Documentation
¶
Overview ¶
Package legacy reads the prosa v1 ".prosa" data bundle (a SQLite catalog + content-addressed zstd-compressed raw source files) so the v3 importers can re-ingest historical sessions whose original source files have since disappeared from the filesystem (e.g., Claude Code retention deleted ~/.claude/projects/.../*.jsonl past a few weeks).
See docs/sources/legacy-bundle.md for the on-disk layout, and internal/cli/sync.go for the `prosa sync --legacy-bundle` wiring that drives this package end-to-end.
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
This section is empty.
Types ¶
type Bundle ¶
type Bundle struct {
Path string
// contains filtered or unexported fields
}
Bundle wraps a read-only handle to a v1 prosa bundle directory. The caller MUST Close it when done.
func Open ¶
Open validates that path looks like a v1 bundle (has prosa.sqlite alongside raw/sources/) and opens the SQLite catalog read-only.
func (*Bundle) Decompress ¶
Decompress reads <bundle>/raw/sources/<ObjectIDHex>.zst, falling back to an uncompressed <ObjectIDHex>.bin object when present. It writes the source bytes to <tmpDir>/<basename(OriginalPath)> and returns the temp path. Preserving the source basename matters for codex (the importer falls back to the filename UUID when session_meta is absent) and for cursor (where the temp file is opened as a SQLite db).
Respects ctx.Done() before opening the source file. The decompress itself is not cancellable mid-stream — zstd reads can take seconds on large files — but that's bounded by per-file size in practice.
func (*Bundle) SourceFiles ¶
func (b *Bundle) SourceFiles(ctx context.Context) ([]SourceFile, error)
SourceFiles returns the catalog rows the v3 pipeline can re-ingest. Only the tools v3 has importers for are returned; v1 also stored `hermes`, which had zero rows in practice and stays excluded.
type SourceFile ¶
type SourceFile struct {
// Tool is the v1 source_tool string: "claude" | "codex" | "cursor" |
// "gemini". sync.go maps this to a v3 importer instance.
Tool string
// OriginalPath is where the source file used to live on disk (e.g.
// /Users/u/.claude/projects/.../<uuid>.jsonl). Used as the basename
// for the decompressed temp file so downstream parsers' filename
// fallback works.
OriginalPath string
// ObjectIDHex is the content hash without the "blake3:" prefix —
// matches the filename in <bundle>/raw/sources/<hex>.zst or .bin.
ObjectIDHex string
// SizeBytes is the uncompressed source file size at v1 import time.
SizeBytes int64
}
SourceFile is one row from the v1 `source_files` table, scoped to the fields the v3 re-ingestion pipeline actually consumes.
Source Files