docs: full README with proper HF structure

Author: aiKunalBisht

README.md

@@ -10,5 +10,82 @@ app_file: app.py
 pinned: false
 ---
 
-# TranscriptAI
-Japanese Business Intelligence Platform
+# TranscriptAI 🧠
+**Japanese Business Intelligence Platform — Speech & Meeting Analyzer**
+
+[![Live Demo](https://img.shields.io/badge/Live%20Demo-HuggingFace-D96080?style=flat-square)](https://huggingface.co/spaces/KunalTheBeast/TranscriptAI)
+[![GitHub](https://img.shields.io/badge/GitHub-Repo-3C2416?style=flat-square&logo=github)](https://github.com/aiKunalBisht/Transcript-ai)
+[![Python](https://img.shields.io/badge/Python-3.10%2B-blue?style=flat-square)](https://python.org)
+[![License](https://img.shields.io/badge/License-MIT-green?style=flat-square)](LICENSE)
+
+> Turns any meeting transcript into structured business intelligence in ~3 seconds.
+> Works with Otter.ai, Zoom, Google Meet, Whisperflow exports out of the box.
+
+---
+
+## How we got here — v1 to v5
+
+| Version | What changed | Accuracy |
+|---|---|---|
+| v1 | Hard exact matching, no cultural awareness | 22–30% |
+| v2 | Fuzzy names, rule-based code-switch, TF-IDF similarity | ~45% |
+| v3 | MeCab keigo detection, bilingual ground truth, soft sentiment | ~60% |
+| v4 | Hallucination guard, 16 nemawashi patterns, APPI PII masking | 75–85% |
+| **v5 (live)** | 2-key Groq rotation, vector cache, health score, trends | **85–95%** |
+
+---
+
+## Why Japanese business is different
+
+| What was said | Generic AI | TranscriptAI |
+|---|---|---|
+| 検討いたします | "We will consider it" — action item | ⚠ Likely soft rejection (72%) |
+| 難しいかもしれません | "May be difficult" — neutral | 🚨 High rejection signal (90%) |
+| 承知いたしました | Acknowledged | 🏯 High keigo — senior speaker |
+
+---
+
+## Features
+
+- **Trilingual** — Japanese / Hindi (Hinglish) / English / Mixed
+- **Keigo detection** via MeCab morphological analysis
+- **16 nemawashi soft rejection patterns** with confidence scores
+- **8 Hindi indirect patterns** (देखते हैं, थोड़ा मुश्किल है)
+- **APPI-compliant PII masking** before any LLM call
+- **Hallucination guard** — 100% rule-based, LLM never validates itself
+- **2-key Groq rotation** — auto-failover on 429 rate limit
+- **Vector cache** — instant return for similar transcripts
+- **Meeting health score** (0–100) with breakdown
+- **Trends dashboard** — soft rejection drift, provider usage over time
+- **Groq Whisper** — MP4/MP3/WAV transcription (same free key)
+- **FastAPI REST** endpoint for CRM integration
+
+---
+
+## Quick start
+
+```bash
+git clone https://github.com/aiKunalBisht/Transcript-ai.git
+cd Transcript-ai
+pip install -r requirements.txt
+export GROQ_API_KEY=your_key_here   # free at console.groq.com
+python -m streamlit run app.py
+```
+
+**HuggingFace Spaces:** Add `GROQ_API_KEY` in Space → Settings → Repository secrets.
+
+---
+
+## Evaluation
+
+| Test case | v1 baseline | v5 live |
+|---|---|---|
+| Sales call · JA/EN mixed | 30.8% | **95.2%** |
+| Internal meeting · Japanese | 22.2% | **81.6%** |
+| Client conflict · EN/JA | 55.9% | **85.8%** |
+
+---
+
+## Built by
+
+[Kunal Bisht](https://github.com/aiKunalBisht) · Benglore, Karnataka, India