Live logs¶
FastProof shows test output while tests run: open a launch, pick a running test and its log grows in the
browser. Results can be opened before they finish (status: "running") and are completed later by the same
client id, so the launch page shows what is running right now.
The reporters stream logs for you. This page describes what they send, for anyone writing their own integration.
Sending logs¶
POST /api/v1/projects/{key}/launches/{id}/logs
Authorization: Bearer <token with launches:write>
Content-Type: application/x-ndjson
{"resultId":"checkout-1","attempt":0,"seq":1,"ts":"2026-11-21T10:00:00.120Z","level":"info","stream":"log","message":"opening the cart"}
{"resultId":"checkout-1","attempt":0,"seq":2,"ts":"2026-11-21T10:00:00.480Z","level":"error","stream":"stderr","message":"timeout waiting for #pay"}
A JSON array of the same objects is accepted too.
| Field | Meaning |
|---|---|
resultId |
The client id of the result, the same one used when the result is reported |
attempt |
Attempt number, starting at 0; retries get their own log |
seq |
Sequence number within the attempt; lines are shown in this order |
ts |
Time of the line (RFC 3339) |
level |
trace, debug, info, warn or error |
stream |
stdout, stderr, log or step |
message |
The line |
step |
Optional step name for stream: step |
Lines are idempotent by (resultId, attempt, seq): resending a batch after a timeout does not duplicate
anything. Line size, lines per request and bytes per launch are limited; a request over
the limits is rejected with a clear error. Mask secrets before sending: the reporters do it at the
source.
Following logs¶
A server-sent event stream of log events, batched to at most 20 per second per subscriber. When a client
reads too slowly, the oldest lines are dropped and a gap event says how many; fetch them from the stored
logs if needed. Filters: resultId, minimum level, and a text q.
With several replicas every replica sees every line: through PostgreSQL LISTEN/NOTIFY in direct mode, or by
reading the ingest.logs Kafka topic in Kafka mode.
Reading stored logs¶
Pages backwards from before. Logs are stored as compressed chunks per attempt in PostgreSQL, and copied to
ClickHouse or OpenSearch when a sink is enabled.
Searching logs¶
GET /api/v1/projects/{key}/search/logs?q=&launchId=&from=&to=&level=&limit=
GET /api/v1/projects/{key}/results/{id}/similar
Search across launches and "similar failures" run on the configured search backend: PostgreSQL by default, ClickHouse or OpenSearch for large volumes (see Scaling and big data).
All endpoints need a token or session with read access to the project; sending needs launches:write.