Skip to content

provider errors give no diagnostic detail, no turn resume, no duplicate-terminal detection #674

Description

@Svector-anu

issue 1: provider errors give no diagnostic detail

session events of type "error" only ever store the top-level message
string, e.g.:

{"type":"error","payload":{"message":"provider error: provider returned error"}}

no http status code, no response body, no upstream error cause is
persisted anywhere in events.jsonl. there is no --verbose/--debug flag
or env var to capture more detail either. when a provider call fails,
there is no way to determine why from the cli or its logs.

repro: session zero_20260713163241_1783960361665303000_1, events
#361, #586, #632, #646, #669, #689 (2026-07-14, 06:12–08:01 utc),
provider=gitlawb-opengateway, model=tencent/hy3. all 6 have identical
generic messages with zero additional context.

issue 2: no resume after a provider error mid-turn

after a provider error interrupts a turn, the next turn does not
resume from where it left off — it re-reads previously-read files
from scratch to rebuild context. work in progress at the time of the
error is lost, not just the single failed request.

issue 3: no visibility into duplicate background terminals across sessions

three separate bun run dev background terminals were left running
concurrently from three different zero sessions, started 06:33, 08:18,
and 10:54 utc on the same date, same command, same working directory.
nothing in the cli flagged that an identical background terminal was
already running before starting another.

Metadata

Metadata

Labels

No labels
No labels

Type

No type

Fields

No fields configured for issues without a type.

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions