Capture • Understand • Connect • Retrieve • Create
Project Athena is an offline-first, privacy-first personal knowledge and AI workspace designed to turn documents, research, notes, and ideas into a connected, searchable body of knowledge.
Knowledge Before Intelligence.
Athena builds a reliable knowledge layer first, then adds retrieval, citations, workspace intelligence, and local AI assistance.
Athena is a long-term software engineering project for building a private local knowledge system that can:
- 📚 Import and organize documents
- 🔎 Search knowledge semantically
- 🧠 Retrieve relevant information with context
- 🔗 Connect related knowledge
- 📝 Preserve citations and references
- 🤖 Integrate local AI providers
- 🖥️ Expose intelligence through a desktop workspace
- 📊 Evaluate retrieval quality and performance
- 🔐 Keep personal research and knowledge local
The goal is not simply to build another chatbot.
The goal is to build personal knowledge infrastructure that can grow for years.
| Principle | Meaning |
|---|---|
| 🧠 Knowledge Before Intelligence | Build reliable knowledge infrastructure before adding more intelligence. |
| 🔌 Integrate Before Innovate | Prefer proven local tools and providers before reinventing them. |
| 🛡️ Zero Regression | Existing functionality must remain stable as the system grows. |
| ❄️ Freeze Before Expansion | Stabilize and validate a milestone before starting the next layer. |
| 🔍 Audit Before Action | Understand repository and architecture state before changing it. |
| 🗺️ Roadmap Discipline | Develop features through explicit engineering milestones. |
| 🔒 Privacy First | Local-first architecture is a core design objective. |
Designed for PDF, DOC/DOCX, TXT, EPUB, images, OCR, structured extraction, chunking, metadata, and source tracking.
Application-level retrieval orchestration with query planning, semantic retrieval, ranking, context assembly, and RAG answers.
A dedicated citation architecture covering citation creation, resolution, presentation, validation, and source traceability.
A knowledge workspace for context-aware queries, connected knowledge, research workflows, and intelligent responses.
Provider-independent integration with local AI tools such as Ollama and LM Studio, plus local embedding models.
┌─────────────────────────┐
│ Athena UI │
│ Knowledge Workspace │
└────────────┬────────────┘
│
┌────────────▼────────────┐
│ Workspace Services │
│ Context • Queries • UX │
└────────────┬────────────┘
│
┌────────────▼────────────┐
│ Retrieval Intelligence │
│ Planner • Search • RAG │
└────────────┬────────────┘
│
┌───────────────────┼───────────────────┐
│ │ │
┌──────▼──────┐ ┌──────▼──────┐ ┌──────▼──────┐
│ Documents │ │ Knowledge / │ │ Citations │
│ Intelligence│ │ Vector Layer│ │ Services │
└──────┬──────┘ └──────┬──────┘ └──────┬──────┘
│ │ │
└───────────────────┼───────────────────┘
│
┌────────────▼────────────┐
│ Local Data │
│ SQLite • Files • Index │
└────────────┬────────────┘
│
┌────────────▼────────────┐
│ Local AI Providers │
│ Ollama • LM Studio • … │
└─────────────────────────┘
| Area | Technologies |
|---|---|
| Language | Python |
| Desktop UI | PySide6 / Qt |
| Database | SQLite |
| Vector / Retrieval | Qdrant and retrieval services |
| Local LLM | Ollama / LM Studio |
| Document Processing | Docling and related tools |
| OCR | RapidOCR / compatible tooling |
| Retrieval | Semantic retrieval, query planning, RAG |
| Evaluation | Benchmark runner, metrics engine, reporting |
| Architecture | Service-oriented, provider-independent, local-first |
A development baseline used:
| Metric | Result |
|---|---|
| Questions | 25 |
| Indexed documents | 78 |
| Chunks | 147 |
| Top-1 Accuracy | 60% |
| Top-3 Accuracy | 88% |
| Top-5 Accuracy | 96% |
| MRR | 0.731 |
| Average latency | 145.10 ms |
These are development benchmark results for a specific corpus and configuration, not universal performance claims.
Athena follows an explicit engineering lifecycle:
Constitution Check
↓
Repository Truth
↓
Sprint Review
↓
Risk Analysis
↓
One-Sprint Execution Plan
↓
Implementation
↓
Quality Gate
↓
Freeze
This helps control architectural drift and prevent feature growth from destabilizing the system.
Athena includes an engineering audit framework for understanding a growing Python codebase.
Capabilities include:
- Source parsing
- Repository scanning
- Module analysis
- Class and function analysis
- Method analysis
- Import analysis
- Dependency graph analysis
- Architecture-oriented reporting
Recorded audit baseline:
Source Files : 392
Modules : 385
Classes : 236
Functions : 197
Methods : 573
Imports : 1247
The goal is to make the system observable, auditable, and maintainable.
Athena follows a staged roadmap rather than uncontrolled feature expansion.
A15 Stabilization
│
▼
A16 AI Provider Platform
│
▼
A17 AI Control Center
│
▼
A18 Document Intelligence
│
▼
A19 Retrieval Intelligence
│
▼
A20 Workspace Intelligence
│
▼
A21 Plugin Ecosystem
│
▼
A22 Assistant Engine
│
▼
A23 Offline AI Worker
Each stage is subject to architecture review and quality gates before expansion.
Athena is evolving from:
Question → Search → Answer
toward:
Knowledge
↓
Context
↓
Retrieval
↓
Workspace
↓
Reasoning
↓
Action
The workspace layer provides the foundation for a future personal assistant without making the assistant the center of the architecture.
Your knowledge should remain yours.
Athena prioritizes:
- Local storage
- Local document processing
- Local embeddings
- Local LLM providers
- Provider abstraction
- Explicit integration boundaries
- Minimal external dependencies
Cloud services can be integrated where useful, but the core architecture is designed to remain meaningful without requiring a cloud AI service.
Athena is not intended to be:
- ❌ Just another chatbot
- ❌ A thin wrapper around an LLM API
- ❌ A collection of disconnected AI experiments
- ❌ A cloud-dependent knowledge service
Athena is being engineered as:
- ✅ Personal knowledge infrastructure
- ✅ Retrieval and evidence system
- ✅ Document intelligence platform
- ✅ Local AI integration layer
- ✅ Workspace intelligence system
- ✅ Foundation for a future personal assistant
Documents + Research + Notes + Data + Conversations
│
▼
┌───────────────┐
│ ATHENA │
├───────────────┤
│ Understand │
│ Retrieve │
│ Connect │
│ Cite │
│ Reason │
│ Assist │
└───────┬───────┘
│
▼
Better Research
Better Decisions
Better Work
The objective is not to make AI replace thinking.
It is to make thinking with your own knowledge dramatically more powerful.
athena/
├── athena_core/
│ ├── retrieval/
│ ├── workspace/
│ ├── citations/
│ ├── documents/
│ ├── providers/
│ └── services/
├── tools/
│ └── aeaf/
├── benchmarks/
├── tests/
├── docs/
├── scripts/
└── README.md
The structure may evolve as the architecture develops.
Python · PySide6 · AI · RAG · Retrieval ·
Knowledge Management · Document Intelligence · Semantic Search ·
Local AI · Ollama · LM Studio · Qdrant · SQLite · Docling ·
OCR · Citations · Workspace Intelligence · Offline AI ·
Privacy-First
Your Private Knowledge. Your Local AI. Your Second Brain.
Knowledge Before Intelligence. Evidence Before Confidence. Architecture Before Features.
Think locally. Learn continuously. Build intelligently.