mirror of
https://github.com/zhukunpenglinyutong/jetbrains-cc-gui.git
synced 2026-11-04 15:43:49 +08:00
57ggfk
e2e2264d76
OpenCode 1.18 run --format json pushes each part only once with terminal state: text/reasoning arrive as complete parts and the tool part arrives as a single completed event. The parser assumed a two-phase started->completed tool flow, so it emitted orphan tool_results that never rendered, and the missing --thinking flag made the CLI suppress reasoning parts entirely. - classify events by part.type (text/reasoning/tool), keeping legacy top-level type matching as fallback - synthesize the tool_use + tool_result pair from the single terminal tool event, keyed on part.callID; started-dedup keeps future two-phase pushes converging on one card - pass the plugin thinking toggle down to --thinking (Java stdin -> channel -> argv); bypassPermissions maps to --auto - emit only the novel suffix when a part is re-pushed with cumulative text, so text never doubles in the UI - capture real 1.18.32 event samples and replay them as test fixtures The streaming toggle stays a documented no-op for OpenCode until a serve-based integration lands (see docs/plans plan for the matrix).
40 lines
2.2 KiB
Markdown
40 lines
2.2 KiB
Markdown
# OpenCode 1.18.32 `run --format json` 原始事件采样
|
||
|
||
> 采集日期：2026-09-29 · opencode CLI 1.18.32（Windows, npm 安装）
|
||
> 模型：`opencode/mimo-v2.6-flash-free`（免费档，仅用于复现事件形态）
|
||
> 用途：`2026-09-29-opencode-process-display-plan.md` 的根因证据；可改造为
|
||
> `ai-bridge/services/opencode/message-service.test.js` 的回放 fixtures。
|
||
|
||
## 采集命令
|
||
|
||
```bash
|
||
# 工具调用回合（bash 快速执行）
|
||
opencode run --format json --model <model> "Use the bash tool to run: echo hello-from-opencode. Then reply briefly."
|
||
# → tool-call-turn.jsonl
|
||
|
||
# 工具调用回合（bash 含 sleep 6，验证是否存在 pending/running 中间事件）
|
||
opencode run --format json --model <model> "Use the bash tool to run: sleep 6 && echo done-sleeping. Then reply briefly."
|
||
# → slow-tool-turn.jsonl（答案：没有中间事件，仍只有单个 completed）
|
||
|
||
# 纯文本长回复（200 词故事，验证流式粒度）
|
||
opencode run --format json --model <model> "Write a 200-word story about a robot."
|
||
# → long-text-turn.jsonl（答案：整个 text part 结束后才输出一条完整文本）
|
||
|
||
# 思考回合，带 --thinking
|
||
opencode run --format json --thinking --model <model> "Think step by step: what is 17*23? ..."
|
||
# → thinking-turn-with-flag.jsonl（出现 {"type":"reasoning","part":{"type":"reasoning",...}}）
|
||
|
||
# 思考回合，不带 --thinking（对照组）
|
||
opencode run --format json --model <model> "Think step by step: what is 17*23? ..."
|
||
# → thinking-turn-without-flag.jsonl（0 条 reasoning 事件；step_finish.tokens.reasoning 仍有值）
|
||
```
|
||
|
||
## 关键结论
|
||
|
||
1. **无流式**：text/reasoning part 仅在完成时（`part.time.end` 已设置）输出**一次完整内容**；
|
||
tool part 仅在 `state.status ∈ {completed, error}` 时输出。无任何 delta 事件。
|
||
2. **`--thinking` 默认关闭**（非交互 run 模式）：reasoning part 完全不输出，加 flag 才有。
|
||
3. **工具是单事件**：`{"type":"tool_use","part":{"type":"tool","tool":"bash","callID":"call_…",
|
||
"state":{"status":"completed","input":{…},"output":"…"}}}` —— 没有 started→completed 两段式。
|
||
4. 事件骨架：`step_start → (reasoning|text|tool_use)* → step_finish`，每步循环一次。
|