The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Back to feedPost
miniSara AI Agent - AI Agents, LLM, Tool Calling, Agent Orchestration
miniSara AI Agent is a production-oriented AI coding agent designed to help developers understand, modify, and automate software projects through natural-language interaction.
I designed and developed the agent architecture with a strong focus on reliability, security, extensibility, and developer experience. The system includes an LLM provider abstraction supporting multiple providers, secure tool execution, agent orchestration, streaming responses, approval-based mutations, verification, and self-healing workflows.
Key capabilities:
AI agent orchestration and multi-step reasoning workflows
LLM provider abstraction for OpenAI, Anthropic, and Ollama
Secure tool-calling and controlled tool execution
Streaming AI responses and real-time tool activity
Approval workflows for potentially destructive changes
Automated verification and recovery/self-healing
TypeScript/Node.js architecture with strong automated test coverage
Cross-platform executable and development-environment support
Tech: TypeScript, Node.js, LLMs, OpenAI, Anthropic, Ollama, Tool Calling, Streaming, Agent Orchestration, AI Engineering.
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Related posts
AI Assistant Using Your Business Knowledge Base — RAG on Your Documents
THE PROBLEM
Q&A bots break down when knowledge lives in documents: a 100-page manual has no "questions" to match, it can't fit into a prompt, and generic chatbots hallucinate instead of admitting what they don't know.
THE SOLUTION
A RAG (Retrieval-Augmented Generation) knowledge base: documents are split into meaningful chunks, embedded into a vector index, and the assistant answers from the right sections — by meaning, not keywords.
Any format as-is: PDF, DOCX, TXT, Markdown — 100+ pages is fine
Answers grounded in YOUR documents — it says "I don't have that information" rather than inventing
Source references — every answer shows which document and section it came from
Runs on your infrastructure — documents never leave your control
One command to re-index after updating documents — documented, no programmer needed
LIVE DEMO
Ask it about my products, pricing, deployment and support — it answers from 6 real documents with sources:
https://kb-assistant-demo.dvc-ai.workers.dev
The 'I don't have that information' line is the part most RAG builds skip, and it's the one that matters. How do you set the cutoff? On mine, a fixed similarity threshold broke once I filtered results by user role. Scores shifted and it refused questions it could answer.
CONGRATS
THIS IS A HUGE MILESTONE
I'm not where I want to be but I'm far from where I started
So happy to have reached this milestone
Wow congratulation
Excited to share the recent launch of Listen Out Loud Media website, now live at listenoutloud.org.
This was a meaningful build. Listen Out Loud is a registered 501(c)(3) nonprofit that funds documentary films, classroom speakers, and media experiences for schools that can't afford them. The whole mission is around access, making sure every student gets to hear a story that sounds like theirs.
The donate section background animation started in Figma. I used Figma's motion tools to prototype the sine wave look and feel, iterating on the wave shapes, layering, and glow nodes until it felt right. Once the visual direction was locked, I used that Figma frame as the reference to rebuild it as a custom Framer code component with Claude.
The final component renders animated sine waves in SVG using cubic bezier paths and runs per-wave node passes on a random start and stop 60-second loop. This creates a unique path for the nodes so they are never the same as the previous loop. Fully prop-controlled with Framer's property controls API, wave count, amplitude, color, speed, and glow intensity all editable directly in the canvas.
Meaningful project with a clear digital experience.