Skip to main content
Open In ColabOpen on GitHub

Taiga

这个笔记本提供了快速入门 langchain_taiga 中Taiga工具的概览。有关每个工具和配置的详细信息，请参阅您仓库中的文档字符串或相关文档页面。

概览​

集成细节​

Class包序列化JS支持Package 最新版本
create_entity_tool, search_entities_tool, get_entity_by_ref_tool, update_entity_by_ref_tool , add_comment_by_ref_tool, add_attachment_by_ref_toollangchain-taigaN/ATBDPyPI - Version

Tool features​

  • create_entity_tool: 在Taiga中创建用户故事、任务和问题。
  • search_entities_tool: 在Taiga中搜索用户故事、任务和问题。
  • get_entity_by_ref_tool: 根据参考获取用户故事、任务或问题。
  • update_entity_by_ref_tool: 按照参考更新用户故事、任务或问题。
  • add_comment_by_ref_tool: 添加对用户故事、任务或问题的评论。
  • add_attachment_by_ref_tool: 向用户故事、任务或问题添加一个附件。

设置​

The integration lives in the langchain-taiga package.

%pip install --quiet -U langchain-taiga
/home/henlein/Workspace/PyCharm/langchain/.venv/bin/python: No module named pip
Note: you may need to restart the kernel to use updated packages.

Credentials​

此集成要求您将TAIGA_URL，TAIGA_API_URL，TAIGA_USERNAME，TAIGA_PASSWORD和OPENAI_API_KEY设置为环境变量以与Taiga进行身份验证。

export TAIGA_URL="https://taiga.xyz.org/"
export TAIGA_API_URL="https://taiga.xyz.org/"
export TAIGA_USERNAME="username"
export TAIGA_PASSWORD="pw"
export OPENAI_API_KEY="OPENAI_API_KEY"

这也是可选但推荐的做法，用于获得最佳的可观测性，请设置LangSmith：

# os.environ["LANGSMITH_TRACING"] = "true"
# os.environ["LANGSMITH_API_KEY"] = getpass.getpass()

Instantiation​

以下是一个示例，展示了如何在langchain_taiga中实例化Taiga工具。根据您的具体需求进行相应调整。

from langchain_taiga.tools.discord_read_messages import create_entity_tool
from langchain_taiga.tools.discord_send_messages import search_entities_tool

create_tool = create_entity_tool
search_tool = search_entities_tool

Invocation​

直接调用带参数​

以下是使用字典作为关键字参数调用工具的一个简单示例。

from langchain_taiga.tools.taiga_tools import (
    add_attachment_by_ref_tool,
    add_comment_by_ref_tool,
    create_entity_tool,
    get_entity_by_ref_tool,
    search_entities_tool,
    update_entity_by_ref_tool,
)

response = create_entity_tool.invoke(
    {
        "project_slug": "slug",
        "entity_type": "us",
        "subject": "subject",
        "status": "new",
        "description": "desc",
        "parent_ref": 5,
        "assign_to": "user",
        "due_date": "2022-01-01",
        "tags": ["tag1", "tag2"],
    }
)

response = search_entities_tool.invoke(
    {"project_slug": "slug", "query": "query", "entity_type": "task"}
)

response = get_entity_by_ref_tool.invoke(
    {"entity_type": "user_story", "project_id": 1, "ref": "1"}
)

response = update_entity_by_ref_tool.invoke(
    {"project_slug": "slug", "entity_ref": 555, "entity_type": "us"}
)


response = add_comment_by_ref_tool.invoke(
    {"project_slug": "slug", "entity_ref": 3, "entity_type": "us", "comment": "new"}
)

response = add_attachment_by_ref_tool.invoke(
    {
        "project_slug": "slug",
        "entity_ref": 3,
        "entity_type": "us",
        "attachment_url": "url",
        "content_type": "png",
        "description": "desc",
    }
)

调用带有ToolCall​

如果您生成了一个模型输出ToolCall，请将其按照以下格式传递给tool.invoke()。

# This is usually generated by a model, but we'll create a tool call directly for demo purposes.
model_generated_tool_call = {
    "args": {"project_slug": "slug", "query": "query", "entity_type": "task"},
    "id": "1",
    "name": search_entities_tool.name,
    "type": "tool_call",
}
tool.invoke(model_generated_tool_call)

链式调用​

以下是一个更完整的示例，展示了你如何在LLM中将create_entity_tool和search_entities_tool工具集成到一个链或代理中。此示例假设您有一个函数（如create_react_agent），该函数能够设置一种类似于LangChain风格的代理，在适当的时候调用工具。

# Example: Using Taiga Tools in an Agent

from langgraph.prebuilt import create_react_agent
from langchain_taiga.tools.taiga_tools import create_entity_tool, search_entities_tool

# 1. Instantiate or configure your language model
# (Replace with your actual LLM, e.g., ChatOpenAI(temperature=0))
llm = ...

# 2. Build an agent that has access to these tools
agent_executor = create_react_agent(llm, [create_entity_tool, search_entities_tool])

# 4. Formulate a user query that may invoke one or both tools
example_query = "Please create a new user story with the subject 'subject' in slug project: 'slug'"

# 5. Execute the agent in streaming mode (or however your code is structured)
events = agent_executor.stream(
    {"messages": [("user", example_query)]},
    stream_mode="values",
)

# 6. Print out the model's responses (and any tool outputs) as they arrive
for event in events:
    event["messages"][-1].pretty_print()

API 参考​

请参阅以下位置的文档字符串：

for usage details, parameters, and advanced configurations.