InsightMesh is an AI-powered, multi-agent feedback intelligence system that automates the processing, categorization, prioritization, explanation, and visualization of user feedback at scale. Built with Google Cloud Vertex AI, Gemini models, and RAG (Retrieval-Augmented Generation), it empowers organizations to understand user sentiment, extract actionable insights, and close the feedback loop—all in real time.
- 💬 Feedback Analyzer: Analyze raw feedback using an intelligent multi-agent pipeline.
- 📊 Insights Dashboard: View categorized and prioritized feedback along with sentiment and timeline.
- 📝 Add Feedback: Inject new feedback into BigQuery for testing and real-world data simulation.
- ✨ Summarizer (Gemini): Summarize large batches of feedback into digestible insights.
- 🧠 RAG Engine: Retrieve relevant contextual data for better explanations and traceability.
- 📦 BigQuery Integration: All feedback records are stored and queried from Google BigQuery.
- Frontend: Streamlit
- Backend: Python, FastAPI (optional extension)
- Agents: Custom-built multi-agent framework (based on ADK pattern)
- Embedding:
text-embedding-005(Vertex AI) - LLMs: Gemini 1.5 Pro
- Cloud: Google Cloud Platform (GCP)
- Vertex AI
- Matching Engine
- BigQuery
- Cloud Run (for deployment)
- Clone the Repository
git clone https://github.com/your-org/insightmesh.git cd insightmesh
Install Requirements
bash Copy code pip install -r requirements.txt Setup Environment Variables Create a .env file with:
env Copy code PROJECT_ID=your-gcp-project LOCATION=us-central1 BQ_DATASET=your_dataset BQ_TABLE=your_table MATCHING_ENGINE_INDEX=your_index_id Run Locally
bash Copy code streamlit run streamlit_ui/app.py 🧪 Testing Instructions Navigate to the Feedback Analyzer tab, input feedback text, and click Analyze.
Go to Add Feedback to manually inject new feedback into BigQuery.
Check the Insights tab to verify data visualization.
Use Summarizer to test summarization with multiple feedbacks.
If RAG is set up, confirm that relevant context appears alongside the analysis.
📦 Deployment (Cloud Run)
bash
Copy code
gcloud builds submit --tag gcr.io/<PROJECT_ID>/insightmesh-app
gcloud run deploy insightmesh-app
--image gcr.io/<PROJECT_ID>/insightmesh-app
--platform managed
--region us-central1
--allow-unauthenticated
🤖 Agent Pipeline Structure
text
Copy code
Feedback -> SentimentAgent -> CategoryAgent -> PriorityAgent -> RAGAgent -> ExplainerAgent -> LoggerAgent -> Dashboard
🎯 Use Cases
Customer Support Analysis
Product Feedback Monitoring
Feature Request Categorization
UX/UI Sentiment Analysis
Internal Feedback Loop Automation
🧠 What We Learned How to integrate Vertex AI’s suite of models into real-world applications
How to structure a modular agent-based pipeline
How to deploy scalable Streamlit apps on Google Cloud
🛣️ Future Enhancements Add auth layer and user roles
Integrate vector search using pgvector or FAISS
Add Slack/email notification triggers
Add training pipeline for custom models
📄 License MIT License © 2025