SRKB — Scientific Research Knowledge Base

Papers, claims, and drafts
in one workspace.

Upload literature into per-project libraries, search with embeddings, chat with citations, and write LaTeX with PDF builds and an in-editor assistant — all scoped to your tenant and team roles.

Team workspaces · email verification on first sign-in · usage and audit built in

Project libraries

Upload PDFs into S3-backed storage, parse and chunk in background jobs, and keep one canonical library per research project.

Grounded chat

Ask questions over your library with retrieval and citations — answers stay tied to the passages they came from.

Semantic search

pgvector search across titles, overviews, extracted claims, benchmarks, datasets, and metrics — always scoped by project.

LaTeX writing IDE

Edit multi-file TeX projects, build PDFs on the worker, preview side-by-side, and iterate with a writing-focused assistant.

Structured knowledge

Claims and related objects are stored as first-class rows with provenance, so downstream agents and humans share the same ground truth.

Bring your own models

Configure provider keys per workspace, enable skills for agents, and enforce per-user quotas with audit trails.

Teams and access

Multi-tenant isolation, owner-managed accounts, email verification, per-project roles, and settings for API keys and usage.

Safety-minded defaults

Row-level tenancy, signed uploads, job isolation on workers, and audit events for sensitive operations.

Community

Research packages and provider rankings.

Browse public project summaries and evidence-backed provider recommendations before signing in.

Open Community

Project packages

login to download

Provider ranking

recommended

From PDF to answer — without blocking the API

Long-running steps run as jobs on the worker tier. The web API stays responsive while parsing, embeddings, builds, and agent turns complete in the background.

STEP 01

Ingest

Upload PDFs (and zip bundles for writing projects). Workers extract text, write chunks and metadata to Postgres, and queue embedding updates.

STEP 02

Explore

Search and chat stay project-scoped. Citations point at stored chunks so you can verify claims against the original paper.

STEP 03

Write & ship

Use the LaTeX IDE for manuscripts, trigger PDF builds, review AI-proposed edits, and export zips when you are ready to submit.

Stop re-reading papers. Start querying them.

Create a tenant, invite collaborators, connect model providers, and keep every artifact under one roof — from literature review to camera-ready PDFs.

Create my workspace