Source → knowledge → answer, one continuous system.
Raw sources flow through a seven-phase pipeline, pass an ingestion gate, and become retained, versioned, traceable knowledge — held in many representations and described by entities, scopes, trust, and lineage. Agents attach anywhere along the way and turn it into cited, accurate answers.
Sources
Tap any phase to focus it. Phases 1–7 flow top to bottom · the agent fabric overlays every stage.
What the system actually does
Every capability below is a real surface you can open today. Follow the link to see it for yourself.
Bring any source
PDFs, scans, audio, video, web pages, native files. They upload into your file system and get an origin lineage stamp the moment they arrive.
FilesA real extraction pipeline
Extract → clean → segment → embed, run per-stage or as one streamed job. Watch the original PDF, raw text, cleaned markdown, and segments side by side in a 4-pane viewer.
LibraryMany representations, one source
The same document is held as text, segments, vectors, summaries, schemas, and indices — each scoped to who's allowed to see it, all tracing back to the original.
Data storesEntities + a knowledge graph
NER lifts entities and concepts out of your sources and connects them. Explore the graph, click a node, and jump to the exact evidence that put it there.
Knowledge graphRetrieval you can trust
Semantic + structural + trust-weighted search in a multi-tab Search Lab. Every hit drills down to its provenance — no black-box answers.
Search LabAgents on every node
Attach an agent anywhere in the pipeline — as a chatbot, a button, a form, an app, or an automation. The same agent reshapes to the job.
AgentsOne question, resolved end-to-end
Attorney → agent
“Depose Dr. Smith tomorrow. Find where he didn't follow AMA guidelines for John Doe, case #123456, and flag prior injuries (preexisting conditions). Settle the case — hit hard, but get every fact right.”
cited · accurate · traceable
The agent resolves it:
- 1Structural filter → case #123456 · Dr. Smith
- 2Semantic search → 'AMA deviations' · 'prior injuries'
- 3Pull canonical → medical reports (JSON) + AMA guide
- 4Verify → trace each fact to its provenance root
- 5Apply skill → 'how to depose a doctor'
From a blank file to a pro-level answer
An on-screen wizard walks you through the whole system using the real UI — not a fake demo. You do each step; the system makes something real happen. Tags show what's live today.
- 1
Upload a PDF and a CSV
PartialDrop both into your files. The PDF feeds the knowledge pipeline; CSV-as-knowledge ingest is still being built.
- 2
Run extraction
LiveSend the PDF through extract → clean → segment → embed and watch it in the 4-pane viewer.
- 3
Build a specialist agent
LiveCreate a custom agent that pulls a specific piece of data out of the document.
- 4
Run, inspect, fix, re-test
LiveRun the agent, see the result, edit the agent to fix issues, run again, finalize.
- 5
Agent battle
LiveRace several models side by side — fast + cheap vs. slow + expensive + better — in the comparison arena.
- 6
Automate the extraction
LiveUploaded files auto-schedule for RAG (≈5 min after arrival). Toggle “Process for RAG immediately” in the New menu to skip the wait, or trigger/refresh any file on demand.
- 7
Upload again, watch it run
LiveDrop a new document and watch its RAG lifecycle move scheduled → running → indexed right on the file — no manual step required.
- 8
Chat, then save a Note
LiveAsk a chat agent a question, then save the answer straight into Notes.
- 9
Edit the note
LiveClean it up and make it better in the notes editor.
- 10
Trigger RAG / NER
LiveProcess the note for RAG, then hit “Run NER now” right in the note toolbar — entities lift out on demand instead of waiting for the backend batch.
- 11
Visualize the graph
LiveOpen the knowledge graph and see entities and relationships light up.
- 12
Manual search
LiveRun a manual retrieval to confirm you can pull good data back out.
- 13
Build your badass agent
LiveCreate a top-tier agent wired to your knowledge base.
- 14
Ask the hard question
LivePose a deep, industry-knowledge question and watch the agent hit your RAG, reason, and answer like a pro — cited and traceable.
The guided wizard that drives you from page to page is on the roadmap. The capabilities it strings together are already real — this page links straight to each one.
See the whole system in motion
Start with a single document and follow it from source to a cited answer.