Diagnosing failures

Every execution is recorded, so when a request fails you don’t need to re-run it with extra flags: look it up by id and compare it with the last time it worked.

What gets recorded

Every execution of run (single requests, --chain steps, --data rows), collection run and shell send, including attempts that never got a response:

resolution problems (missing variables, bad types)
connection errors and timeouts
pre-request and post-response script errors
failed assertions and non-2xx responses

Dry runs are not recorded. Records store the method, URL, headers, body, variables with their origins, response headers, error, failure phase and assertion results, with secrets masked before saving. Response bodies are stored as received.

1. Get the execution id

A failed run prints it on stderr:

request failed: HTTP 500 (id: 078b9896) — inspect with: gurl diag 078b9896

Other ways to find it:

gurl history "health"                 # newest first, with short ids and errors
gurl history "health" --json          # [{"id","short_id","request","time","ok","status",...}]
gurl timeline --since 1h              # across all requests
gurl run "health" --json | jq -r .short_id

2. Inspect it

gurl diag 078b9896        # any unique prefix works; or: gurl diag --id 078b9896

diag shows:

Summary: result, status, failure phase (request_build, http, extraction, assertion, status, pre_request_script, post_response_script, run_if, next_request), duration, size
Request as sent
Variables, each with its variable origin (environment, collection, data, cli, or a runtime value) — secret values show as *****
Response headers and body (first 4KB; --full for all of it)
Assertions
Compared with last success: what changed since the last successful run of the same request — status, duration, URL, headers and body, variable values and origins, response headers and top-level JSON fields

--json prints the record and the comparison as JSON.

3. Read the failure phase

Phase	Usually means	Next step
`request_build`	A placeholder couldn’t be filled	Check the problems list; `gurl run <name> --dry-run`
`http`	Connection error or timeout	Host, network, `--timeout`
`status`	Non-2xx response	Response body in `diag`; compare with last success
`assertion`	An `--assert` failed	Assertion messages in `diag`
`extraction`	An `--extract` rule found nothing	Test the rule with `gurl extract` on the stored body
`pre_request_script` / `post_response_script`	A script threw	Script error in `diag`

Compare the latest responses

gurl diff "health"                 # last two responses
gurl diff "health" --json          # structured RFC 6902 changes, or a text diff

Secrets in records

Records never contain secrets: secret and credential-looking values are masked before saving, so diag compares secret variables by origin only (“came from environment both times”), never by value. In agent mode, diag and diff also mask secrets in response bodies and headers, using the secrets of both compared runs.