VidChain 0.2.0__tar.gz

vidchain-0.2.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Rahul Sharma
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

vidchain-0.2.0/PKG-INFO

@@ -0,0 +1,184 @@
+Metadata-Version: 2.4
+Name: VidChain
+Version: 0.2.0
+Summary: A Lightweight Video RAG Framework for Multimodal Reasoning
+Author-email: Rahul Sharma <rahulsharma.hps@gmail.com>
+License: MIT
+Project-URL: Homepage, https://github.com/rahulsiiitm/videochain-python
+Project-URL: Bug Tracker, https://github.com/rahulsiiitm/videochain-python/issues
+Keywords: video-rag,multimodal,ai,computer-vision,whisper,ollama,yolo
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Developers
+Classifier: Intended Audience :: Science/Research
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Requires-Python: >=3.11
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: opencv-python
+Requires-Dist: scenedetect
+Requires-Dist: openai-whisper
+Requires-Dist: sentence-transformers
+Requires-Dist: faiss-cpu
+Requires-Dist: litellm
+Requires-Dist: google-generativeai
+Requires-Dist: moviepy
+Requires-Dist: pillow<12.0
+Requires-Dist: torch
+Requires-Dist: torchvision
+Requires-Dist: torchaudio
+Requires-Dist: numpy
+Requires-Dist: tqdm
+Requires-Dist: python-dotenv
+Requires-Dist: ultralytics
+Requires-Dist: librosa
+Requires-Dist: easyocr
+Dynamic: license-file
+
+# VidChain: Video Intelligence RAG Framework
+> Edge-optimized multimodal RAG framework for video understanding — transforms raw footage into a structured, queryable knowledge base.
+
+![Python](https://img.shields.io/badge/Python-3.11+-blue) ![CUDA](https://img.shields.io/badge/CUDA-12.1-green) ![License](https://img.shields.io/badge/License-MIT-yellow) ![Status](https://img.shields.io/badge/Status-beta-orange) [![PyPI version](https://badge.fury.io/py/vidchain.svg)](https://pypi.org/project/VidChain/)
+
+---
+
+## Overview
+
+vidchain v0.2.0 is a lightweight, modular framework that combines computer vision, smart OCR, speech recognition, and LLM reasoning into a unified **late-fusion pipeline**. Designed to run efficiently on consumer-grade GPUs (tested on NVIDIA RTX 3050), it extracts human-readable stories from raw sensor data, making on-device video intelligence practical without massive cloud dependency.
+
+At the heart of the framework is **B.A.B.U.R.A.O.** (*Behavioral Analysis & Broadcasting Unit for Real-time Artificial Observation*), an elite AI copilot that uses abductive reasoning to translate raw, flickering object/action logs into flowing, conversational narratives.
+
+---
+
+## Core Pipeline
+```text
+Video Input → Adaptive Keyframes → Dual-Brain Vision (YOLO + MobileNet) + OCR → Audio Transcription → Semantic Chunking → FAISS Vector DB → B.A.B.U.R.A.O. RAG
+````
+
+-----
+
+## Key Capabilities
+
+### 🧠 Dual-Brain Vision Engine
+
+Instead of basic classification, vidchain uses a two-pronged visual approach:
+
+  * **The "Noun" Engine (YOLOv8):** Detects specific objects (e.g., "1 person, 2 laptops").
+  * **The "Verb" Engine (MobileNetV3):** Classifies the intent or state of the scene (e.g., NORMAL, SUSPICIOUS, VIOLENCE).
+
+### 🔤 Context-Aware OCR
+
+Powered by EasyOCR, the system intelligently scans for text *only* when YOLO detects readable surfaces (monitors, laptops, books, whiteboards), saving massive compute power while capturing ground-truth data (e.g., reading the brand "ASUS Vivobook" off a laptop).
+
+### B.A.B.U.R.A.O. RAG Engine (Conversational)
