AI_JSONParseError should support retry for mid-stream parse errors #12595

Description

When using streamText with OpenAI-compatible providers (via @ai-sdk/openai-compatible), the AI SDK may throw AI_JSONParseError during stream consumption when a provider or gateway returns malformed SSE data. Currently, this error:

1. Has no isRetryable property (only APICallError has it)
2. Is never retried by the SDK's built-in retryWithExponentialBackoff mechanism
3. Occurs after the HTTP request succeeds (HTTP 200), so the request-level retry is already bypassed

These errors are typically transient — caused by SSE chunks being concatenated incorrectly at proxy/gateway level, network issues corrupting stream data mid-flight, or providers temporarily returning invalid JSON in stream.

Evidence from Production

We observed this error when using Kimi K2.5 via the Kilo AI Gateway (api.kilo.ai):

{
  "name": "AI_JSONParseError",
  "cause": {},
  "text": "{\"id\":\"chatcmpl-jQugNdata:{\"id\":\"chatcmpl-iU6vkr3fItZ0Y4rTCmIyAnXO\",\"object\":\"chat.completion.chunk\",...}"
}

The data: SSE prefix of the second event was embedded inside the first event's JSON, indicating SSE chunking corruption at the gateway level.

This is a known class of issue — similar reports:

• StreamText error handling still not working properly in v4.0.18 #4099: streamText error handling
• AI_JSONParseError with generateText no tools #8577: AI_JSONParseError with generateText (mentions isRetryable)
• Cloudflare Tunnel + Vercel AI SDK = [AI_JSONParseError]: JSON parsing failed ollama/ollama#5417: Cloudflare Tunnel + Vercel AI SDK = AI_JSONParseError

Current Behavior

The SDK's parsing pipeline (parseJsonEventStream) uses safeParseJSON which returns { success: false, error: JSONParseError } on parse failure. The OpenAI-compatible model handler then emits this as { type: 'error', error: ... } in the stream. There is no way for consumers to retry the stream, skip bad events, or distinguish transient parse errors from permanent ones.

How Other CLI Agents Handle This

OpenAI Codex CLI: Skips unparseable SSE events with continue; and keeps processing. JSON parse errors are logged at debug level. Stream-level errors trigger up to 5 retries with exponential backoff. CodexErr::Json is explicitly marked retryable.

Qwen Code: SDK's parseJsonLineSafe() silently skips failed lines with a warning log.

Proposed Solutions

Option A: Add isRetryable property to AI_JSONParseError for stream context.

Option B: Add onStreamParseError callback to streamText allowing consumers to skip bad events.

Option C: Allow configuring stream-level retry (streamRetries: 3).

Environment

• AI SDK Version: Observed across v4.x and v5.x
• Provider: @ai-sdk/openai-compatible (Kilo AI Gateway)
• Runtime: Bun 1.3.x

Workaround

We detect AI_JSONParseError in our error handler and classify it as retryable with exponential backoff.

See: link-assistant/agent#169