Spaces:
Sleeping
title: Federal FOIA Intelligence Search
emoji: ποΈ
colorFrom: blue
colorTo: indigo
sdk: gradio
sdk_version: 6.3.0
app_file: app.py
pinned: true
license: mit
short_description: 'FOIA DECLASSIFIED DOCUMENTS SEARCH '
ποΈ Federal FOIA Intelligence Search
Public Electronic Reading Rooms Β· Evidence-Grade Citations Β· Responsible AI
Executive Summary
Federal FOIA Intelligence Search is a read-only, transparency-first research application designed to help users discover, organize, cite, and export publicly released U.S. government records from official FOIA Electronic Reading Rooms.
This application is intentionally built to meet:
- π° Investigative journalism standards
- βοΈ Legal evidentiary and citation expectations
- π§ Responsible AI principles
- π‘οΈ Hugging Face Trust & Safety requirements
At all times, the system prioritizes public accountability, verifiability, and user control over automation or opaque data processing.
Core Principles
- Public Records Only
- Link-Out, Not Scraping
- User-Initiated Actions Only
- Explicit AI Opt-In
- Court-Ready Outputs
- Zero Background Data Collection
Supported Federal Agencies (Link-Out Only)
The app generates official search links to the following FOIA Electronic Reading Rooms:
- CIA β Central Intelligence Agency
- FBI β Federal Bureau of Investigation
- DOJ β Department of Justice
- DHS β Department of Homeland Security
- U.S. Department of State
- GSA β General Services Administration
- NSA β National Security Agency
β οΈ Important:
This application does not scrape, crawl, mirror, or store documents from any agency.
Each result redirects the user to the agencyβs official FOIA website.
How the App Works (Detailed)
Step 1 β User Search
The user enters a search term in the unified FOIA search bar.
Step 2 β Federated Link Generation
For each supported agency, the app:
- Generates an agency-specific FOIA search URL
- Measures request latency (for transparency)
- Produces structured metadata (agency, URL, timestamp)
Step 3 β Results Presentation
Results are displayed in two formats:
- A structured data table
- A polished card-based results gallery
No documents are downloaded automatically.
Step 4 β Optional Actions
Users may then:
- View or download documents from the agency website
- Generate citations, appendices, and exports
- Explicitly opt-in to AI analysis
Feature Overview
π Polished FOIA Search Interface
- Unified search bar across agencies
- Clear agency attribution
- Latency badges for transparency
- No hidden background requests
π Exhibit-Aware Bluebook Citations
Each FOIA result is assigned:
- A cryptographic citation hash
- A sequential Exhibit ID (A-1, A-2, etc.)
- A Bluebook-formatted citation
Designed for:
- Court filings
- Academic papers
- Investigative reporting
- FOIA compliance audits
βοΈ Litigation Appendix & Table of Authorities (PDF)
Generates a court-ready PDF containing:
- Title page
- Generation timestamp
- Sequential exhibit list
- Full Bluebook citations
- Implicit Table of Authorities
π The appendix contains no AI-generated facts β citations only.
π FOIA Request Generator
Provides ready-to-edit FOIA request drafts that:
- Reference the selected agency
- Include the userβs topic
- Follow standard FOIA request conventions
β οΈ Templates must be reviewed and customized before submission.
π§ AI Analysis (Strictly Opt-In)
AI features are disabled by default.
To enable AI, the user must explicitly:
- Enable AI analysis
- (Optionally) allow PDF text extraction
AI analysis includes:
- Source attribution
- Context boundaries
- Mandatory disclosure block
- Cryptographic integrity hash
π AI Citation Cross-Checking
When AI is enabled:
- All analysis is tied to a selected FOIA record
- Unsupported claims are prevented
- Missing citation context is flagged
This reduces hallucinations and misuse.
π Analytical Visualizations
Available visual tools:
- Agency Coverage Heatmap
- Domain / Entity Frequency Graph
- Timeline of FOIA publication dates
These operate only on metadata, not document contents.
π Journalist & Reviewer Export Bundles
One-click ZIP export containing:
citations.txtβ Bluebook citationslinks.csvβ structured dataset- Shareable evidence bundle
π€ Share Pages (Metadata Only)
Users may generate a temporary Share ID that:
- Preserves citation metadata
- Never hosts documents
- Exists only in memory
AI Disclosure & Integrity
Every AI output includes:
- Explicit disclosure notice
- User-initiated confirmation
- Integrity hash for auditability
AI outputs are not evidence and not legal advice.
Privacy & Security
- No accounts
- No cookies
- No analytics
- No persistent storage
- In-memory session state only
Warnings & Disclaimers β οΈ
Not Legal Advice
This application does not provide legal advice.
Not Evidence
AI summaries are not admissible evidence.
Public Records Only
The app cannot access unreleased, classified, or restricted materials.
No Government Affiliation
This project is independent and not affiliated with any U.S. government entity.
Tips & Best Practices
- β Always verify citations at the source
- β Use appendices for formal submissions
- β Treat AI as a research assistant, not an authority
- β Do not submit AI text as sworn testimony
- β Do not assume FOIA completeness
Known Limitations
- No full-text indexing
- PDF extraction may fail on scanned documents
- Latency reflects link generation only
Considerations for Responsible Use
- FOIA records may be incomplete or redacted
- Agencies publish at different cadences
- Absence of evidence is not evidence of absence
Hugging Face Trust & Safety Alignment
This Space:
- Does not scrape or crawl
- Does not bypass authentication
- Does not host sensitive data
- Does not train on user input
- Does not automate surveillance
All AI functionality is:
- User-controlled
- Transparent
- Auditable
Future & Planned Expansions (Non-Operational)
Potential future enhancements may include:
- Additional federal agencies
- State-level FOIA reading rooms
- Advanced citation validation
- Redaction comparison tools
- FOIA release trend analysis
- Multi-appendix case bundling
All future features will maintain:
- Public-only access
- Explicit user consent
- Responsible AI design
Final Note
This application was built with a single guiding principle:
Transparency without automation abuse. Evidence without distortion. AI only when invited.
ποΈ Public Records. Verified Sources. Responsible Intelligence.