Differences

This shows you the differences between two versions of the page.

@@ Line 1: / Line 1: @@
+{{htmlmetatags>metatag-keywords=(rti wiki ai,rti ai chat,rti chatbot,rti chat india,rti question answer tool,rti ai assistant,rti chat free,ai rti helper,rti bot free)&metatag-description=(Free RTI chatbot. Ask anything about India's RTI Act — runs entirely in your browser with an open-source LLM. No sign-up, no API key, no data leaves your device.)}}
+====== RTI Wiki AI — Free, Private, Open Source ======
+{{url>https://righttoinformation.wiki/tools/chat-app.html 100%x700px noborder|RTI Wiki AI}}
+<WRAP center round help 95%>
+**Tool not loading?** Open directly at [[https://righttoinformation.wiki/tools/chat-app.html|/tools/chat-app.html]].
+</WRAP>
+===== What this is =====
+An **RTI Wiki AI — chat about the RTI Act, 2005** that runs **entirely in your browser**. Ask questions in plain English — fees, timelines, grounds for refusal, case law, appeal procedure, state-specific rules, DPDP amendments — and the chatbot answers. **Free as of now.** No sign-up. No API key. No data leaves your device.
+===== How it works =====
+  - You pick an open-source model size (330 MB fastest, 720 MB balanced, 1.8 GB best).
+  - First use downloads the model once (cached in your browser).
+  - You type a question.
+  - The model runs on your device's GPU via **WebGPU** and streams an answer.
+  - The chat remembers the last few turns for context.
+The model is **primed with the RTI Act framework** — all 10 grounds of rejection, §19 appeals, §20 penalties, key sections, post-DPDP-2025 amendments, and 10+ landmark Supreme Court / High Court cases — before your question ever reaches it.
+===== What it is NOT =====
+  * **Not legal advice.** A 1-3 billion parameter model can be wrong. Cross-check with statutory text and the curated pages.
+  * **Not up-to-the-minute.** The model's prior knowledge is from its training date; for the latest rules, check our [[:blog:start|blog]].
+  * **Not a document-aware assistant yet.** Phase 2 (coming) will add retrieval-augmented generation over all 300+ pages of this site.
+===== Requirements =====
+  * **WebGPU** — Chrome 113+, Edge, or any WebGPU-capable browser. Safari / Firefox support is experimental.
+  * **First run download** — 330 MB to 1.8 GB depending on model choice. One-time; cached thereafter.
+===== Privacy =====
+  * The **question you type never leaves your browser.**
+  * The **model weights** come from a public CDN (HuggingFace via WebLLM), but only once — cached afterwards.
+  * **No tracking, no analytics, no logging** of your questions on our side.
+===== Attribution =====
+  * **Qwen 2.5** by [[https://qwenlm.github.io/|Alibaba Qwen]].
+  * **Llama 3.2** by [[https://www.llama.com/|Meta]] under the Llama community licence.
+  * **Quantization** by [[https://mlc.ai|MLC AI]].
+  * **Runtime** — [[https://github.com/mlc-ai/web-llm|WebLLM]] (Apache-2.0).
+===== Related tools =====
+  * **[[:tools:question-builder|Question Builder]]** — describe a problem, get the records to ask for.
+  * **[[:tools:generator|RTI Application Generator]]** — produce a ready-to-file RTI.
+  * **[[:tools:pio-reply-checker|PIO Reply Checker]]** — analyse a PIO reply (also has an AI deep-analysis option).
+  * **[[:tools:first-appeal|First Appeal Builder]]** — if the PIO refuses, draft the §19(1) appeal.
+===== Related reading =====
+  * [[:explanations:grounds-for-rejection|Grounds for RTI rejection]]
+  * [[:blog:how-to-write-rti-application|Ask for records, not answers]]
+  * [[:faq|RTI FAQ — 25 most-asked questions]]
+  * [[:file-rti-online-india|How to file RTI online]]
+  * [[:landmark-cic-decisions|10 landmark CIC decisions]]
+===== Sources =====
+  * Right to Information Act, 2005 (as amended 2019, 2023)
+  * Digital Personal Data Protection Act, 2023 — §44(3)
+  * Meta Llama 3.2 community licence; Alibaba Qwen 2.5; MLC AI; WebLLM (Apache-2.0)
+----
+//Last reviewed: 21 April 2026.//
+{{tag>tool chat ai rag webllm tools}}

Was this helpful? — views