Hybrid RAG SEC AI

Live pipeline control room for AI-powered SEC filing retrieval and answer generation.

Runtime profile

Two-panel dashboard for demos, debugging, and technical presentations. The UI listens to live pipeline events and renders the final grounded answer from the production backend.

Prompt panel

Run a live filing query

Send a prompt to the production RAG backend and watch each pipeline stage update in real time.

Ask questions about SEC filings and financial reports.

  • What legal risks did Apple mention in its 10-K filings?
  • Summarize risk factors in NVIDIA's annual report.
  • What litigation risks appear in the filings?

Transport

SSE for live execution + POST for the final answer.

Suggested questions

Question bank

Query history

Previous runs

0 stored
No queries yet. Submit a prompt to create a replayable session history.

Pipeline view

Execution map

stream idleready
01

Prompt

Incoming user request enters the runtime graph.

02

Embedding

Query gets transformed into vector space for retrieval.

03

Hybrid Retrieval

Qdrant and BM25 search run and merge their candidates.

04

Rerank

CrossEncoder rescoring refines the highest-value chunks.

05

Context Build

Top grounded excerpts are formatted for answer generation.

06

LLM

DeepSeek generates the final grounded response.

07

Answer

Final answer and sources are returned to the interface.

Execution log

Live pipeline trace

IdleStandby
Start a query to watch backend events arrive here in real time.

Final answer

Grounded response

stream idle
The final answer will appear here once the backend completes the run.