+
+Unlike standard RAGs that read out robotic timelines, B.A.B.U.R.A.O. acts as a human copilot:
+
+  * **Abductive Reasoning:** If it sees a "laptop" and a "keyboard", it deduces the scene is a "computer desk."
+  * **Sensor Filtering:** Automatically ignores momentary hardware glitches/hallucinations (e.g., a TV briefly misidentified as an oven).
+  * **Natural Translation:** Translates raw model labels like `VIOLENCE` into contextual human behaviors like "the person became visibly frustrated and hit the desk."
+
+### Edge-First GPU Optimization
+
+Engineered to prevent VRAM crashes. Smart memory routing disables PyTorch's buggy layer fusion during YOLO inference and safely manages VRAM across concurrent vision, audio, and language models.
+
+-----
+
+## Installation
+
+```bash
+# 1. Install the core package
+pip install vidchain
+
+# 2. IMPORTANT: Install GPU-accelerated PyTorch (CUDA 12.1 recommended)
+pip install torch torchvision torchaudio --index-url [https://download.pytorch.org/whl/cu121](https://download.pytorch.org/whl/cu121) --force-reinstall
+```
+
+> ⚠️ **Requirement:** NVIDIA drivers and CUDA are strongly recommended. To verify your hardware is correctly mapped, run the built-in diagnostic script: `python scripts/check_gpu.py`
+
+-----
+
+## Quick Start
+
+### 1 — Analyze a Video (Build Knowledge Base)
+
+Analyze a video file, extract multimodal context, and generate a structured JSON timeline:
+
+```bash
+vidchain-analyze sample.mp4
+```
+
+*This command automatically builds a FAISS index and drops you into the interactive B.A.B.U.R.A.O. chat terminal.*
+
+### 2 — Train the Action Engine
+
+Fine-tune the MobileNetV3 "Verb" classifier on your domain-specific dataset:
+
+```bash
+vidchain-train
+```
+
+Place labeled training images under `data/train/` before running.
+
+-----
+
+## Knowledge Base Schema
+
+The framework utilizes **Semantic Chunking** to compress repetitive frames. The `knowledge_base.json` outputs a clean, fused timeline:
+
+```json
+{
+    "time": 0.97,
+    "type": "ocr",
+    "content": "ASUS Vivabook"
+},
+{
+    "time": 3.87,
+    "type": "visual",
+    "content": "Duration: [3.87s - 6.77s] | Subjects: 1 laptop, 1 tv | Action State: SUSPICIOUS"
+},
+{
+    "time": 19.34,
+    "type": "visual",
+    "content": "Duration: [19.34s - 19.34s] | Subjects: 1 tv | Action State: VIOLENCE"
+}
+```
+
+-----
+
+## Tech Stack
+
+| Component | Technology |
+|---|---|
+| Object Detection (Nouns) | YOLOv8s |
+| Intent Classification (Verbs) | MobileNetV3 (Custom fine-tuned) |
+| Text Extraction (OCR) | EasyOCR |
+| ASR (Audio) | OpenAI Whisper (Base) |
+| Vector Database | FAISS + Sentence-Transformers (`all-MiniLM-L6-v2`) |
+| LLM Routing | LiteLLM (`gemini-2.5-flash` default, Ollama supported) |
+| GPU Runtime | CUDA 12.1 (Optimized for 4GB+ VRAM) |
+
+-----
+
+## Roadmap
+
+  - [ ] **Real-time streaming pipeline** — live ingestion and indexing with low-latency event detection.
+  - [ ] **Advanced temporal reasoning** — multi-clip reasoning and cross-camera subject tracking.
+  - [ ] **Interactive Dashboard** — PyQt5 HUD for video playback, timeline visualization, and KB exploration.
+
+-----
+
+## Contributing
+
+Contributions, issues, and feature requests are highly welcome\! Open a GitHub issue or submit a pull request.
+
+-----
+
+## Author
+
+**Rahul Sharma** — B.Tech CSE, IIIT Manipur
+
+## License
+
+Distributed under the [MIT License](https://www.google.com/search?q=LICENSE).

