Knowledge Intelligence Platform (KIP) — Agentic Multimodal Research Platform (v2.0)

A production-grade, open-source Agentic Multimodal Research & Knowledge Intelligence Platform engineered for deep codebase analysis, GitHub repository ingestion, local folder indexing, visual diagram intelligence, and grounded technical research synthesis with exact citations.

---

🌟 Key Capabilities (v2.0 Enterprise)

• 🤖 Autonomous Deep Research Agent: Decomposes complex engineering questions into multi-hop sub-goals, explores AST code symbols, inspects file slices, and synthesizes grounded research dossiers.
• 🐙 GitHub Repository Ingestion: Shallow-clone and REST API tree ingestion for public and private repositories with branch/tag switching and commit context.
• 📁 Local Folder & Codebase Indexer: Recursive filesystem walker with .gitignore compliance, binary filtering, and batch ingestion.
• 🧩 Language-Aware Code & AST Chunking: Syntactic boundary chunker preserving functions, classes, and comments across Python, TypeScript, JavaScript, Go, Rust, Java, and C++.
• 🖼️ Multimodal Visual Intelligence: Extracts and captions system architecture diagrams, flowcharts, screenshots, and visual plots using OCR and Vision LLMs.
• 🔍 Hybrid Retrieval with RRF: Reciprocal Rank Fusion of dense embeddings and BM25 keyword search for robust precision across code and prose.
• 🛡️ Zero-Hallucination Grounding & Citations: Every sentence in synthesized research and Q&A is verified against source passages and exact file:line citations.
• ⚡ Zero-Dependency Core: Runs out-of-the-box offline using SQLite vector storage and deterministic hashing embeddings.

---

🏗️ System Architecture

                               ┌────────────────────────────────────────────────┐
                               │           Web UI (React 18 + Vite)             │
                               │  - Chat & Citations    - Deep Research Agent   │
                               │  - Knowledge Assets    - GitHub/Folder Ingest  │
                               └───────────────────────┬────────────────────────┘
                                                       │ REST API (FastAPI)
                               ┌───────────────────────▼────────────────────────┐
                               │             KIP Backend Services               │
                               │  - DocumentService    - GitHubIngestService    │
                               │  - IngestService      - DeepResearchAgent      │
                               └───────┬───────────────┬────────────────┬───────┘
                                       │               │                │
            ┌──────────────────────────▼───┐   ┌───────▼────────────┐   └──────────┐
            │   Hybrid Retrieval Engine    │   │  AST Code Chunker  │              │
            │  - Dense Hashing / ST Vectors│   │  - Polyglot parser │              ▼
            │  - BM25 & SQLite FTS5 Search │   │  - Symbol Extractor│     ┌─────────────────┐
            │  - Reciprocal Rank Fusion    │   └────────────────────┘     │ Multimodal Asset│
            │  - Cross-Encoder Reranker    │                              │ - Diagram OCR   │
            └──────────────┬───────────────┘                              │ - Vision LLM    │
                           │                                              └─────────────────┘
            ┌──────────────▼────────────────────────────────────────────────────┐
            │                 Grounded Evidence & Citation Gate                 │
            │      - Pre-generation gate      - Post-generation support check   │
            │      - Exact file:line tracing  - Grounding confidence score      │
            └───────────────────────────────────────────────────────────────────┘

---

⚡ Quick Start

1. Local Development (Windows / Linux / macOS)

Terminal 1: Backend Server (FastAPI)

cd backend
python -m venv .venv
.\.venv\Scripts\activate      # On Linux/macOS: source .venv/bin/activate
pip install -e .[dev]
uvicorn kip.api:app --reload --port 8000

API Docs & Swagger UI: http://localhost:8000/docs

Terminal 2: Frontend Workspace (React + Vite)

cd frontend
npm install
npm run dev

Frontend UI: http://localhost:5173

---

2. Run Full Stack with Docker Compose

# Clone and enter the repository
git clone https://github.com/Om-Talaviya/knowledge-intelligence-platform.git
cd knowledge-intelligence-platform

# Copy environment template
cp .env.example .env

# Launch all services (PostgreSQL, Qdrant, Backend, Frontend)
docker compose up -d

Frontend available at http://localhost

---

📡 API Endpoints Reference

Autonomous Research Agent

• POST /api/agent/research — Executes autonomous deep research across indexed repositories and codebases.

GitHub Repository Ingestion

• POST /api/github/ingest — Clones, extracts symbols, chunks, and indexes a GitHub repo.
• GET /api/github/status/{job_id} — Polls repository ingestion job progress.
• POST /api/github/webhook — Webhook listener for automated re-indexing on push events.

Local Folder Ingestion

• POST /api/ingest/local — Scans and indexes a local codebase directory tree.
• GET /api/ingest/status/{job_id} — Polls directory ingestion progress.

Documents & Knowledge Assets

• POST /api/documents/upload — Upload PDF, DOCX, Markdown, or TXT documents.
• GET /api/documents — List paginated documents.
• GET /api/documents/{id}/chunks — Inspect document chunk breakdown.
• DELETE /api/documents/{id} — Delete document and remove vector embeddings.

Chat & Q&A

• POST /api/chat/ask — Ask a question with grounded retrieval and source citations.
• GET /api/chat/conversations — Retrieve conversation history.

---

🧪 Running Tests & Verification

# 1. Zero-dependency internal verification suite (640+ checks)
cd backend
python -m selfcheck

# 2. Pytest unit & integration test suite (90 tests)
pytest tests/ -v

# 3. Frontend linting and TypeScript compilation
cd frontend
npm run lint
npm run build

---

📄 License

MIT License.