Langchain Logseq

Collection of Langchain utilities for working with Logseq files.

Components

This section provides an overview of the components provided, listed by type

Retrievers

Retrievers inject context into a conversation. Works in tandem with a Contextualizer and Document Loader.

• Input:
  • natural-language user-input, usually query-like
  • (optional) chat history
• Output:
  • list of Documents to provide context for an LLM to answer the user-input

Implementations

• LogseqJournalDateRangeRetriever
  • retrieve Logseq journal Documents, intended for queries that require context from a date range
  • required to set up:
    • RetrieverContextualizer
    • LogseqJournalLoader
  • examples:
    • "What did I do over Christmas break 2024?"
    • "How did I spend the last Independence Day?"

Contextualizers

Contextualizers serve as the bridge between natural-language input and a downstream component that handles fetching of relevant Documents.

• Input:
  • natural-language user-input, usually query-like
  • (optional) chat history
• Output:
  • structured downstream query, based on

In this library, an instance of RetrieverContextualizer is provided directly to Retrievers during the latter's instantiation. To set up the RetrieverContextualizer, provide RetrieverContextualizerProps, which includes:

• llm - this is the backbone of the contextualizer
• prompt - instructions provided to the LLM
• output_schema - (optional) structured schema used to fetch relevant Documents
  • if no schema is provided, a string shall be returned instead
• other flags and settings

Loaders

Loaders are one type of component that can fetch relevant Documents. Loaders are typically specific to a corresponding Retriever component.

• Input:
  • each loader specifies its own schema
    • the Contextualizer is usually responsible for creating an instance of the query obj to act upon
• Output:
  • list[Document]

Implementations

• LogseqJournalFilesystemLoader
  • loads from the filesystem, where journal files are expected to be present at specified path

Scripts

PGVector

upload_journal

usage: python scripts/upload_journal_to_pgvector.py [-h] [-p PATH] from_date to_date