VectorOps Assist is a hybrid AI support framework built to help teams handle customer queries faster, cheaper, and more reliably.
Instead of relying only on rule-based chatbots or expensive LLMs, VectorOps combines the best of both: deterministic rules for simple tasks and smart GenAI reasoning for complex cases.
It’s designed for engineers who want full control, transparency, and production-ready behavior right out of the box.
Support workflows typically fall into two categories:
Cheap and fast, but easily break when customers ask anything unexpected.
Intelligent but:
- slow
- expensive
- unpredictable
- unsafe for critical flows like refunds or billing
Teams must choose between cost and intelligence.
VectorOps Assist eliminates this trade-off.
VectorOps Assist introduces a two-tier architecture:
Handles:
- cancellations
- order status
- simple billing issues
- refund requests
- FAQ responses
Instant, predictable, cost-free.
Handles:
- deep reasoning
- multi-step actions
- CRM & database lookups
- refund eligibility
- ticket creation
- long-term memory
LLM never fabricates logic — all critical decisions run through Python tools.
This hybrid design reduces LLM usage by up to 90%.
Notebook-style interface:
- live Python execution
- tool call visualization
- log inspector
- rule engine editor
- memory viewer
Built with React + Tailwind.
- lightweight keyword/regex intent detection
- template-based replies
- micro memory sliding window
- ultra-fast routing
Powered by:
- ReAct reasoning loop
- TTL-based memory manager
- Function-calling interface
- Dynamic tool discovery (MCP-like)
get_order_status(order_id)check_refund(order_id)file_complaint_ticket(...)query_customer_db(sql_query)
Includes simulated:
- order DB
- ticketing system
- refund policies
- CRM
Gives the agent real production-like behavior.
Automatically generates:
server.py(FastAPI wrapper)Dockerfile
Deployable to:
- Cloud Run
- ECS
- Any container-based infra
- Node.js 18+
- Gemini API Key
- (Optional) Python 3.11
git clone https://github.com/iamkajalbiswas/VectorOps-Assist
cd VectorOps-Assist2. Install Dependencies
npm install3. Add API Key
Create .env:
API_KEY=your_gemini_api_key4. Start Workbench
npm startOpen browser: 👉 http://localhost:8080
This step loads all core components:
- imports
- backend simulation
- rule engine
- agent core
Run simple test prompts to see instant deterministic replies.
The agent will:
- analyze intent
- decide if a tool is needed
- execute the tool
- compose the final answer
Compare Tier 1 and Tier 2 outputs side-by-side to understand hybrid behavior.
Running the export cell generates:
server.pyDockerfile
These files can be deployed directly to container environments.
- Hybrid Tier-1/Tier-2 intelligence
- ReAct reasoning loop
- Deterministic function-calling safety
- Full support workbench UI
- Memory system with TTL and compression
- Container-ready deployment
- Clean logging & observability
Support teams need:
- speed
- accuracy
- transparency
- reliability
- low cost
VectorOps intelligently splits workloads between rules and AI to deliver all of these benefits simultaneously.
- Real MCP plugin support
- Multi-agent collaboration
- Analytics dashboard
- Kubernetes deployment templates
- Automation marketplace
- Enhanced reasoning visualization
Contributions are welcome!
Submit PRs to improve tools, UI, workflows, or core agent logic.
MIT License — see LICENSE for full details.