v0.6.4 — Indexed document inspection

CtxVault v0.6.4 adds the ability to inspect and export the text currently stored in a semantic vault, without touching the source file on disk.

Added

Indexed document inspection — ctxvault docs <vault> --show <doc-id> displays the text reconstructed from stored chunks alongside document metadata: source path, file type, chunk count, indexed timestamp, and a SHA-256 hash of the reconstructed text. A warning is shown whenever the document spans more than one chunk, since adjacent chunks may produce minor overlap at boundaries.

Indexed document export — ctxvault docs <vault> --export <doc-id> --output <path> writes the reconstructed text to a file. Useful for diffing against a potentially modified or corrupted source file before deciding whether to re-index.

GET /docs/{doc_id}/content API endpoint — returns the same reconstructed text and metadata programmatically. Requires vault_name as a query parameter. Returns 404 if the document id is not found, 400 if called on a skill vault.

reconstructed_text_hash field — SHA-256 of the reconstructed chunk text. Documented explicitly as a hash of the reconstructed content, not the original source file, to avoid confusion when comparing against an external checksum.

Notes

The text returned by both the CLI and the API is reconstructed by joining indexed chunks in order, separated by double newlines. It is a processed approximation of the original document content, not an exact copy. Markdown formatting and other source-level structure is not preserved. This feature is intended for recovery and debugging workflows, not as a replacement for versioning vault data with Git.

What's unchanged

All existing CLI commands, API endpoints, MCP tools, and vault operations work exactly as before.