TokenForge Router

Created by team OG_Mode on July 07, 2026

Github Copilot ChatGPT Anthropic Claude

Video Captioning

Grounded Multi-Style Video Captioning Agent is a Dockerized AI system for AMD Developer Hackathon ACT II, Track 2. It reads video-captioning tasks from `/input/tasks.json`, downloads each video, samples resized frames across the full clip using ffmpeg, and uses a Fireworks-hosted vision model to understand the scene. The agent first creates a grounded factual summary of visible settings, subjects, actions, objects, mood, and notable details. It then converts that summary into four required caption styles: formal, sarcastic, humorous-tech, and humorous-non-tech. To improve reliability, the system includes JSON repair, raw model-response handling, and deterministic grounded fallbacks, so even imperfect model outputs can still produce valid `/output/results.json` captions without generic placeholders.

Category tags:

Github Presentation Demo

Explore more applications

HopRouter: Smart Local/Remote AI Router

A token-efficient AI agent that routes tasks between a free local model and Fireworks AI across 8 task categories, minimizing token cost without sacrificing accuracy.

HopRouter

AMD Developer CloudQwen3HuggingFace Hub

HdhshU

Hny nsunuvsnumsumcsu mucmsumsucmuvumsvumsvumsvsumyemymsvusmusmvumsuvdmunsyvnsumsuvmsvmsuvmdvmdmdu dimbdm dmbudmbudbmudmbudmbudmbidmbif mbmdvmd mdand

Effectonious

Assistants APIAudioCraft

Dobbies

Dobbies is an automated security auditor for AI Agents — combining LLM-powered static analysis with AI-vs-AI attack simulations to detect OWASP LLM Top 10 vulnerabilities before they reach production. Pre-deployment, proactive, actionable.

nodesemesta

AMD Developer CloudVercelkiro

Token-Efficient Routing Agent

An AI agent that routes queries to a small or large model based on complexity, cutting inference costs while preserving answer quality — built on Fireworks AI.

Ayesha Syed

AI/ML APIYi-LLMS

TokenRouter the prove-or-escalate agent

An agent that spends Fireworks tokens only when it has to: free plain-code solvers first, then a bundled local Gemma, then Fireworks only after a verified miss. Two-thirds of tasks cost zero tokens, and a wrong free answer is structurally impossible.

SprintSprint

Ruththiragayan Sutharsan
Undergraduate

Upcoming AI Hackathons
For Innovators & Creators

Explore more applications

HopRouter: Smart Local/Remote AI Router

A token-efficient AI agent that routes tasks between a free local model and Fireworks AI across 8 task categories, minimizing token cost without sacrificing accuracy.

HopRouter

AMD Developer CloudQwen3HuggingFace Hub

HdhshU

Hny nsunuvsnumsumcsu mucmsumsucmuvumsvumsvumsvsumyemymsvusmusmvumsuvdmunsyvnsumsuvmsvmsuvmdvmdmdu dimbdm dmbudmbudbmudmbudmbudmbidmbif mbmdvmd mdand

Effectonious

Assistants APIAudioCraft

Dobbies

Dobbies is an automated security auditor for AI Agents — combining LLM-powered static analysis with AI-vs-AI attack simulations to detect OWASP LLM Top 10 vulnerabilities before they reach production. Pre-deployment, proactive, actionable.

nodesemesta

AMD Developer CloudVercelkiro

Token-Efficient Routing Agent

An AI agent that routes queries to a small or large model based on complexity, cutting inference costs while preserving answer quality — built on Fireworks AI.

Ayesha Syed

AI/ML APIYi-LLMS

TokenRouter the prove-or-escalate agent

An agent that spends Fireworks tokens only when it has to: free plain-code solvers first, then a bundled local Gemma, then Fireworks only after a verified miss. Two-thirds of tasks cost zero tokens, and a wrong free answer is structurally impossible.

SprintSprint