RagSolux

RagSolux is a desktop app that ingests CSVs and other structured files, builds a vector index, and lets you query the content through an LLM chat interface. It’s designed for developers and analysts who want repeatable, configurable retrieval-augmented Q&A over tabular data. The workflow keeps processing transparent with logs, progress controls, and tunable retrieval/generation settings.

Key features

• Multi-format ingestion — Import common tabular and document formats for indexing and chat-based querying.
• Embedding and indexing — Splits data into chunks/rows, generates embeddings, and stores them in a vector index.
• LLM and embedding selection — Configure which chat model and embedding model the app uses.
• Chat over your data — Ask natural-language questions and get answers grounded in retrieved segments.
• Retrieval tuning — Adjust diversity (MMR-style search) and other knobs to balance coverage vs. precision.
• Prompt controls — Use optional prompt overrides with validation to keep outputs consistent.
• Persistent conversation history — Keep prior context available across sessions.
• Desktop-focused UX — Dark theme, status panel, and progress/cancel controls for long processing runs.
• Processing visibility — Console-style logs show parsing, embedding, and indexing progress.

Supported technologies

• Desktop platforms

  • Windows
  • macOS
  • Linux

• Supported file types

  • CSV
  • XLS / XLSX
  • JSON
  • HTML
  • XML

• RAG stack

  • LangChain (ChatOpenAI interface)
  • FAISS vector store

• Retrieval and generation

  • MMR-style retrieval with configurable parameters
  • Fine-grained generation controls (e.g., verbosity, reasoning effort)

Use cases

• Ask questions about large CSV exports (sales, logs, telemetry) without writing custom queries.
• Explore spreadsheet-based datasets with conversational filtering and summaries.
• Build a quick “local knowledge base” from structured files for internal analysis.
• Compare answers under different retrieval settings to validate consistency.
• Share a repeatable QA workflow with teammates using the same configuration.

Notes

• Requires an API key for the configured LLM/embedding service; a custom base URL can be set if needed.
• Indexing progress and status are visible during ingestion, with cancel support for long jobs.