github/langchain
1
0
Fork 0
mirror of https://github.com/hwchase17/langchain.git synced 2025-06-19 13:23:35 +00:00
Code Issues Packages Projects Releases Wiki Activity

x

Browse Source
This commit is contained in:
Eugene Yurtsev 2024-10-11 15:48:21 -04:00
parent 8688c21d2a
commit 1b3e07b7e3
1 changed files with 34 additions and 37 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

71
docs/docs/concepts/index.mdx
View File

@ -12,43 +12,40 @@ In this section, you'll find explanations of the key concepts, providing a deepe
The conceptual guide will not cover step-by-step instructions or specific implementation details — those are found in the [How-To Guides](/docs/how_to/) and [Tutorials](/docs/tutorials) sections. For detailed reference material, please visit the [API Reference](https://python.langchain.com/api_reference/).
| Concept | Description |
|----------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------------|
| [Runnable interface](/docs/concepts/runnables) | A standard interface for creating and invoking custom chains, with methods like invoke, stream, and batch, both sync and async. |
| [LangChain Expression Language (LCEL)](/docs/concepts/lcel) | A declarative way to chain LangChain components with features like streaming, async support, retries, and more. |
| [Chat models](/docs/concepts/chat_models) | Models that process sequences of messages as input and output, with support for roles like 'user', 'assistant', and 'system'. |
| [LLMs](/docs/concepts/llms) | Older or lower-level models that process plain text input and output plain text, often replaced by chat models. |
| [Messages](/docs/concepts/messages) | Different message types representing the roles and content of conversational exchanges in LangChain. |
| [Prompt templates](/docs/concepts/prompts) | Templates that help guide a model's response by formatting user inputs and parameters into prompts. |
| [Output parsers](/docs/concepts/output_parsers) | Components that transform model output into structured formats, useful for LLMs generating structured data. |
| Example selectors | Classes responsible for selecting and formatting examples into prompts to improve model performance. |
| Chat history | A class that stores and manages the history of inputs and outputs in a conversation, keeping track of previous interactions. |
| Documents | An object that contains the text and metadata associated with a piece of information in LangChain. |
| Document loaders | Classes responsible for loading document data from various external sources like Slack, Google Drive, or Notion. |
| [Text splitters](/docs/concepts/text_splitters) | Tools for splitting text into smaller, semantically meaningful chunks to fit into model context windows. |
| [Embedding models](/docs/concepts/embedding_models) | Create vector representations of text for similarity search and retrieval in natural language tasks. |
| [Vector stores](/docs/concepts/vectorstores) | Tools for storing and searching embedded data, allowing you to perform vector searches based on similarity. |
| [Retrievers](/docs/concepts/retrievers) | Interfaces that return relevant documents based on an unstructured query, more general than vector stores. |
| Key-value stores | Storage mechanism for key-value pairs, helpful for caching embeddings or storing multiple vectors per document. |
| Tools | Utilities designed to be invoked by models, allowing models to interact with code or external APIs. |
| Toolkits | Collections of tools designed for specific tasks, often with convenient loading methods. |
| Agents | Systems that use LLMs as reasoning engines to choose actions and determine inputs, iterating until task completion. |
| Callbacks | LangChain's system for logging and tracking various stages of execution, including models, chains, and tools. |
| Streaming | Allows consuming partial output as it is generated, helping to reduce latency in complex chains and agents. |
| [Function/tool calling](/docs/concepts/#function-tool-calling) | Allows a model to generate output arguments that invoke external tools or functions for more complex tasks. |
| Structured output | Constrains a model's output to a specific format, such as JSON, to improve the usability of generated responses. |
| [Few-shot prompting](/docs/concepts/#few-shot-prompting) | A prompting technique where example inputs and outputs are added to a model prompt to improve performance. |
| Retrieval | The process of providing relevant data to an LLM at query time to improve its response to the user. |
| Query Translation | Techniques for improving retrieval accuracy by refining or altering the original query. |
| Routing | Methods for routing queries to appropriate data sources based on content or context. |
| Query Construction | Techniques for transforming natural language queries into queries specific to the data source's format, such as SQL. |
| Indexing | Creating and storing documents for efficient search and retrieval, often by using embedding models and vector stores. |
| Post-processing | Techniques for filtering or ranking retrieved documents to improve the quality of results passed to the LLM. |
