SSE streaming parse error with GLM-4.7 responses #8282

Description

Using opencode with GLM-4.7 (Z.AI) causes frequent AI_JSONParseError during streaming. The SSE handler incorrectly concatenates chunks, creating malformed JSON.

Error:
AI_JSONParseError: JSON parsing failed: Text: {"id":"20260114043705b7f7de19603a48fd","created":1768336625,"object":"chat.completion.chunk","model":"glm-4.7","choices":[{"index":0,"delta":{"role":"adata: {"id":"20260114043917eff1a1dd435a4df0","created":1768336757,"object":"chat.completion.chunk","model":"glm-4.7","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"Now"}}]}. Error message: JSON Parse error: Expected '}'

Root Cause:
SSE chunks (each prefixed with data: ) are being concatenated instead of parsed separately. The result is {"id":"...data: {"id":"... - an incomplete first JSON object.

Expected:

• Split on data: prefix
• Parse each chunk as separate JSON
• Handle models with reasoning_content

Actual:

• Chunks concatenated without parsing, causing parse failure.

Plugins

No response

OpenCode version

1.1.18

Steps to reproduce

1. Use opencode with GLM-4.7
2. Run prompts that generate longer responses (multiple chunks)
3. Occurs frequently with reasoning tokens

Screenshot and/or share link

No response

Operating System

Ubuntu 22.04.3

Terminal

GNOME Terminal 3.52.0