vidchain-0.2.0/README.md

@@ -0,0 +1,146 @@
+# VidChain: Video Intelligence RAG Framework
+> Edge-optimized multimodal RAG framework for video understanding — transforms raw footage into a structured, queryable knowledge base.
+
+![Python](https://img.shields.io/badge/Python-3.11+-blue) ![CUDA](https://img.shields.io/badge/CUDA-12.1-green) ![License](https://img.shields.io/badge/License-MIT-yellow) ![Status](https://img.shields.io/badge/Status-beta-orange) [![PyPI version](https://badge.fury.io/py/vidchain.svg)](https://pypi.org/project/VidChain/)
+
+---
+
+## Overview
+
+vidchain v0.2.0 is a lightweight, modular framework that combines computer vision, smart OCR, speech recognition, and LLM reasoning into a unified **late-fusion pipeline**. Designed to run efficiently on consumer-grade GPUs (tested on NVIDIA RTX 3050), it extracts human-readable stories from raw sensor data, making on-device video intelligence practical without massive cloud dependency.
+
+At the heart of the framework is **B.A.B.U.R.A.O.** (*Behavioral Analysis & Broadcasting Unit for Real-time Artificial Observation*), an elite AI copilot that uses abductive reasoning to translate raw, flickering object/action logs into flowing, conversational narratives.
+
+---
+
+## Core Pipeline
+```text
+Video Input → Adaptive Keyframes → Dual-Brain Vision (YOLO + MobileNet) + OCR → Audio Transcription → Semantic Chunking → FAISS Vector DB → B.A.B.U.R.A.O. RAG
+````
+
+-----
+
+## Key Capabilities
+
+### 🧠 Dual-Brain Vision Engine
+
+Instead of basic classification, vidchain uses a two-pronged visual approach:
+
+  * **The "Noun" Engine (YOLOv8):** Detects specific objects (e.g., "1 person, 2 laptops").
+  * **The "Verb" Engine (MobileNetV3):** Classifies the intent or state of the scene (e.g., NORMAL, SUSPICIOUS, VIOLENCE).
+
+### 🔤 Context-Aware OCR
+
+Powered by EasyOCR, the system intelligently scans for text *only* when YOLO detects readable surfaces (monitors, laptops, books, whiteboards), saving massive compute power while capturing ground-truth data (e.g., reading the brand "ASUS Vivobook" off a laptop).
+
+### B.A.B.U.R.A.O. RAG Engine (Conversational)
+
+Unlike standard RAGs that read out robotic timelines, B.A.B.U.R.A.O. acts as a human copilot:
+
+  * **Abductive Reasoning:** If it sees a "laptop" and a "keyboard", it deduces the scene is a "computer desk."
+  * **Sensor Filtering:** Automatically ignores momentary hardware glitches/hallucinations (e.g., a TV briefly misidentified as an oven).
+  * **Natural Translation:** Translates raw model labels like `VIOLENCE` into contextual human behaviors like "the person became visibly frustrated and hit the desk."
+
+### Edge-First GPU Optimization
+
+Engineered to prevent VRAM crashes. Smart memory routing disables PyTorch's buggy layer fusion during YOLO inference and safely manages VRAM across concurrent vision, audio, and language models.