| Generation | Methods for self-correcting errors in responses, such as hallucinations, by iterating on the answer. |
| Text splitting | Mechanisms for splitting text into smaller chunks, either by character, sentence, or semantically related units. |
| Evaluation | The process of assessing the quality and performance of an LLM's responses in an application. |
| Tracing | A system for tracking and observing the sequence of operations in a LangChain application to help diagnose issues. |
- [Runnable interface](/docs/concepts/runnables): A standard interface for creating and invoking custom chains, with methods like invoke, stream, and batch, both sync and async.
- [LangChain Expression Language (LCEL)](/docs/concepts/lcel): A declarative way to chain LangChain components with features like streaming, async support, retries, and more.
- [Chat models](/docs/concepts/chat_models): Models that process sequences of messages as input and output, with support for roles like 'user', 'assistant', and 'system'.
- [LLMs](/docs/concepts/llms): Older or lower-level models that process plain text input and output plain text, often replaced by chat models.
- [Messages](/docs/concepts/messages): Different message types representing the roles and content of conversational exchanges in LangChain.
- [Prompt templates](/docs/concepts/prompts): Templates that help guide a model's response by formatting user inputs and parameters into prompts.
- [Output parsers](/docs/concepts/output_parsers): Components that transform model output into structured formats, useful for LLMs generating structured data.
- Example selectors: Classes responsible for selecting and formatting examples into prompts to improve model performance.
- Chat history: A class that stores and manages the history of inputs and outputs in a conversation, keeping track of previous interactions.
- Documents: An object that contains the text and metadata associated with a piece of information in LangChain.
- Document loaders: Classes responsible for loading document data from various external sources like Slack, Google Drive, or Notion.
- [Text splitters](/docs/concepts/text_splitters): Tools for splitting text into smaller, semantically meaningful chunks to fit into model context windows.
- [Embedding models](/docs/concepts/embedding_models): Create vector representations of text for similarity search and retrieval in natural language tasks.
- [Vector stores](/docs/concepts/vectorstores): Tools for storing and searching embedded data, allowing you to perform vector searches based on similarity.
- [Retrievers](/docs/concepts/retrievers): Interfaces that return relevant documents based on an unstructured query, more general than vector stores.
- Key-value stores: Storage mechanism for key-value pairs, helpful for caching embeddings or storing multiple vectors per document.
- Tools: Utilities designed to be invoked by models, allowing models to interact with code or external APIs.
- Toolkits: Collections of tools designed for specific tasks, often with convenient loading methods.
- Agents: Systems that use LLMs as reasoning engines to choose actions and determine inputs, iterating until task completion.
- Callbacks: LangChain's system for logging and tracking various stages of execution, including models, chains, and tools.
- Streaming: Allows consuming partial output as it is generated, helping to reduce latency in complex chains and agents.
- [Function/tool calling](/docs/concepts/#function-tool-calling): Allows a model to generate output arguments that invoke external tools or functions for more complex tasks.
- Structured output: Constrains a model's output to a specific format, such as JSON, to improve the usability of generated responses.
- [Few-shot prompting](/docs/concepts/#few-shot-prompting): A prompting technique where example inputs and outputs are added to a model prompt to improve performance.
- Retrieval: The process of providing relevant data to an LLM at query time to improve its response to the user.
- Query Translation: Techniques for improving retrieval accuracy by refining or altering the original query.
- Routing: Methods for routing queries to appropriate data sources based on content or context.
- Query Construction: Techniques for transforming natural language queries into queries specific to the data source's format, such as SQL.
- Indexing: Creating and storing documents for efficient search and retrieval, often by using embedding models and vector stores.
- Post-processing: Techniques for filtering or ranking retrieved documents to improve the quality of results passed to the LLM.
- Generation: Methods for self-correcting errors in responses, such as hallucinations, by iterating on the answer.
- Text splitting: Mechanisms for splitting text into smaller chunks, either by character, sentence, or semantically related units.
- Evaluation: The process of assessing the quality and performance of an LLM's responses in an application.
- Tracing: A system for tracking and observing the sequence of operations in a LangChain application to help diagnose issues.