Privacy Policy

1. Introduction & Commitment

Saathi Bot (“Saathi Bot”, “we”, “us”, or “our”) provides an embeddable, retrieval-augmented generation (RAG) conversational chatbot platform designed for modern websites and SaaS businesses. We operate from Bengaluru, Karnataka, India.

We believe customer data privacy is non-negotiable. This Privacy Policy details our operational data flows, the categories of personal data collected, why and how that data is processed, retention schedules, and your statutory rights under India's Digital Personal Data Protection Act, 2023 (DPDP Act 2023), the Information Technology Act, 2000, and where applicable, the General Data Protection Regulation (GDPR / UK GDPR).

2. Categories of Data Collected

Depending on whether you are an Account Holder (bot owner / organization member) or an End-User Visitor interacting with an embedded chat widget, we collect the following:

A. Account Holders & Organization Data

• Identity & Authentication: Full name, verified email address, organization name, avatar image, and authentication credentials managed securely via Supabase Auth.
• Knowledge Base Documentation & Content: Public URLs submitted for web scraping, sitemaps, uploaded PDF manuals, text files, and custom Q&A pairs.
• Bring-Your-Own-Key (BYOK) Credentials: When you provide your own API keys (Google Gemini, Groq, OpenAI, Claude, DeepSeek), these keys are stored either encrypted using AES-256 for background tasks or strictly in your browser's local memory via our Zero-DB vault guarantee.
• Configuration & Domain Whitelists: Allowed origin domains, widget styling parameters, launcher button colors, and custom system prompts.

B. End-User Visitors (Website Chat Widget)

• Chat Queries & Session Messages: The prompt questions submitted to the widget to generate grounded answers.
• Hashed Origin & IP Fingerprint: To enforce rate limits and detect DDoS/abuse, we compute a cryptographically salted one-way SHA-256 hash of the visitor's IP and user-agent string. We do not store raw, unhashed visitor IP addresses in plain text.
• Anonymous Visitor Identifier: A randomly generated UUID stored in browser localStorage to maintain conversation continuity across browser tab refreshes.

3. Purposes & Legal Basis for Processing

We process information strictly for the following operational needs:

• Executing the RAG Pipeline: Converting submitted documents into chunked vector embeddings (using Google's text-embedding-004), performing cosine similarity retrieval via PostgreSQL pgvector, and generating factual answers.
• Active Learning & Question Clustering: Aggregating unanswered or low-similarity questions (< 0.55 similarity threshold) into semantic clusters in the bot owner's dashboard so documentation gaps can be patched.
• Security, Abuse Prevention & Origin Enforcement: Verifying HTTP Origin and Referer headers against authorized domain whitelists to block unauthorized widget hijacking.
• Platform Health & Quota Monitoring: Measuring monthly message volume, bot counts, and latency to enforce plan quotas and ensure service stability.

We never sell, rent, or monetize your data, your uploaded documents, or visitor chat transcripts to third-party data brokers or advertising networks.

4. Third-Party Subprocessors & AI Processing

To deliver vector retrieval and generative responses, data passes through carefully vetted infrastructure providers under strict confidentiality:

• Supabase Inc.: Primary PostgreSQL database, Row-Level Security (RLS) enforcement, pgvector similarity search, file object storage, and secure authentication.
• Google LLC (Gemini API): Generation of 768-dimensional vector embeddings and generative completion for Gemini-configured bots. Customer data sent to Google Enterprise APIs is not used to train Google foundational models.
• Groq Inc.: Ultra-low latency Llama-3.3 inference execution when enabled by the bot configuration.
• Vercel Inc.: Application edge routing, Next.js serverless execution, and global CDN delivery.

For a complete breakdown of services, functions, and server locations, review our Third-Party Subprocessors Table.

5. Data Retention & Automatic Pruning

We retain data only as long as necessary to provide the service:

• Bot Embeddings & Sources: Retained while the bot and sources are active. When a source or bot is deleted by the organization owner, an automated relational cascade deletes all associated chunks, vector embeddings, domains, and conversations immediately from our primary database.
• Visitor Chat Transcripts: Retained for 30 to 90 days (depending on your plan configuration) to allow dashboard analysis, after which background cron jobs prune historical messages.
• AI Gateway Request Logs: Anonymized token counts, latency metrics, and error codes are retained for 30 days for operational debugging and cost calculation.

6. Your Statutory Rights (India DPDP Act 2023 & GDPR)

Under India's Digital Personal Data Protection Act, 2023 and European data protection laws, you possess enforceable rights over your personal data:

• Right to Access & Summary: You may request an export of all personal data held in your account.
• Right to Correction & Erasure: You can edit or delete profile information and purge any bot, document, or knowledge source directly from the dashboard.
• Right to Grievance Redressal: You have the right to seek redressal for any unresolved privacy or data processing grievance.
• Right to Nominate: As provided under the DPDP Act 2023, an individual may nominate another person to exercise rights in the event of death or incapacity.

7. Children's Privacy

Saathi Bot is a business-to-business (B2B) SaaS platform intended exclusively for commercial entities, developers, and individuals aged 18 and older. We do not knowingly collect personal data from minors. If you believe a minor has submitted personal information, contact us immediately.

8. Grievance Officer & Privacy Contact

In accordance with the DPDP Act 2023 and Information Technology Rules, the details of our Grievance Officer are: