Run LLaVA vision-language models for image question answering and multi-turn chat.
npx skills add https://github.com/Signmanal/VIGIL --skill llava-signmanal
Please help me install this Agent Skill. Skill: llava Source: https://github.com/Signmanal/VIGIL/tree/main/optional-skills/mlops/llava Command: npx skills add https://github.com/Signmanal/VIGIL --skill llava-signmanal
This Skill eliminates the need for specialized computer vision expertise to extract insights, answer questions, and hold natural language conversations about image content, cutting down hours of manual visual inspection for analysts and researchers.
Use the llava skill to analyze the attached incident image and list all visible objects, any readable text, and a summary of the scene context.
High-intent search queries and answers about installing and using this skill.
Visual question answering works by applying vision-language models to process image pixels and text prompts together. This Skill leverages LLaVA architectures to enable multi-turn image chat where you can ask follow-up questions without re-uploading the image.
To perform multi-turn image chat, upload your image and provide an initial prompt like asking for a summary of visible objects and readable text. The Skill maintains context so you can ask follow-up questions to drill into specific details without recontextualizing the image.
This Skill handles document verification, security investigation, and threat research by reading text from document images and summarizing scene content. It processes visual artifacts to support analysts requiring multi-turn image chat for thorough visual inspection tasks.
Limitations of using multimodal AI for scene understanding include dependency on GPU-enabled hardware and the computational overhead of running large vision-language models. Processing requires PyTorch and Transformers, which may restrict deployment in resource-constrained environments.
hotriluanProcess multimedia inputs for multimodal analysis, transcription, captioning, and generation.
ChimeraFoundationaEnable image-text interaction for multimodal conversation and visual question answering.
AshesOfTheUndeadAnalyze static images for object detection, scene understanding, and OCR.
GarrettRoiCombine CLIP vision encoders with Vicuna and LLaMA for visual question answering.
