Returns the token content as a string (lossy UTF-8 conversion).

Returns the context file name for a given agent. Each agent reads a specific markdown file at session start for project context. Learnings are append

Returns the full path to the agent's context file in a repository.

Converts a line insertion to CRDT operations. # Arguments `trunk_id` - The file containing the line `after_branch` - The branch to insert after (Non

Convert token changes to insert-only LeafOps (for added/deleted lines).

Create an AliveVertex from an already-resolved span, if it's alive. This function: 1. Checks if the span is alive (has non-deleted edges) 2. Checks i

Creates a new branch (line) in a trunk. # Arguments `id` - The branch ID `trunk_id` - The parent trunk `after` - Insert after this branch, or `None`

Create a Insertion for file content. This creates a span containing actual file content, with proper up and down context for positioning in the graph

Creates a new leaf (token) in a branch. # Arguments `id` - The leaf ID `branch_id` - The parent branch `kind` - The token kind `content_range` - Byt

Create a new project in a workspace.

Create a new workspace.

Returns the duration of the current turn in milliseconds, if a turn is active.

Deep-merge `extra` into `root`: objects merge recursively, arrays union by value (no duplicates), scalars overwrite. Used for non-hook settings such a

Keep the default random view_name; do not pin it to `dev`.

Marks the leaf as deleted.

DELETE request, returns `()` on success.

Build an [`EdgeUpdate`] that marks every content vertex in the file as deleted. This is the single place where "find all content vertices → create de

Marks a line for deletion. Note: This creates a delete without content. For deletes with content (for diff display), use the `LineOps::delete()` cons

Delete a tag by name from the current view.

Create an input for a deleted file.

Get a human-readable description.

Detect the encoding of the given content. This function analyzes the content to determine its encoding: 1. Checks for BOM (Byte Order Mark) 2. Check

Download a change file by hash. # Arguments `hash` - The base32-encoded hash of the change. # Returns The raw change file data.

Embed a batch of texts using the resolved provider. Returns one vector per input text.

* Complete a tool invocation and return the elapsed duration. * * Removes the start time entry from the session. If the tool call * was not p

Ensure all of `change_id`'s spans are loaded into the in-memory index. Runs a single GRAPH range scan the first time a change is queried; later calls

Load `change_id`'s spans into the in-memory index on first access.

Derive a subject URI from a vault path and entry type.

Get the error completion style. Format: `✗ Message` (in red)

Get the exit code that should be used for this error. Different errors have different severity levels, reflected in their exit codes: - 1: General e

Convert a [`CompactGraphNode`] to a `GraphNode<Option<Hash>>`.

Check that the writer is in the expected state.

Export a change from redb to V3 `.change` file format. Reads all sections from redb and assembles a complete V3 change file. This is used for push/pu

Extract a command string from a verification node's detail.

Extract the first sentence from a single paragraph of text. Looks for sentence-ending punctuation followed by whitespace or EOF. Skips known abbrevia

Get the files that need to be restored, paired with their status. Derives the list from a status that was already computed by the caller (so restore

Add a path filter.

Finalize hunks by converting them to the serializable format. This validates that all hunks are properly formed and ready for serialization. # Argum

Find all attestations that cover a specific change. Uses REV_DEPS to find nodes that depend on the given change, then filters by `node_type::ATTESTAT

Resolve the span ending at `target` (or, failing that, containing it), preferring an empty marker at exactly `target`, if this change is loaded.

Find the sequence number for a change by its hash. # Arguments `txn` - Read transaction `view` - View to search `hash` - Hash of the change to find

Find the sequence number for a change by its hash. # Arguments `txn` - Read transaction `view` - View to search `hash` - Hash of the change to find

Retrieve content vertices via global GRAPH DFS.

Find the first `simple_identifier` child of a node.

Finds the insertion position for a new element among concurrent insertions. Given a new element ID and a list of existing elements that were all inse

Find the sequence number for a hash in the current view.

Find the `type_identifier` child of a declaration node.

Flush and close the writer, returning any errors. This is automatically called when the writer is dropped, but calling it explicitly allows you to ha

Finalize an uncompressed tar buffer into an [`OciLayer`]: compute the uncompressed digest (`diff_id`), gzip the bytes, and compute the gzipped digest

Helper: create a ChangeHeader with a fixed timestamp for deterministic tests.

Format a duration in milliseconds to a human-readable string.

Format the local state for display.

Format the remote state for display. # Arguments `state` - The remote's state response `entries` - The remote's changelist entries # Returns A for

Get a specific change by sequence number. # Arguments `txn` - Read transaction `view` - View to query `sequence` - Sequence number to retrieve # Re

Get the chunk manifest for a change. Returns the ordered list of (chunk_index, chunk_hash) pairs. This is used for delta transfer negotiation.

Get file content from a git tree.

Retrieve all lines for a file by trunk ID.

Get the files modified by a specific change. This function returns the paths of all files that were added, modified, or deleted by a specific change.

Get the hash display length.