Documentation
¶
Overview ¶
Package seed seeds demo data by replaying bundled capture corpora through the normal ingest write path.
The old demo seed fabricated derived rows (hand-written node chains) directly into storage; that data could drift from what real capture produces and exercised none of the pipeline. This seed instead replays real captured sessions — the gzipped raw_turns JSONL corpora also used by the derive regression tests — through an in-process ingest server, then runs the deriver. Seeded data is therefore indistinguishable from live capture: it lands in raw_turns, creates sessions rows via the same ingest transaction, and projects traces/ spans through the same derive pass.
Re-running the seed is a no-op: raw-turn deduplication (org_id + request_id) absorbs replayed wire turns, the transcript dedup key includes a content hash, and the derive pass is idempotent.
Index ¶
Constants ¶
This section is empty.
Variables ¶
var ErrUnsupportedDriver = errors.New("demo seeding requires the raw-turn layer (Postgres driver)")
ErrUnsupportedDriver is returned when the storage driver cannot host the raw-turn layer or the derive pass (e.g. the in-memory driver).
Functions ¶
This section is empty.
Types ¶
type Result ¶
type Result struct {
// Sessions is the number of demo sessions the corpora replay into.
Sessions int `json:"sessions"`
// RawTurns is the total number of corpus rows replayed.
RawTurns int `json:"raw_turns"`
// RawTurnsInserted counts rows that landed as new raw turns;
// RawTurnsDeduped counts replays the raw layer's dedup absorbed
// (a re-seed reports everything deduped).
RawTurnsInserted int64 `json:"raw_turns_inserted"`
RawTurnsDeduped int64 `json:"raw_turns_deduped"`
}
Result summarizes one seeding run.
func Run ¶
func Run(ctx context.Context, driver storage.Driver, logger *slog.Logger, orgID string) (*Result, error)
Run seeds the demo corpora into the given driver for orgID (the nil-UUID sentinel or "" both mean the default tenant). It replays every bundled corpus through an in-process ingest server — the same handlers wire capture lands on — and then derives each seeded session synchronously so the trace/span projection is queryable the moment the call returns (no derive worker required).
Source Files