confluence_search.py

https://github.com/RomainNeup/open-webui-utilities This tool searches for Confluence pages using a given query string with RAG (Retrieval Augmented Generation) capabilities.

• Configuration:
  • Username: [email protected]
  • API Key: 1234ABCD
  • Base URL: https://example.atlassian.net/wiki
  • Ssl Verify: Default to true, disable if you have a self-signed certificate
  • Result Limit: Maximum number of pages to retrieve from Confluence API
  • RAG Settings:
    • embedding_model_save_path: Path to save embedding models
    • embedding_model_name: Name/path of the embedding model to use
    • cpu_only: Run the tool on CPU only (vs GPU)
    • chunk_size: Maximum size of each content chunk for processing
    • chunk_overlap: Overlap between consecutive chunks
    • max_results: Maximum number of relevant chunks to return
    • similarity_threshold: Minimum similarity score to include results
    • ensemble_weighting: Balance between keyword (0) and semantic (1) search
    • enable_hybrid_search: Enable combined semantic and keyword search
    • full_context: Return full document content vs just relevant chunks
    • max_page_size: Maximum size in characters for a Confluence page
    • batch_size: Number of documents to process in each batch
• User Settings:
  • api_key_auth: Use API key authentication (true) or personal access token (false)
  • username: Override default username for authentication
  • api_key: Override default API key or personal access token
  • split_terms: Split search query into words for better search results
  • included_confluence_spaces: Comma-separated list of Confluence spaces to search in (leave empty for all). Be aware that it can conflict with the excluded_confluence_spaces setting.
  • excluded_confluence_spaces: Comma-separated list of Confluence spaces to exclude from search (leave empty for none). Be aware that it can conflict with the included_confluence_spaces setting.
• Usage: Run the script and provide the search query as an argument. The tool will return the most relevant Confluence pages and content sections.