Data

Tools

Indexes

Applications

Experiments

Awesome

An open API service indexing awesome lists of open source software.

https://github.com/eshaan-nair/synq

SYNQ: ContextBridge — persistent memory layer for AI agents. SYNQ is a Chrome extension + local backend that gives your AI assistant a memory it was never designed to have.
https://github.com/eshaan-nair/synq

ai-tools backend chrome-extension groq-ai-api javascript memory neo4j-graph productivity typescript

Last synced: about 2 months ago
JSON representation

SYNQ: ContextBridge — persistent memory layer for AI agents. SYNQ is a Chrome extension + local backend that gives your AI assistant a memory it was never designed to have.

Awesome Lists containing this project

README 

          






# SYNQ



### Your AI forgets. SYNQ makes it remember.



Every time you open a new chat, your AI starts from zero. You re-explain your stack, your decisions, your bugs — over and over. **SYNQ eliminates this permanently** by giving your AI assistant persistent, cross-session memory.







[![Stars](https://img.shields.io/github/stars/Eshaan-Nair/Synq?style=for-the-badge&logo=github&labelColor=0B0E14&color=6366F1)](https://github.com/Eshaan-Nair/Synq/stargazers)

[![Forks](https://img.shields.io/github/forks/Eshaan-Nair/Synq?style=for-the-badge&logo=github&labelColor=0B0E14&color=06B6D4)](https://github.com/Eshaan-Nair/Synq/forks)

[![Issues](https://img.shields.io/github/issues/Eshaan-Nair/Synq?style=for-the-badge&logo=github&labelColor=0B0E14&color=02C39A)](https://github.com/Eshaan-Nair/Synq/issues)

[![CI](https://img.shields.io/github/actions/workflow/status/Eshaan-Nair/Synq/ci.yml?style=for-the-badge&label=CI&labelColor=0B0E14&color=02C39A)](https://github.com/Eshaan-Nair/Synq/actions)

[![License: MIT](https://img.shields.io/badge/License-MIT-F8FAFC?style=for-the-badge&labelColor=0B0E14)](LICENSE)

[![Version](https://img.shields.io/badge/version-1.3.2-6366F1?style=for-the-badge&labelColor=0B0E14)](CHANGELOG.md)







Chrome

Neo4j

ChromaDB

Groq

React

TypeScript



**Works with Claude · ChatGPT · Gemini — simultaneously, on the same machine.**



*Switch accounts. Switch AIs. Switch chats. SYNQ still remembers everything.*



https://github.com/user-attachments/assets/ab003d01-3e36-405c-a7a4-9eae417b77ca






---



## The Problem



You're deep into a complex project. You've had 12 conversations with Claude about your architecture, your auth flow, your database schema, and the obscure bug you cracked at 2 AM.



Then you open a new chat.



**It's all gone.** You spend the next 20 minutes re-explaining your stack, what you decided, what you already tried. Every conversation starts from zero. You're not talking to an AI with memory — you're talking to a stranger who happens to be very smart.



## The Solution



SYNQ is a **Chrome extension + local backend** that captures your conversations, distills them into a semantic knowledge graph, and **automatically injects the most relevant context into every new prompt** — using a zero-loss RAG pipeline.



```

 Without SYNQ                           With SYNQ

 ─────────────────────────────          ─────────────────────────────

 New chat → blank slate                 New chat → instant context

 Re-explain your stack                  AI already knows your stack

 Re-explain your decisions              AI remembers your decisions

 Re-explain what you tried              AI recalls past attempts

 20 minutes wasted                      Start working immediately

```



When you type a prompt, SYNQ silently prepends the most relevant context from your history:



```

[SYNQ: Relevant context from your previous session]

### Context 1 (relevance: 87%)

We decided to use Mongoose for the ODM and React Query for server state.

The JWT refresh flow is currently blocked by a frontend state bug.

[END SYNQ CONTEXT]



How do we fix the refresh token issue?

```



**The AI responds as if it was in every conversation with you from the start.**



---



## Table of Contents



- [Key Features](#key-features)

- [How It Works](#how-it-works)

- [Architecture](#architecture)

- [Quick Start](#quick-start)

- [Usage Guide](#usage-guide)

- [Tech Stack](#tech-stack)

- [Privacy & Security](#privacy--security)

- [What's New in v1.3.2](#whats-new-in-v132)

- [Self-Hosting](#self-hosting)

- [Contributing](#contributing)

- [License](#license)



---



## Key Features



### Auto-Connect RAG

Intercepts every prompt and injects relevant context automatically. Zero manual steps after initial setup.



