Set the maximum number of entries to return. # Arguments `n` - Maximum number of entries

List all tags across all views.

List files in a directory. Returns a sorted list of entries as `(filename, is_directory)` tuples. # Errors Returns an error if the path is not a di

List identities matching a filter.

Load global configuration from the default location

Load the full content blob for a change by concatenating all chunks. This is equivalent to reading all CONTENT chunks and joining them. # Returns T

Load all graph sections for a change. Reads from CHANGE_GRAPH only. This is the "thin apply" path — the minimum data needed to apply a change to the

Load an existing session or create a new one. This is the resilient path — if a hook arrives for an unknown session (e.g., the session state was lost

Look up the index for a given hash. Returns `None` if the hash is not in the table. This is an O(1) operation using the internal `HashMap`. # Argume

Helper: create a Change with a hunk.

Helper to create a test edge.

Orchestrator rooted at a real initialized repository (recording works).

Mark an inode as recorded at a specific position. This prevents duplicate recording of the same inode. # Arguments `inode` - The inode being record

Materialize only specific files to the working copy. This is used after `insert` operations to only rewrite files that were actually affected by the

Materialize a view's visible file content into `dir` (created if needed). Writes each visible file's bytes exactly as they appear on `view` (the view

Merge stats from a single operation.

Merges another stats instance into this one.

Merge another set of hashes into this table. This is useful during parallel recording where each thread collects referenced hashes independently, the

Get the merkle state if the channel has state.

Get the change message.

Get the file metadata.

Returns a human-readable name for error messages.

Check whether a change needs semantic regeneration. Returns `true` if the change has GRAPH sections but no SEMANTIC sections in the redb store. Retur

Open or create a pristine database at the given path This will create all necessary tables if they don't exist.

Returns the expected ordering index for this section type. Lower values must appear earlier in the file. Sections with the same ordering index (like

Parse a markdown file's YAML frontmatter and body. Expects the file to optionally start with `---\n...\n---\n`. Returns `(frontmatter_json, body)` wh

Format a file path in bold. Use this for displaying file paths to make them stand out. # Arguments `text` - The path to format # Returns A styled

Populate the session tables from a Sherpa provenance graph. Called during `save_provenance_graph` when the graph has `profile == "sherpa-trace/1.0.0"

Set the previous provenance graph hash (for session resume chaining).

Set the previous attestation hash (for session resume chaining).

Print header information for a change diff. Displays the change hash, message, author, and timestamp before showing the actual diff content.

Print the outcome of a cross-view insert operation.

Process an equal (unchanged) region. Equal regions update position tracking but don't create changes. They may cause pending changes to be flushed if

Process a replacement with individual parameters. # Arguments `old_pos` - Starting position in old content `old_len` - Number of lines being replace

Get the default progress bar style. Format: `[████████░░░░░░░░] 50/100 Message (50%)`

Get the filesystem path for a provenance graph with the given hash. Provenance graphs use the same two-level directory structure as changes and attes

Add an operation to the result.

Read all remaining sections and verify the content hash in one call. This is a convenience method that combines [`read_all_sections`](Self::read_all_

Collect all existing learning lines (flat list) for dedup checking.

Record a successfully downloaded change. Increments the download counter and adds the bytes to the transfer total. # Arguments `bytes` - The size o

Record a conflict with the given type.

Record that a directory was created.

Record an error.

Record a processed file.

Record a skipped file.

Returns the list of files that were recorded in this turn.

Regenerate SEMANTIC sections for all changes that need it. Scans the redb store for changes that have GRAPH sections but no SEMANTIC sections, and re

Regenerate SEMANTIC sections for a single change from its GRAPH + CONTENT. This is the core regeneration function. It: 1. Loads the GRAPH sections to

Remove a single file from tracking. This is the low-level function that actually modifies the database. # Arguments `txn` - A mutable transaction `

Add a repository pattern.

Resolve an index back to its 32-byte hash. Returns `None` if the index is out of bounds or is the "none" sentinel ([`HASH_INDEX_NONE`]). # Arguments

Try to resolve identity from git config in the repository.

Pure identity resolution from a URL plus the configured server bindings. userinfo → server-host match (longest suffix) → first-label subdomain. Split

Resolve an inode to its graph position. This function looks up the position in the repository graph where a file's content root is located. # Argume

Resolve an index back to its hash, returning an error if not found. This is a convenience method that wraps [`resolve`](Self::resolve) and converts `

Resolve an index to a hash, erroring if it's `NONE` or out of bounds. Unlike [`resolve_or_none`](Self::resolve_or_none), this method treats [`HASH_IN

Persist the graph to disk at `{session_dir}/graph.json`. Uses atomic write (temp file + rename) to prevent corruption.

Save a change using pre-serialized V3 bytes (hash-stable). This writes the exact V3 bytes to disk, ensuring the file hash matches the hash registered

Search the content index. Returns matching lines with file paths, line numbers, and content. Supports path filtering, file type filtering, and case-i

Internal: read only sections of a specific type, skipping all others. Skipped sections are still fed through the hasher for verification.