# SmartChat

- **Event:** [WeaveHacks 3: Self-Improving Agents Hackathon with Weights & Biases](https://cerebralvalley.ai/e/weave-hacks-3-self-improving-agents-hackathon-with-weights-and-biases-7014fe80)
- **When:** Jan 31 at 9:00 AM – Feb 1 at 5:00 PM (PST)
- **Where:** 400 California St, San Francisco, CA
- **Team:** [Amit Bhadoria](https://cerebralvalley.ai/u/amit_b), [Siya Bhadoria](https://cerebralvalley.ai/u/siyabhadoria)
- **GitHub:** https://github.com/siyabhadoria/smartchat
- **Demo video:** https://www.loom.com/share/3bd55f4ff95a455e905dde58ccdf45f4
- **Gallery:** https://cerebralvalley.ai/e/weave-hacks-3-self-improving-agents-hackathon-with-weights-and-biases-7014fe80/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/weave-hacks-3-self-improving-agents-hackathon-with-weights-and-biases-7014fe80/hackathon/gallery/19

SmartChat is an AI agent built on the CoALA framework. It uses a three-tier memory system to give context-aware responses and improve from feedback.

It has Episodic Memory, which stores past conversations in time order using the Soorma episodic service so the agent understands conversational flow. Semantic Memory stores extracted facts in a vector database and retrieves them with RAG to ground responses in knowledge. Working Memory holds the agent’s active state, including feedback penalties and reasoning traces that power features like “Why did you say that?”

SmartChat includes a real-time voice interface using Daily Pipecat.  

For observability, SmartChat uses Weights and Biases Weave. Key functions are traced, including retrieved facts, prompts, and model outputs. User feedback is logged to show how behavior changes over time.

Negative feedback lowers semantic fact relevance and injects corrections into episodic memory, enabling self-correction.

---

Markdown version of https://cerebralvalley.ai/e/weave-hacks-3-self-improving-agents-hackathon-with-weights-and-biases-7014fe80/hackathon/gallery/19. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt