opencode+Ollama 配置本地模型，使用时opencode生成的prompt特别大，导致本地模型卡死 #17304

Question

我使用opencode+Ollama 配置本地模型
，在配置文件opencode.json中增加本地模型配置项：
"ollama": {
"name": "Ollama本地模型",
"npm": "@ai-sdk/openai-compatible",
"models": {
"kamekichi128/qwen3-4b-instruct-2507:latest": {
"name": "Qwen3-4B-Instruct (kamekichi128)",
"parameters": {"temperature": 0.2}
},
"qwen2.5-coder:3b": {
"name": "qwen2.5-coder:3b",
"promptTemplate": "{{content}}" ,
"parameters": {"temperature": 0.2}
}
},
"options": {"baseURL": "http://localhost:11434/v1","systemPrompt": "","promptTemplate": "{{content}}"}
}
使用时启动Ollama服务，在Ollama客户端验证服务可用，并且反应速度很快。而在opencode中加载本地模型qwen2.5-coder:3b使用时，对话仅仅是输入“你好”，都需要40S时间才有回应，
Ollama日志显示prompt=9641，太大了，导致模型处理速度太慢。
Ollama日志：
llama_context: n_ctx_seq (2048) < n_ctx_train (32768) -- the full capacity of the model will not be utilized
llama_context: CPU output buffer size = 0.59 MiB
llama_kv_cache: CPU KV buffer size = 72.00 MiB
llama_kv_cache: size = 72.00 MiB ( 2048 cells, 36 layers, 1/1 seqs), K (f16): 36.00 MiB, V (f16): 36.00 MiB
llama_context: Flash Attention was auto, set to enabled
llama_context: CPU compute buffer size = 300.75 MiB
llama_context: graph nodes = 1231
llama_context: graph splits = 1
time=2026-03-09T09:44:38.187+08:00 level=INFO source=server.go:1388 msg="llama runner started in 1.34 seconds"
time=2026-03-09T09:44:38.188+08:00 level=INFO source=sched.go:567 msg="loaded runners" count=1
time=2026-03-09T09:44:38.188+08:00 level=INFO source=server.go:1350 msg="waiting for llama runner to start responding"
time=2026-03-09T09:44:38.188+08:00 level=INFO source=server.go:1388 msg="llama runner started in 1.35 seconds"
time=2026-03-09T09:44:38.477+08:00 level=WARN source=runner.go:153 msg="truncating input prompt" limit=2048 prompt=9641 keep=4 new=2048
[GIN] 2026/03/09 - 09:44:48 | 200 | 4.2154ms | 127.0.0.1 | GET "/api/tags"
[GIN] 2026/03/09 - 09:45:04 | 500 | 28.0329633s | 127.0.0.1 | POST "/v1/chat/completions"

是什么原因导致prompt那么大？