether to collect directories. # Arguments `collect` - Whether to include directories in results # Example ```rust use atomic_core::output::r

Set whether to collect files. # Arguments `collect` - Whether to include files in results # Example ```rust use atomic_core::output::repo::TreeCol

Collect view-scoped TREE lifecycle events before applying hunk-level TREE mutations. Recording all adds, moves, and deletes, not just deferred Git imp

Find the length of common prefix and suffix between two sequences, applying the **suffix-limiting heuristic** used by the display path. This optimiza

Convert an `Insertion<Option<Hash>>` to a [`CompactInsertion`].

Compare content with size limit check. Like `compare_content`, but treats files over the size limit as binary. # Arguments `old_content` - Content

Compute the USD cost for a pair of token counts. Uses [`COST_PER_M_INPUT`] and [`COST_PER_M_OUTPUT`] so every layer that touches token accounting pro

Compute token-level changes between two lines. This is the core of the token-level diff - it compares the tokens of two lines and produces a sequence

Condense a transcript from raw bytes, auto-detecting format. Currently supports Claude Code JSONL. Future: Gemini JSON, other formats.

Create token changes for a line that was entirely added.

Set the delegating identity (for delegated identities).

Delete a change from the store. Removes all entries for this change from CHANGE_META, CHANGE_GRAPH, CHANGE_SEMANTIC, CHANGE_CHUNKS, and CHANGE_UNHASH

Create statistics for a deleted file. # Arguments `path` - The file path `lines` - Number of lines in the deleted file

Deserialize the payload from postcard format. Convenience wrapper around [`postcard::from_bytes`] that handles the error conversion. # Type Paramete

Deserialize bytes (JSON) to a list of conflicts.

Detect the encoding of content. # Arguments `content` - The content to analyze # Returns The detected encoding.

Walk up from `start` looking for a [`SANDBOX_POINTER`] file. If found, return `(working_root, canonical_root, view)`.