+
+-----
+
+## Installation
+
+```bash
+# 1. Install the core package
+pip install vidchain
+
+# 2. IMPORTANT: Install GPU-accelerated PyTorch (CUDA 12.1 recommended)
+pip install torch torchvision torchaudio --index-url [https://download.pytorch.org/whl/cu121](https://download.pytorch.org/whl/cu121) --force-reinstall
+```
+
+> ⚠️ **Requirement:** NVIDIA drivers and CUDA are strongly recommended. To verify your hardware is correctly mapped, run the built-in diagnostic script: `python scripts/check_gpu.py`
+
+-----
+
+## Quick Start
+
+### 1 — Analyze a Video (Build Knowledge Base)
+
+Analyze a video file, extract multimodal context, and generate a structured JSON timeline:
+
+```bash
+vidchain-analyze sample.mp4
+```
+
+*This command automatically builds a FAISS index and drops you into the interactive B.A.B.U.R.A.O. chat terminal.*
+
+### 2 — Train the Action Engine
+
+Fine-tune the MobileNetV3 "Verb" classifier on your domain-specific dataset:
+
+```bash
+vidchain-train
+```
+
+Place labeled training images under `data/train/` before running.
+
+-----
+
+## Knowledge Base Schema
+
+The framework utilizes **Semantic Chunking** to compress repetitive frames. The `knowledge_base.json` outputs a clean, fused timeline:
+
+```json
+{
+    "time": 0.97,
+    "type": "ocr",
+    "content": "ASUS Vivabook"
+},
+{
+    "time": 3.87,
+    "type": "visual",
+    "content": "Duration: [3.87s - 6.77s] | Subjects: 1 laptop, 1 tv | Action State: SUSPICIOUS"
+},
+{
+    "time": 19.34,
+    "type": "visual",
+    "content": "Duration: [19.34s - 19.34s] | Subjects: 1 tv | Action State: VIOLENCE"
+}
+```
+
+-----
+
+## Tech Stack
+
+| Component | Technology |
+|---|---|
+| Object Detection (Nouns) | YOLOv8s |
+| Intent Classification (Verbs) | MobileNetV3 (Custom fine-tuned) |
+| Text Extraction (OCR) | EasyOCR |
+| ASR (Audio) | OpenAI Whisper (Base) |
+| Vector Database | FAISS + Sentence-Transformers (`all-MiniLM-L6-v2`) |
+| LLM Routing | LiteLLM (`gemini-2.5-flash` default, Ollama supported) |
+| GPU Runtime | CUDA 12.1 (Optimized for 4GB+ VRAM) |
+
+-----
+
+## Roadmap
+
+  - [ ] **Real-time streaming pipeline** — live ingestion and indexing with low-latency event detection.
+  - [ ] **Advanced temporal reasoning** — multi-clip reasoning and cross-camera subject tracking.
+  - [ ] **Interactive Dashboard** — PyQt5 HUD for video playback, timeline visualization, and KB exploration.
+
+-----
+
+## Contributing
+
+Contributions, issues, and feature requests are highly welcome\! Open a GitHub issue or submit a pull request.
+
+-----
+
+## Author
+
+**Rahul Sharma** — B.Tech CSE, IIIT Manipur
+
+## License
+
+Distributed under the [MIT License](https://www.google.com/search?q=LICENSE).

