RSI TECH
Contact
Work index

open-source / lab / ongoing

LiteratureAtlas

An offline-first research atlas turns local PDFs into summaries, embeddings, maps, Q&A, and analytics.

System fingerprint

Select a signal node to inspect its operating principle.

  1. PDF ingestionKeep source material local.
  2. Local intelligenceSummarize and embed on device.
  3. OutputsPersist inspectable local artifacts.
  4. AtlasConnect maps, Q&A, and analytics.

01 / Evidence

Outcome

  • Offline-first PDF ingestion and local summarization and embedding.
  • Interactive maps, Q&A, and analytics over research material.
  • A Python and DuckDB analytics pipeline with optional Rust helpers.

02 / Operating environment

Context

On-device SwiftUI atlas for PDF ingestion, local summaries/embeddings, interactive maps, Q&A, and analytics.

Researchers working with local PDF collections

03 / Boundary conditions

Constraints

  1. C01

    PDF ingestion and generated outputs must remain local.

  2. C02

    The workflow must run without sending data to the cloud.

  3. C03

    Acceleration is optional rather than a requirement for the core workflow.

04 / Decision path

Architecture

Select a signal node to inspect its operating principle.

  1. PDF ingestionKeep source material local.
  2. Local intelligenceSummarize and embed on device.
  3. OutputsPersist inspectable local artifacts.
  4. AtlasConnect maps, Q&A, and analytics.

05 / Engineering choices

Key decisions

  • D1Use a SwiftUI app core with a Python and DuckDB analytics backend.
  • D2Keep clear boundaries from the app through stored outputs and analytics to the UI.
  • D3Use optional Rust helpers for acceleration.

06 / Proof discipline

Evaluation

  • Store outputs locally so the workflow can be inspected without cloud dependencies.
  • Keep the analytics pipeline separate from the application core.

07 / Scope

Ownership

SwiftUI application, local intelligence workflow, analytics, and optional acceleration

  1. O01

    SwiftUI application core

  2. O02

    Local summarization and embedding workflow

  3. O03

    Analytics backend and optional Rust acceleration

08 / Explicit compromises

Trade-offs

  • Prefer local privacy and offline operation over cloud dependencies.
  • Keep Rust acceleration optional so the core workflow remains portable.

Implementation stack

SwiftUI / On-device LLM / Python / DuckDB / Rust (optional)

Next evidence

More work

delivered / Security AnalyticsSecurity Ops: User Behavior Analytics & Anomaly DetectionBehavioral baselines and explainable anomaly signals for SOC triage and investigation.delivered / Data PlatformsAutomated Time-Series Forecasting FrameworkA repeatable forecasting workflow made model comparisons safer and iteration easier to hand over.