### 100% Local & Private

Ollama, ChromaDB, and Neo4j run on your machine. Your data never leaves your network.



### Zero Data Loss

Sliding window chunker preserves every word — no filtering, no minimum length, no information dropped.



### Privacy Scrubbing

API keys, JWTs, emails, connection strings auto-redacted before any text reaches external services.



### Semantic Knowledge Graph

22 entity types, 20+ relation types — captures both technical architecture and personal context.



### D3.js Graph Visualization

Force-directed graph with degree-scaled nodes, curved edges, hover tooltips, and zoom controls.



### ⏯Pause / Resume Toggle

Suspend context injection without losing session state. Resume instantly when ready.



### Multi-Platform

Claude, ChatGPT, and Gemini — all supported simultaneously with platform-specific selectors.



---



## How It Works



```

  ┌─────────────────────────────────────────────────────────────────┐

  │  1. CAPTURE                                                     │

  │     Finish a chat → click Save Chat in the extension popup      │

  │     Extension scrapes full conversation (user + AI turns)       │

  │     FNV-1a fingerprint deduplication prevents double-saves      │

  └─────────────────────────┬───────────────────────────────────────┘

                            │

  ┌─────────────────────────▼───────────────────────────────────────┐

  │  2. PRIVACY SCRUB                                               │

  │     API keys, JWTs, emails, connection strings → [REDACTED]     │

  │     Happens in the browser before anything is transmitted       │

  └─────────────────────────┬───────────────────────────────────────┘

                            │

              ┌─────────────┴─────────────┐

              │                           │

  ┌───────────▼──────────┐   ┌────────────▼─────────────┐

  │  3A. VECTOR TRACK    │   │  3B. GRAPH TRACK         │

  │  (RAG — all local)   │   │  (Knowledge Graph)       │

  │                      │   │                          │

  │  Sliding window      │   │  Groq LLaMA 3.1 →       │

  │  chunker (300w/80ov) │   │  summarize facts →       │

  │  → Ollama embeddings │   │  extract triples →       │

  │  → ChromaDB cosine   │   │  Neo4j MERGE             │

  └───────────┬──────────┘   └────────────┬─────────────┘

              │                           │

              └─────────────┬─────────────┘

                            │

  ┌─────────────────────────▼───────────────────────────────────────┐

  │  4. AUTO-CONNECT (every prompt, automatically)                  │

  │     User types → intercepted → ChromaDB cosine search →        │

  │     Top-3 chunks prepended to prompt → AI gets full context     │

  └─────────────────────────────────────────────────────────────────┘

```



> **Full pipeline details:** [RAG_PIPELINE.md](RAG_PIPELINE.md) · **Architecture deep-dive:** [ARCHITECTURE.md](ARCHITECTURE.md)



---



## Architecture



```

SYNQ/

├── backend/

│   └── src/

│       ├── routes/             # chat · context · graph · rag

│       ├── services/           # chroma · chunker · embeddings

│       │                       # extractor · mongo · neo4j

│       └── utils/              # logger · privacy

├── dashboard/

│   └── src/

│       ├── api/                # rag · synq

│       ├── components/         # ChatViewer · GraphView

│       └── App.tsx

├── extension/

│   ├── popup/                  # popup.html · popup.ts · popup.css

│   └── src/

│       ├── platforms/          # claude · chatgpt · gemini · index

│       ├── content.ts          # DOM scraping, prompt interception

│       └── background.ts       # service worker, backend proxy

├── scripts/

│   └── check-selectors.js      # Playwright smoke test

├── docker-compose.yml          # Neo4j + MongoDB + ChromaDB

└── start.sh / start.bat        # One-command launcher

```



### Service Map



| Service | Port | Purpose |

|:---|:---|:---|

| **Backend** | `3001` | Express API — routes, rate limiting, security headers |

| **Neo4j** | `7474` / `7687` | Knowledge graph — HTTP browser + Bolt protocol |

| **MongoDB** | `27017` | Sessions, FullChat storage, active session singleton |

| **ChromaDB** | `8000` | Vector store — cosine similarity with nomic-embed-text |

| **Ollama** | `11434` | Local embedding generation — 768-dim vectors |

| **Dashboard** | `5173` | React dev server — graph visualization, session management |



---



## Quick Start



### Prerequisites



| Requirement | Link |

|:---|:---|

| **Node.js 20 LTS** | [nodejs.org](https://nodejs.org) |

| **Docker Desktop** | [docker.com](https://www.docker.com/products/docker-desktop) (WSL2 on Windows) |

| **Ollama** | [ollama.com](https://ollama.com) — then run `ollama pull nomic-embed-text` |

| **Groq API Key** | [console.groq.com](https://console.groq.com) — free, no credit card |



### 1. Clone & Configure



```bash

git clone https://github.com/Eshaan-Nair/Synq.git

cd Synq



# macOS / Linux

cp backend/.env.example backend/.env



# Windows

copy backend\.env.example backend\.env

```



Open `backend/.env` and set your `GROQ_API_KEY`. All other values have working local defaults.