vidchain-0.2.0/VidChain.egg-info/PKG-INFO

@@ -0,0 +1,184 @@
+Metadata-Version: 2.4
+Name: VidChain
+Version: 0.2.0
+Summary: A Lightweight Video RAG Framework for Multimodal Reasoning
+Author-email: Rahul Sharma <rahulsharma.hps@gmail.com>
+License: MIT
+Project-URL: Homepage, https://github.com/rahulsiiitm/videochain-python
+Project-URL: Bug Tracker, https://github.com/rahulsiiitm/videochain-python/issues
+Keywords: video-rag,multimodal,ai,computer-vision,whisper,ollama,yolo
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Developers
+Classifier: Intended Audience :: Science/Research
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Requires-Python: >=3.11
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: opencv-python
+Requires-Dist: scenedetect
+Requires-Dist: openai-whisper
+Requires-Dist: sentence-transformers
+Requires-Dist: faiss-cpu
+Requires-Dist: litellm
+Requires-Dist: google-generativeai
+Requires-Dist: moviepy
+Requires-Dist: pillow<12.0
+Requires-Dist: torch
+Requires-Dist: torchvision
+Requires-Dist: torchaudio
+Requires-Dist: numpy
+Requires-Dist: tqdm
+Requires-Dist: python-dotenv
+Requires-Dist: ultralytics
+Requires-Dist: librosa
+Requires-Dist: easyocr
+Dynamic: license-file
+
+# VidChain: Video Intelligence RAG Framework
+> Edge-optimized multimodal RAG framework for video understanding — transforms raw footage into a structured, queryable knowledge base.
+
+![Python](https://img.shields.io/badge/Python-3.11+-blue) ![CUDA](https://img.shields.io/badge/CUDA-12.1-green) ![License](https://img.shields.io/badge/License-MIT-yellow) ![Status](https://img.shields.io/badge/Status-beta-orange) [![PyPI version](https://badge.fury.io/py/vidchain.svg)](https://pypi.org/project/VidChain/)
+
+---
+
+## Overview
+
+vidchain v0.2.0 is a lightweight, modular framework that combines computer vision, smart OCR, speech recognition, and LLM reasoning into a unified **late-fusion pipeline**. Designed to run efficiently on consumer-grade GPUs (tested on NVIDIA RTX 3050), it extracts human-readable stories from raw sensor data, making on-device video intelligence practical without massive cloud dependency.
+
+At the heart of the framework is **B.A.B.U.R.A.O.** (*Behavioral Analysis & Broadcasting Unit for Real-time Artificial Observation*), an elite AI copilot that uses abductive reasoning to translate raw, flickering object/action logs into flowing, conversational narratives.
+
+---
+
+## Core Pipeline
+```text
+Video Input → Adaptive Keyframes → Dual-Brain Vision (YOLO + MobileNet) + OCR → Audio Transcription → Semantic Chunking → FAISS Vector DB → B.A.B.U.R.A.O. RAG
+````
+
+-----
+
+## Key Capabilities
+
+### 🧠 Dual-Brain Vision Engine
+
+Instead of basic classification, vidchain uses a two-pronged visual approach:
+
+  * **The "Noun" Engine (YOLOv8):** Detects specific objects (e.g., "1 person, 2 laptops").
+  * **The "Verb" Engine (MobileNetV3):** Classifies the intent or state of the scene (e.g., NORMAL, SUSPICIOUS, VIOLENCE).
+
+### 🔤 Context-Aware OCR
+
+Powered by EasyOCR, the system intelligently scans for text *only* when YOLO detects readable surfaces (monitors, laptops, books, whiteboards), saving massive compute power while capturing ground-truth data (e.g., reading the brand "ASUS Vivobook" off a laptop).
+
+### B.A.B.U.R.A.O. RAG Engine (Conversational)
+
+Unlike standard RAGs that read out robotic timelines, B.A.B.U.R.A.O. acts as a human copilot:
