多數開源）¶

繁體中文 | 简体中文 | English

同一個 ReAct loop、不同 SDK 的 3 個關鍵差異。對應 SKILL.md Step 3。

TL;DR 對照表¶

部分	Anthropic SDK	OpenAI-compat SDK
Tool schema 包法	`tools=[{name, description, input_schema}]`	`tools=[{"type": "function", "function": {name, description, parameters}}]`
Schema 欄位名	`input_schema`	`parameters`
抓 tool call	`[b for b in resp.content if b.type == "tool_use"]`	`resp.choices[0].message.tool_calls`
Args 格式	`call.input` 已是 dict	`call.function.arguments` 是 JSON string、要 `json.loads(...)`
判完成	`resp.stop_reason == "end_turn"`	`resp.choices[0].finish_reason == "stop"`
Assistant 接回	`messages.append({"role": "assistant", "content": resp.content})`	`messages.append({"role": "assistant", "content": msg.content, "tool_calls": msg.tool_calls})`
Tool result 接回	`messages.append({"role": "user", "content": [{"type": "tool_result", "tool_use_id": call.id, "content": obs}]})`	`messages.append({"role": "tool", "tool_call_id": tc.id, "content": obs})`
Exception class	`anthropic.RateLimitError` / `anthropic.APIConnectionError`	`openai.RateLimitError` / `openai.APIConnectionError`

並排程式碼對照（單輪 tool call）¶

Anthropic¶

import anthropic

client = anthropic.Anthropic()

TOOLS = [{
    "name": "get_weather",
    "description": "Use this when the user asks about current weather.",
    "input_schema": {
        "type": "object",
        "properties": {"city": {"type": "string"}},
        "required": ["city"]
    }
}]

resp = client.messages.create(
    model="claude-haiku-4-5",
    max_tokens=512,
    tools=TOOLS,
    messages=[{"role": "user", "content": "台北現在有下雨嗎？"}]
)

# 抓 tool call
calls = [b for b in resp.content if b.type == "tool_use"]
city = calls[0].input["city"]   # 已是 dict

OpenAI-compat（Ollama）¶

from openai import OpenAI
import json

client = OpenAI(base_url="http://localhost:11434/v1", api_key="ollama")

TOOLS = [{
    "type": "function",
    "function": {
        "name": "get_weather",
        "description": "Use this when the user asks about current weather.",
        "parameters": {       # ← 名字不同
            "type": "object",
            "properties": {"city": {"type": "string"}},
            "required": ["city"]
        }
    }
}]

resp = client.chat.completions.create(
    model="qwen2.5:3b",
    tools=TOOLS,
    messages=[{"role": "user", "content": "台北現在有下雨嗎？"}]
)

# 抓 tool call
tc = resp.choices[0].message.tool_calls[0]
args = json.loads(tc.function.arguments)   # ← 要 json.loads
city = args["city"]

並排 ReAct loop（多輪）¶

Anthropic 完整 loop¶

messages = [{"role": "user", "content": "..."}]
for step in range(5):
    resp = client.messages.create(model=MODEL, max_tokens=1024, tools=TOOLS, messages=messages)
    messages.append({"role": "assistant", "content": resp.content})   # 整個 content list 直接接
    if resp.stop_reason == "end_turn":
        break
    tool_results = []
    for call in [b for b in resp.content if b.type == "tool_use"]:
        obs = TOOL_IMPL[call.name](call.input)
        tool_results.append({"type": "tool_result", "tool_use_id": call.id, "content": obs})
    messages.append({"role": "user", "content": tool_results})   # tool results 包在 user message

OpenAI-compat 完整 loop¶

messages = [{"role": "user", "content": "..."}]
for step in range(5):
    resp = client.chat.completions.create(model=MODEL, tools=TOOLS, messages=messages)
    msg = resp.choices[0].message
    messages.append({"role": "assistant", "content": msg.content, "tool_calls": msg.tool_calls})
    if not msg.tool_calls:
        break
    for tc in msg.tool_calls:
        args = json.loads(tc.function.arguments)
        obs = TOOL_IMPL[tc.function.name](args)
        messages.append({"role": "tool", "tool_call_id": tc.id, "content": obs})   # 自己的 role

4 個容易踩錯的地方¶

parameters vs input_schema：複製貼上時最常踩——抄 Anthropic schema 直接給 OpenAI-compat 用會 silent fail（Ollama 不會 raise、就是不呼叫 tool）。
call.input vs json.loads(arguments)：OpenAI-compat 忘記 json.loads 會拿到 string 不是 dict、KeyError。
Tool result 接回 role 不同：Anthropic 用 role="user" + [{"type": "tool_result", ...}]；OpenAI-compat 用 role="tool" + 純 string content。
stop_reason vs finish_reason：兩個都存在、但欄位名跟值不同。Anthropic "end_turn" / "tool_use"；OpenAI "stop" / "tool_calls"。

完整對照範例¶

每個 Stage 3 練習都同時 ship 兩個 starter：

starter.py = OpenAI-compat / Ollama
starter_anthropic.py = Anthropic

對照 ../../stage-3/02-multi-tool-selection/ ~ ../../stage-3/06-schema-design/ 任一個 folder。