The tracked-vault path for an intent's attestation: `attestations/<sanitized-normalized-id>/attested.md`.

Add multiple authors.

Build the delegation scope.

Build a Go-like source file with `count` distinct top-level functions, each with a unique, greppable marker line. Mirrors the shape of the real-world

Build a graph where the vertex at `parent_idx` has the given child vertex indices. Vertices are pushed in order 0..N where 0 is DUMMY.

Collect all change `NodeId`s visible from a view and expand their indexed dependency closure without loading `.change` files. This is the dependency-

Returns the number of registered agents.

Create a directory and all parent directories.

Create a multi-progress container for parallel progress bars. A [`MultiProgress`] allows displaying multiple progress bars simultaneously, useful for

Create a test change with some content.

Skip TLS certificate verification (dangerous!). Only use this for testing or when connecting to servers with self-signed certificates that you trust.

Remove learnings that already exist in the context file. Returns a new `Learnings` with only the Repo and Workflow entries that are not already prese

Set an optional description.

Detect the encoding of content. Analyzes content to determine if it's valid UTF-8 text or binary data. # Arguments `content` - The content to analy

Ensure the workspace directory for a view exists. Creates `.atomic/workspaces/<view_name>/` and any intermediate directories. This is called from `i

Get the file extension for this format.

Extract edges from content text (wiki-links, file paths).

Extract unhashed turn data from a change. Returns `None` if the change has no unhashed data or no `"agent_turn"` key.

Get metadata for a file or directory.

Filter to a specific path prefix.

Recompute the contents hash after modifying contents.

Find the vault path for an intent by its display ID. First checks the manifest's `vault_path` field (fast path), then falls back to scanning vault en

Find the repository root starting from a specific path. Like [`find_repository_root`], but starts from the given path instead of the current working

Indices of unclassified tool-derived nodes grouped into sequences separated by structural boundaries (goals, patches, human gates, errors).

Find lines that appear exactly once in both sequences and match. Returns a list of matches sorted by position in the old sequence.

Finish a progress bar with an error message. This clears the progress bar and displays an error message in red. # Arguments `pb` - The progress bar

Format condensed entries as human-readable text. Output format: ```text [User] Fix the auth bug in login.rs [Assistant] I'll fix the token validation

Format learnings as individual list items grouped by category. Only includes **Repo** and **Workflow** learnings. Code learnings (with file:line refe

Format entries for oneline output. # Arguments `entries` - History entries to format `hash_length` - Number of hash characters to display # Returns

Format the change for short output.

Build hunks from a diff result with context lines. This function groups diff operations into hunks with the specified number of context lines around

Generate diff operations between two content blobs. Converts content to lines and runs the specified diff algorithm. # Arguments `old_content` - Ol

Get the default remote, if one is configured. Returns the first remote marked as default, or "origin" if it exists and no other default is set. # Ex

Get the raw contents of a file. Returns `None` if the file doesn't exist or is a directory. # Example ```rust use atomic_core::output::Memory; let

Get the message for the revised change.

Get the Merkle state at a given sequence number.

Filter by identity type.

Set whether to include deleted content.

Set whether to include deleted vertices. # Arguments `include` - Whether to include deleted content # Example ```rust use atomic_core::output::rep

Initialize the embeddings table.

Mark a vertex as "skip" — it belongs to a resolved SCC but is not the lead vertex.

Returns true if the token is alive (not deleted).

Check if this status represents a change that needs to be recorded. Returns `true` for statuses that indicate uncommitted changes.

Check if the builder is empty (no changes recorded). # Returns `true` if no hunks have been recorded yet.

Check if no files were collected.

Check if the tracker is empty.

Determine if a declaration is exported (publicly visible). Walks the immediate children of the declaration node looking for access-level modifiers. S

Iterate over all attestation hashes on disk.

Returns the number of selected layers.

Get the number of signatures.

Lift the given inputs into a canonical `MemoryNode`, mapping a lift failure (missing `uid`, missing `memoryKind`, …) to a clean argument error.

List all tracked files. # Arguments `txn` - A transaction (read or write) # Returns A vector of all tracked files and directories.

Load an intent's attestation sidecar and classify it against the current source (`inputs`). A missing sidecar is `None`; a corrupt/malformed one is tr

Load an attestation from the repository by hash.

Load a provenance graph from the repository by hash.

Create a short prefix from a session ID for use in node IDs. Takes up to the first 8 characters. If the session ID is a UUID, this is the first segme

Merge another outcome into this one. This is useful when collecting results from multiple output operations (e.g., outputting different subtrees). #

Merge results from a file output.

Merge new items into existing section headings without duplicating headings. If `### Repo` already exists in the section, new repo items are appended

Merge `source` paths into `target`, deduplicating.

Iterate over modified files.