+
+  * **Abductive Reasoning:** If it sees a "laptop" and a "keyboard", it deduces the scene is a "computer desk."
+  * **Sensor Filtering:** Automatically ignores momentary hardware glitches/hallucinations (e.g., a TV briefly misidentified as an oven).
+  * **Natural Translation:** Translates raw model labels like `VIOLENCE` into contextual human behaviors like "the person became visibly frustrated and hit the desk."
+
+### Edge-First GPU Optimization
+
+Engineered to prevent VRAM crashes. Smart memory routing disables PyTorch's buggy layer fusion during YOLO inference and safely manages VRAM across concurrent vision, audio, and language models.
+
+-----
+
+## Installation
+
+```bash
+# 1. Install the core package
+pip install vidchain
+
+# 2. IMPORTANT: Install GPU-accelerated PyTorch (CUDA 12.1 recommended)
+pip install torch torchvision torchaudio --index-url [https://download.pytorch.org/whl/cu121](https://download.pytorch.org/whl/cu121) --force-reinstall
+```
+
+> ⚠️ **Requirement:** NVIDIA drivers and CUDA are strongly recommended. To verify your hardware is correctly mapped, run the built-in diagnostic script: `python scripts/check_gpu.py`
+
+-----
+
+## Quick Start
+
+### 1 — Analyze a Video (Build Knowledge Base)
+
+Analyze a video file, extract multimodal context, and generate a structured JSON timeline:
+
+```bash
+vidchain-analyze sample.mp4
+```
+
+*This command automatically builds a FAISS index and drops you into the interactive B.A.B.U.R.A.O. chat terminal.*
+
+### 2 — Train the Action Engine
+
+Fine-tune the MobileNetV3 "Verb" classifier on your domain-specific dataset:
+
+```bash
+vidchain-train
+```
+
+Place labeled training images under `data/train/` before running.
+
+-----
+
+## Knowledge Base Schema
+
+The framework utilizes **Semantic Chunking** to compress repetitive frames. The `knowledge_base.json` outputs a clean, fused timeline:
+
+```json
+{
+    "time": 0.97,
+    "type": "ocr",
+    "content": "ASUS Vivabook"
+},
+{
+    "time": 3.87,
+    "type": "visual",
+    "content": "Duration: [3.87s - 6.77s] | Subjects: 1 laptop, 1 tv | Action State: SUSPICIOUS"
+},
+{
+    "time": 19.34,
+    "type": "visual",
+    "content": "Duration: [19.34s - 19.34s] | Subjects: 1 tv | Action State: VIOLENCE"
+}
+```
+
+-----
+
+## Tech Stack
+
+| Component | Technology |
+|---|---|
+| Object Detection (Nouns) | YOLOv8s |
+| Intent Classification (Verbs) | MobileNetV3 (Custom fine-tuned) |
+| Text Extraction (OCR) | EasyOCR |
+| ASR (Audio) | OpenAI Whisper (Base) |
+| Vector Database | FAISS + Sentence-Transformers (`all-MiniLM-L6-v2`) |
+| LLM Routing | LiteLLM (`gemini-2.5-flash` default, Ollama supported) |
+| GPU Runtime | CUDA 12.1 (Optimized for 4GB+ VRAM) |
+
+-----
+
+## Roadmap
+
+  - [ ] **Real-time streaming pipeline** — live ingestion and indexing with low-latency event detection.
+  - [ ] **Advanced temporal reasoning** — multi-clip reasoning and cross-camera subject tracking.
+  - [ ] **Interactive Dashboard** — PyQt5 HUD for video playback, timeline visualization, and KB exploration.
+
+-----
+
+## Contributing
+
+Contributions, issues, and feature requests are highly welcome\! Open a GitHub issue or submit a pull request.
+
+-----
+
+## Author
+
+**Rahul Sharma** — B.Tech CSE, IIIT Manipur
+
+## License
+
+Distributed under the [MIT License](https://www.google.com/search?q=LICENSE).

