以编程方式使用文档

Contextual AI 提供专为准确可靠的企业 AI 应用设计的最新 RAG 组件。我们的 LangChain 集成为我们的专业模型提供独立的 API 端点：

  • - Grounding 语言模型 (GLM)：世界上最可靠的 grounding 语言模型，通过优先考虑对检索知识的忠实度来最大程度减少幻觉。GLM 提供卓越的事实准确性，并带有内联归属，使其成为可靠性和准确性至关重要的企业 RAG 和代理应用的理想选择。
  • - 指令跟随重排器：首个遵循自定义指令的重排器，能够根据特定标准（如时效性、来源或文档类型）智能地优先排序文档。我们的重排器在行业基准测试中优于竞争对手，解决了企业知识库中冲突信息带来的挑战。

Contextual AI 由 RAG 技术的发明者创立，其专业组件帮助创新团队加速开发生成级 RAG 代理，这些代理能够提供异常准确的响应。

Grounding 语言模型 (GLM)

Grounding 语言模型 (GLM) 专为减少企业 RAG 和代理应用中的幻觉而设计。GLM 提供：

  • - 在 FACTS 基准测试中达到 88% 事实准确性的卓越性能 (查看基准测试结果)
  • - 严格基于提供的知识来源的响应，并带有内联归属 (阅读产品详情)
  • - 直接集成在生成响应中的精确来源引用
  • - 优先考虑检索上下文而非参数知识 (查看技术概述)
  • - 在信息不可用时清楚表达不确定性

GLM 可作为 RAG 管道中通用 LLM 的直接替代品，显著提高关键任务企业应用的可靠性。

指令跟随重排器

全球首个指令跟随重排器以前所未有的控制和准确性革新文档排序。关键能力包括：

  • - 自然语言指令，可根据时效性、来源、元数据等优先排序文档 (查看工作原理)
  • - 在 BEIR 基准测试中得分 61.2，显著优于竞争对手 (查看基准测试数据)
  • - 智能解决来自多个知识来源的冲突信息
  • - 作为现有重排器的直接替代品，无缝集成
  • - 通过自然语言命令动态控制文档排序

重排器在处理可能存在矛盾信息的企业知识库方面表现出色，允许您精确指定在各种场景中哪些来源应优先考虑。

在 LangChain 中使用 Contextual AI

详情请参阅 Contextual 聊天集成文档.

此集成允许您轻松将 Contextual AI 的 GLM 和指令跟随重排器整合到您的 LangChain 工作流中。GLM 确保您的应用提供严格基于事实的响应，而重排器通过智能优先排序最相关文档来显著提高检索质量。

无论您是为受监管行业还是安全敏感环境构建应用，Contextual AI 都能提供您的企业用例所需的准确性、控制力和可靠性。

立即开始免费试用，体验最扎实的语言模型和指令遵循重排序器，为企业 AI 应用赋能。

扎根式语言模型

# Integrating the Grounded Language Model



from langchain_contextual import ChatContextual

# Set credentials
if not os.getenv("CONTEXTUAL_AI_API_KEY"):
    os.environ["CONTEXTUAL_AI_API_KEY"] = getpass.getpass(
        "Enter your Contextual API key: "
    )

# initialize Contextual llm
llm = ChatContextual(
    model="v1",
    api_key="",
)
# include a system prompt (optional)
system_prompt = "You are a helpful assistant that uses all of the provided knowledge to answer the user's query to the best of your ability."

# provide your own knowledge from your knowledge-base here in an array of string
knowledge = [
    "There are 2 types of dogs in the world: good dogs and best dogs.",
    "There are 2 types of cats in the world: good cats and best cats.",
]

# create your message
messages = [
    ("human", "What type of cats are there in the world and what are the types?"),
]

# invoke the GLM by providing the knowledge strings, optional system prompt
# if you want to turn off the GLM's commentary, pass True to the `avoid_commentary` argument
ai_msg = llm.invoke(
    messages, knowledge=knowledge, system_prompt=system_prompt, avoid_commentary=True
)

print(ai_msg.content)
According to the information available, there are two types of cats in the world:

1. Good cats
2. Best cats

指令遵循重排序器

from langchain_contextual import ContextualRerank

if not os.getenv("CONTEXTUAL_AI_API_KEY"):
    os.environ["CONTEXTUAL_AI_API_KEY"] = getpass.getpass(
        "Enter your Contextual API key: "
    )


api_key = ""
model = "ctxl-rerank-en-v1-instruct"

compressor = ContextualRerank(
    model=model,
    api_key=api_key,
)

from langchain_core.documents import Document

query = "What is the current enterprise pricing for the RTX 5090 GPU for bulk orders?"
instruction = "Prioritize internal sales documents over market analysis reports. More recent documents should be weighted higher. Enterprise portal content supersedes distributor communications."

document_contents = [
    "Following detailed cost analysis and market research, we have implemented the following changes: AI training clusters will see a 15% uplift in raw compute performance, enterprise support packages are being restructured, and bulk procurement programs (100+ units) for the RTX 5090 Enterprise series will operate on a $2,899 baseline.",
    "Enterprise pricing for the RTX 5090 GPU bulk orders (100+ units) is currently set at $3,100-$3,300 per unit. This pricing for RTX 5090 enterprise bulk orders has been confirmed across all major distribution channels.",
    "RTX 5090 Enterprise GPU requires 450W TDP and 20% cooling overhead.",
]

metadata = [
    {
        "Date": "January 15, 2025",
        "Source": "NVIDIA Enterprise Sales Portal",
        "Classification": "Internal Use Only",
    },
    {"Date": "11/30/2023", "Source": "TechAnalytics Research Group"},
    {
        "Date": "January 25, 2025",
        "Source": "NVIDIA Enterprise Sales Portal",
        "Classification": "Internal Use Only",
    },
]

documents = [
    Document(page_content=content, metadata=metadata[i])
    for i, content in enumerate(document_contents)
]
reranked_documents = compressor.compress_documents(
    query=query,
    instruction=instruction,
    documents=documents,
)