Skip to content

User Manual

  • Get running

    Start the stack, verify health, and create your first corpus.

  • Index a corpus

    Turn a folder (repo, docs, or subtree) into searchable chunks + embeddings + graph context.

  • Search & answer

    Run fused tri-brid retrieval (vector + sparse + graph) and optionally generate answers as one step in the broader MLOps workflow.

  • Use the UI

    Learn the Dashboard / RAG / Eval / Admin tabs and how they map to the API.

  • Tune safely

    Understand the “Pydantic is the law” contract and adjust config without drifting.

Quickstart UI tour Configuration API

Naming: ragweld vs TriBridRAG

This repo/product is called ragweld. Many internal names still say tribrid (config keys, module names, older docs). Treat tribrid as stable internal naming; don’t mass-rename it.

What ragweld does (in one paragraph)

You point ragweld at a corpus (a folder on disk). It indexes that corpus into chunks, embeddings, and (optionally) a graph. Then when you ask a question, ragweld runs three retrieval “legs” in parallel (vector, sparse, graph), fuses the results, can rerank, and returns matches (and optionally an answer). Around that retrieval core, the platform adds synthetic data generation, eval drilldowns, training studios, tracing, and routing controls.

The mental model

flowchart TB
  subgraph Indexing
    A["Folder on disk"] --> B["Load → Chunk → Embed"]
    B --> C["Postgres: chunks + vectors + FTS"]
    B --> D["Neo4j: graph context (optional)"]
  end

  subgraph Retrieval
    Q["Query"] --> V["Vector"]
    Q --> S["Sparse"]
    Q --> G["Graph"]
    V --> F["Fusion"]
    S --> F
    G --> F
    F --> R["Rerank (optional)"]
    R --> O["Matches / Answer"]
    F --> O
  end

Reading paths (pick one)

  1. Follow the Quickstart.
  2. Then skim Searching & answering so you know which knobs matter.
  1. Read the high-level Architecture.
  2. Skim Retrieval overview to understand the tri-brid approach.
  3. Use the UI tour to see what’s actually operable.
  1. Deployment
  2. Operations & metrics
  3. Troubleshooting

What you’ll use most

  • UI: http://127.0.0.1:5173/web/dashboard (default dev URL)
  • API base: http://127.0.0.1:8012/api (default dev URL)
  • Source of truth: server/models/tribrid_config_model.py (Pydantic config + API shapes)