### 2. Start Everything (One Command)



```bash

# macOS / Linux

chmod +x start.sh && ./start.sh



# Windows

start.bat

```



The start script will:

- ✅ Check for Ollama and pull `nomic-embed-text` if missing

- ✅ Start Docker containers (Neo4j + MongoDB + ChromaDB)

- ✅ Install dependencies and build the extension

- ✅ Launch backend (port 3001) and dashboard (port 5173)

- ✅ Health-check the backend before proceeding



### 3. Load the Chrome Extension



1. Open Chrome → `chrome://extensions`

2. Enable **Developer mode** (toggle, top-right)

3. Click **Load unpacked** → select the `Synq/extension` folder

4. The **SYNQ** icon appears in your toolbar



Open Claude, ChatGPT, or Gemini — the SYNQ badge appears automatically.



Alternative: Manual Setup (separate terminals)



```bash

# Terminal 1 — Backend

cd backend && npm install && npm run dev



# Terminal 2 — Dashboard

cd dashboard && npm install && npm run dev



# Build extension (one time)

cd extension && npm install && npm run build



# Start databases

docker compose up -d



# Start Ollama (if not running)

ollama serve

```



---



## Usage Guide



### Saving a Conversation



1. Have a conversation on Claude, ChatGPT, or Gemini

2. Click the **SYNQ** icon in the Chrome toolbar

3. Enter a project name (e.g., `AuthService v2`) and click **Save Chat**

4. SYNQ scrubs PII, chunks the conversation, embeds it locally, and runs graph extraction — typically under 5 seconds



### Auto-Connect (RAG) — The Main Workflow



Once a session is saved and active, **SYNQ auto-connects every time you open a supported platform**. Type your next prompt normally — SYNQ searches its vector store and prepends the most relevant context before sending.



**To pause** injection without losing your session: click **Pause SYNQ** in the popup.



### Classic Inject — For Cold Starts



Click **Inject Context (one-time)** in the popup. SYNQ pulls the structured project summary from the knowledge graph and pastes it directly into the chat input — you send it manually to prime the conversation.



### Dashboard



Open **http://localhost:5173** to access:



| Tab | What It Shows |

|:---|:---|

| **Graph** | Interactive D3.js force graph — hover nodes to see connections and labels |

| **History** | All extracted semantic triples with timestamps, newest first |

| **Chat** | Full saved conversation with color-coded user/assistant bubbles |



Use **Load into Extension** to set any session as the active context across all open AI tabs.



---



## Tech Stack



| Layer | Technology | Notes |

|:---|:---|:---|

| **Extension** | TypeScript, Chrome Manifest V3 | Content scripts, service worker, Shadow DOM |

| **Backend API** | Node.js, Express 5, TypeScript | Rate limiting, helmet, CORS, input validation |

| **Graph Database** | Neo4j 5.18 | Typed triples, MERGE idempotency |

| **Vector Database** | ChromaDB 0.6.3 (cosine) | Local, persistent, pinned version |

| **Embeddings** | Ollama `nomic-embed-text` | 768-dim, CPU-only, ~270MB, parallel generation |

| **LLM Processing** | Groq LLaMA 3.1 8B Instant | Free tier, fact extraction + graph summarization |

| **Dashboard UI** | React 19, Vite 7, D3.js v7 | Force graph, TypeScript strict mode |

| **Infrastructure** | Docker Compose | Named volumes, pinned image versions |

| **Testing** | Jest + ts-jest | Unit tests for chunker, privacy scrubber, RAG |

| **CI/CD** | GitHub Actions | Backend build+test, dashboard build, extension build |



---



## Privacy & Security



**Your data stays on your machine.** All conversation data, embeddings, and graph triples are stored in local Docker volumes. Nothing syncs externally.