vidchain-0.2.0/VidChain.egg-info/SOURCES.txt

@@ -0,0 +1,26 @@
+LICENSE
+README.md
+pyproject.toml
+VidChain.egg-info/PKG-INFO
+VidChain.egg-info/SOURCES.txt
+VidChain.egg-info/dependency_links.txt
+VidChain.egg-info/entry_points.txt
+VidChain.egg-info/requires.txt
+VidChain.egg-info/top_level.txt
+vidchain/__init__.py
+vidchain/cli.py
+vidchain/processor.py
+vidchain/rag.py
+vidchain/vision.py
+vidchain/core/__init__.py
+vidchain/core/fusion.py
+vidchain/core/ollama_engine.py
+vidchain/llm/__init__.py
+vidchain/loaders/__init__.py
+vidchain/loaders/audio_loader.py
+vidchain/loaders/video_loader.py
+vidchain/processors/__init__.py
+vidchain/processors/audio_model.py
+vidchain/processors/ocr_model.py
+vidchain/processors/vision_model.py
+vidchain/storage/__init__.py

vidchain-0.2.0/VidChain.egg-info/dependency_links.txt

@@ -0,0 +1 @@
+

vidchain-0.2.0/VidChain.egg-info/entry_points.txt

@@ -0,0 +1,3 @@
+[console_scripts]
+vidchain-analyze = vidchain.cli:main
+vidchain-train = scripts.train_vision:main

vidchain-0.2.0/VidChain.egg-info/requires.txt

@@ -0,0 +1,18 @@
+opencv-python
+scenedetect
+openai-whisper
+sentence-transformers
+faiss-cpu
+litellm
+google-generativeai
+moviepy
+pillow<12.0
+torch
+torchvision
+torchaudio
+numpy
+tqdm
+python-dotenv
+ultralytics
+librosa
+easyocr

vidchain-0.2.0/VidChain.egg-info/top_level.txt

@@ -0,0 +1 @@
+vidchain

vidchain-0.2.0/pyproject.toml

@@ -0,0 +1,65 @@
+[build-system]
+requires = ["setuptools>=61.0"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "VidChain"
+version = "0.2.0"
+description = "A Lightweight Video RAG Framework for Multimodal Reasoning"
+readme = "README.md"
+requires-python = ">=3.11"
+authors = [{ name = "Rahul Sharma", email = "rahulsharma.hps@gmail.com" }]
+license = { text = "MIT" }
+keywords = ["video-rag", "multimodal", "ai", "computer-vision", "whisper", "ollama", "yolo"]
+
+# Detailed Classifiers for PyPI
+classifiers = [
+    "Development Status :: 3 - Alpha",
+    "Intended Audience :: Developers",
+    "Intended Audience :: Science/Research",
+    "License :: OSI Approved :: MIT License",
+    "Programming Language :: Python :: 3.11",
+    "Topic :: Scientific/Engineering :: Artificial Intelligence",
+]
+
+# Core Dependencies
+dependencies = [
+    "opencv-python",
+    "scenedetect",
+    "openai-whisper",       # Provides the 'whisper' import
+    "sentence-transformers",
+    "faiss-cpu",
+    "litellm",              # Added for your 'gemini/gemini-2.5-flash' routing
+    "google-generativeai",
+    "moviepy",
+    "pillow<12.0",          # 🛑 PINNED: Fixes the MoviePy incompatibility crash
+    "torch",                # 🛑 UNPINNED: Allows users to install their own cu118/cu121 versions
+    "torchvision",          # 🛑 UNPINNED: Prevents pip from forcing a CPU downgrade
+    "torchaudio",           # Added to complete the PyTorch hardware triad
+    "numpy",
+    "tqdm",
+    "python-dotenv",
+    "ultralytics",          # YOLO-based Noun Engine
+    "librosa",              # Audio energy extraction
+    "easyocr",              # GPU-accelerated text extraction
+]
+
+# --- CONSOLIDATED CLI ENTRY POINTS ---
+[project.scripts]
+vidchain-analyze = "vidchain.cli:main"
+vidchain-train = "scripts.train_vision:main"
+
+[project.urls]
+"Homepage" = "https://github.com/rahulsiiitm/videochain-python"
+"Bug Tracker" = "https://github.com/rahulsiiitm/videochain-python/issues"
+
+[tool.setuptools.packages.find]
+where = ["."]
+include = ["vidchain*"]
+
+[tool.setuptools.package-data]
+vidchain = ["py.typed"]
+
+# UV specific configuration for CUDA-enabled Torch
+[[tool.uv.index]]
+url = "https://download.pytorch.org/whl/cu121"

vidchain-0.2.0/setup.cfg

@@ -0,0 +1,4 @@
+[egg_info]
+tag_build = 
+tag_date = 0
+

vidchain-0.2.0/vidchain/__init__.py

@@ -0,0 +1,7 @@
+from .core.fusion import FusionEngine
+from .processors.vision_model import VisionEngine
+from .processors.audio_model import AudioProcessor
+from .loaders.video_loader import VideoLoader
+from .loaders.audio_loader import AudioLoader
+
+__version__ = "0.1.0"