Browse applications built on Streamlit technology. Explore PoC and MVP applications created by our community and discover innovative use cases for Streamlit technology.
Uses Gemma 4 to classify complex vs. simple requests, and routes them to a local LLM / cloud provider as needed.
SovereignSwarm is an autonomous video captioning engine using Fireworks AI and OpenCV to extract chronological frames and generate structured, multi-tone captions (Formal, Sarcastic, Humorous Tech, Everyday) in an enterprise-ready JSON format.
A reinforcement-learning router that learns which AI model to call for each query, cutting cost by ~93% without ever hardcoding a routing rule.
A video captioning agent that watches a clip once, then rewrites what it saw as Formal, Sarcastic, or Humorous captions
Penny routes each task to the cheapest Fireworks model that can still get it right — or to zero-token local solving when it doesn't need a model at all — re-deciding cost in real time per task instead of picking one model for everything.
Helps in optimizing your model costs by selecting the appropriate models with token efficiency
A local-first AI routing agent that uses a fine-tuned DistilBERT classifier to choose cheap or stronger Fireworks models, reducing routing token cost while preserving accuracy across eight task categories.
An AMD-powered HS-code classifier that turns a plain-English product description into its tariff code, import duty, and licensing restrictions — every number backed by an official government source URL.
Transform your codebase into a VC-grade startup pitch deck in seconds. Commit-To-Pitch analyzes your code, extracts technical intent, and generates a compelling, investable pitch using advanced AI.
Real-time quantum device digital twin with AMD MI300X GPU acceleration and intelligent triage agent that maintains state sync under extreme throughput conditions.
Generates accurate captions for short videos in four distinct styles—formal, sarcastic, humorous-tech, and humorous-non-tech—using AI-powered video understanding to create engaging, context-aware summaries with consistent formatting and tone.
Your autonomous growth team. Paste a product URL and a goal — get audience research, ad copy, creative, landing pages, A/B testing, and a "ship this" call in minutes.
ReachIQ AI's Video RAG engine ingests competitor YouTube videos, builds a vector knowledge base, and generates strategic competitive intelligence reports — powered by Gemma 3 4B running natively on AMD Developer Cloud via ROCm + vLLM.
Cost-minimizing AI routing agent, routes each task to the cheapest sufficient model. Tries to use local models for simple tasks and escalates to Fireworks AI models if needed. Agentic task classification, model tier selection, and automatic fallback.
An AI agent that watches a short video and captions it in four distinct voices: formal, sarcastic, tech-humor, and everyday-humor
An automated tool that watches video clips and writes creative captions for them in different writing styles.
Life broke the plan. Rebound rebuilds it—using diagnostics, topic dependencies, exam importance, and time left to decide what to protect, what to cut, and what to study next.
VidCap watches a short video clip and writes four captions in clearly distinct voices — formal, sarcastic, humorous-tech, humorous-non-tech all grounded in one verified fact sheet, in about 15 seconds per clip via Fireworks AI.
A local DistilBERT binary router sends each query to the cheapest suitable Fireworks model, preserving answer quality while using zero routing tokens across eight task categories.
FollowThrough is aAn AI agent that turns meeting transcripts into verified, ready-to-send action items — with a built-in fact-checker so it never invents a task.
An AI platform that creates video parodies by re-voicing existing footage or generating 9:16 vertical short-form videos from scratch using Moonshot Kimi K2.6, FLUX.1, and Google Gemma models with an asynchronous multi-threaded Streamlit interface.
A multi-model AI pipeline that analyzes video frames chronologically and synthesizes verified captions across four distinct tonal registers.
An AI agent that verifies job postings live — checking WHOIS domain records, real search results, and known scam patterns — and shows the exact evidence behind every risk score, not just a verdict.
An intelligent AI model router that automatically selects the best LLM for each request based on configurable routing strategies such as cost, latency, capability, and availability.
Privacy-first AI coding assistant powered by AMD ROCm and Fireworks AI, delivering lightning-fast local inference, intelligent cloud routing, real-time code analysis, bug detection, and personalized developer guidance.
Adaptive Router is an intelligent orchestrator that dynamically routes AI prompts to the most cost-effective model based on task complexity, slashing inference costs by up to 86% while maintaining 100% accuracy and injecting live web context.
Hybrid token-efficient AI agent that analyzes prompt complexity and routes requests to either a local Qwen model or a Fireworks-hosted LLM, reducing API usage while preserving response quality and explainable decision-making.
An AI-powered Video Captioning Agent that automatically generates accurate, context-aware captions, timestamps, and scene descriptions for videos, improving accessibility, searchability, and content understanding.
An AI-driven hardware orchestrator mimicking the human nervous system to eliminate software redundancy loops, dynamically optimizing task routing, thermal management, and power efficiency across complex CPU and GPU topologies.
Four-agent AI system that classifies production incidents, routes them to the right team, writes role-appropriate Jira tickets, and detects architectural patterns — all with a human approval gate before anything is sent.