fix(core): bump stream stall watchdog default to 5 min (#155)

Long agentic turns (parallel subagent dispatch, deep mid-stream planning
pauses) were tripping the 90 s per-chunk idle watchdog and surfacing as
"Stream stalled: no data received for 90s" even on healthy connections.
Bump the default to 300 s — still catches genuine frozen connections,
but gives legitimate slow-but-steady streams the headroom they need.
PROTO_STREAM_STALL_TIMEOUT_MS env override unchanged.

Co-authored-by: Automaker <automaker@localhost>
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: 3 people

docs/reference/settings.md

@@ -84,11 +84,11 @@ In addition to `settings.json`, the `.proto/` directory can contain:
 
 These variables take effect at runtime and do not require a settings file entry.
 
-| Variable                        | Default | Description                                                                                                                                                      |
-| ------------------------------- | ------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------- |
-| `PROTO_STREAM_STALL_TIMEOUT_MS` | `90000` | Max ms to wait between streaming chunks before the stall watchdog fires (throws a retryable error). Increase if complex agentic responses are hitting the limit. |
-| `PROTO_SYSTEM_DEFAULTS_PATH`    | —       | Override path to the system defaults settings file                                                                                                               |
-| `PROTO_SYSTEM_SETTINGS_PATH`    | —       | Override path to the system settings override file                                                                                                               |
+| Variable                        | Default  | Description                                                                                                                                                      |
+| ------------------------------- | -------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `PROTO_STREAM_STALL_TIMEOUT_MS` | `300000` | Max ms to wait between streaming chunks before the stall watchdog fires (throws a retryable error). Increase if complex agentic responses are hitting the limit. |
+| `PROTO_SYSTEM_DEFAULTS_PATH`    | —        | Override path to the system defaults settings file                                                                                                               |
+| `PROTO_SYSTEM_SETTINGS_PATH`    | —        | Override path to the system settings override file                                                                                                               |
 
 ### `modelProviders`
 

packages/core/src/core/turn.ts

@@ -37,13 +37,13 @@ import { withChunkTimeout, StreamStallError } from '../utils/streamStall.js';
 
 /**
  * Max ms to wait between individual stream chunks before declaring a stall.
- * 90 s gives complex agentic responses (e.g. parallel subagent dispatch)
- * enough room for Gemini's mid-stream planning pauses while still catching
- * genuine frozen connections well before the 120 s SDK timeout.
+ * 5 min gives complex agentic responses (e.g. parallel subagent dispatch
+ * and long mid-stream planning pauses) ample headroom while still catching
+ * genuine frozen connections.
  * Override via PROTO_STREAM_STALL_TIMEOUT_MS env var.
  */
 const STREAM_STALL_TIMEOUT_MS = parseInt(
-  process.env['PROTO_STREAM_STALL_TIMEOUT_MS'] ?? '90000',
+  process.env['PROTO_STREAM_STALL_TIMEOUT_MS'] ?? '300000',
   10,
 );