The only external service is **Groq**, used solely for graph triple extraction — and the text sent there is PII-scrubbed first. RAG chunking, embedding, and retrieval are 100% local.



| Control | Implementation |

|:---|:---|

| **Auto-redaction** | API keys, JWTs, emails, connection strings scrubbed before transmission |

| **Rate limiting** | 200 req/min global · 10 req/min on `/api/chat/save` |

| **CORS locked** | Only `localhost:5173`, `localhost:4173`, and `chrome-extension://` |

| **Input validation** | All routes validate sessionId, platform, text length |

| **Helmet headers** | Security headers on every response |

| **Shared secret** | Optional `X-SYNQ-Secret` header for request authentication |

| **Opt-in only** | SYNQ reads the DOM only when you save a chat or Auto-Connect is active |



> See [SECURITY.md](SECURITY.md) for the vulnerability reporting policy.



---



## What's New in v1.3.2



### UI/UX & Extension Refinements (v1.3.2)



- **Collapsible Sidebar:** Left sidebar with Facts and Chat tabs is now collapsible for a better viewing experience.

- **Graph Settings:** Added granular controls for dashboard graph parameters (node size, position, tension).

- **Performance Fixes:** Fixed staggered graph rendering causing lag in large knowledge graphs.

- **Extension Controls:** Added the ability to save the same chat multiple times, explicit 'Unload' session buttons, and an instant badge toggle for SYNQ.



### Zero-Loss RAG Pipeline



The v1.0 Groq-based topic splitter was **lossy** — it silently discarded personal facts and rejected short messages. v1.2 replaced it entirely:



| | Groq Topic Splitter (v1.0) | Sliding Window Chunker (v1.2+) |

|:---|:---|:---|

| Personal facts ("my dog is Noob") | ❌ Deleted as "filler" | ✅ Preserved verbatim |

| Short messages | ❌ Rejected | ✅ Always included |

| API call on save | ❌ Yes (1–3s, costs quota) | ✅ None (pure function) |

| Information loss | ❌ Significant | ✅ **Zero** |



### Auto-Connect



No more manual Connect button. Open a supported AI platform with an active session and the interceptor **auto-attaches**. Just type — SYNQ handles the rest.



### Knowledge Graph Expansion



Entity types grew from 12 → 22, now capturing the full picture:



`Project · Technology · Feature · Bug · Decision · Concept · Library · API · Database · Framework · Auth · Architecture · Person · Pet · Goal · Problem · Preference · Habit · Location · Organization · Tool · Pattern`



### Graph Visualization Overhaul



- Degree-scaled nodes (8–60px radius) — hubs are visually dominant

- Curved quadratic bezier edges with hover-reveal labels

- Per-type colored glow filters and 3D radial gradients

- Hover tooltip with name, type, and connection count

- Zoom controls (+/−/reset)



> See [CHANGELOG.md](CHANGELOG.md) for the full history.



---



## Self-Hosting



All data lives in local Docker volumes. For custom passwords, port changes, backups, data reset, and reverse proxy configuration:



**[Self-Hosting Guide →](SELF_HOSTING.md)**



---



## Contributing



Contributions are welcome — bug fixes, new platform support, UI improvements, documentation, and test coverage.



**[Contributing Guide →](CONTRIBUTING.md)** — fork/clone/branch workflow, commit format, and step-by-step guide for adding a new AI platform.



**[Code of Conduct →](CODE_OF_CONDUCT.md)**



**Good first issues** are tagged [`good first issue`](https://github.com/Eshaan-Nair/Synq/issues?q=is%3Aissue+label%3A%22good+first+issue%22) in the issue tracker.



---



## Documentation



| Document | Description |

|:---|:---|

| [ARCHITECTURE.md](ARCHITECTURE.md) | Full architecture, data flow, security model, data models |

| [RAG_PIPELINE.md](RAG_PIPELINE.md) | Pipeline details, similarity scoring, tuning parameters |

| [PLATFORM_SELECTORS.md](PLATFORM_SELECTORS.md) | Per-platform CSS selectors and staleness tracking |

| [SELF_HOSTING.md](SELF_HOSTING.md) | Ports, passwords, backups, reverse proxy setup |

| [CONTRIBUTING.md](CONTRIBUTING.md) | Fork workflow, commit format, adding new platforms |

| [CHANGELOG.md](CHANGELOG.md) | Full version history |



---



## License



MIT — see [LICENSE](LICENSE).



---









**Stop re-explaining yourself.**

**Give your AI the memory it should have had from day one.**







*Built by [Eshaan Nair](https://github.com/Eshaan-Nair) — a developer, for developers.*