Differences

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

@@ Line 1: / Line 1: @@
+{{htmlmetatags>metatag-keywords=(pio reply checker,rti reply analysis,rti appeal grounds finder,rti reply checker tool,pio refusal valid,rti reply scrutiny,rti tool free)&metatag-description=(Paste a PIO reply. The tool flags invalid grounds, deemed refusal, missing §10 severability, §11 procedure skips, fee over-demands — each with case-law authority and next step.)}}
+====== PIO Reply Checker ======
+{{url>https://righttoinformation.wiki/tools/pio-reply-checker-app.html 100%x1500px noborder|PIO Reply Checker}}
+<WRAP center round help 95%>
+**Tool not loading?** Open it directly at [[https://righttoinformation.wiki/tools/pio-reply-checker-app.html|/tools/pio-reply-checker-app.html]].
+</WRAP>
+===== What this tool does =====
+Paste the PIO's reply into the text box. In seconds the tool returns:
+  * **A verdict** — strong grounds for appeal, likely appealable, or legally sound.
+  * **A findings list** — each invalid-ground pattern identified, categorised by severity.
+  * **Statutory + case-law authority** for every finding.
+  * **A concrete next-step action** per finding — what to add to your appeal.
+  * **Timeline block** — when your RTI was filed, when the 30-day window expired, when your First Appeal is due.
+  * **Direct CTA** to the First Appeal Builder (Tool 2) when appeal is warranted.
+All analysis runs in your browser. Nothing is sent to any server.
+===== What the tool checks =====
+==== Invalid refusals (high severity) ====
+  * **§7(2) deemed refusal** — date comparison between your RTI and the reply.
+  * **§7(9) blanket refusal** — "voluminous" / "disproportionate" without offering inspection.
+  * **§8(1)(a) "Confidential" stamp** cited without a harm-test.
+  * **§8(1)(b) sub-judice** cited without an express court order.
+  * **§6(2) motive inquiry** — PIO asking why you want the information.
+  * **§7(8)(i) unreasoned refusal** — boilerplate or bald order.
+==== Questionable grounds (medium severity) ====
+  * **§8(1)(d) commercial** without identified third-party and harm.
+  * **§8(1)(e) fiduciary** applied to non-fiduciary records.
+  * **§8(1)(h) investigation** where the inquiry may be closed.
+  * **§8(1)(i) cabinet** for a decided matter.
+  * **§8(1)(j) personal** without §8(2) balancing (post-DPDP 2025).
+  * **"Not information" dismissal** without §5(3) assistance.
+  * **§11 third-party procedure** skipped.
+  * **§10 severability** missing (blanket refusal).
+  * **Fee demand** beyond statutory rates (Rs. 10 + Rs. 2/page).
+==== Positive signals ====
+  * Inspection offered (correct under §7(9)).
+  * Specific harm-test cited.
+  * §10 severability applied.
+  * First Appellate Authority named.
+===== How the verdict is computed =====
+  * **High-severity finding present →** "Strong grounds for appeal".
+  * **Two or more medium findings →** "Grounds for appeal likely".
+  * **One medium finding →** "One questionable refusal".
+  * **Only positive signals →** "Reply appears legally sound".
+  * **Nothing matched →** "No automatic triggers" — manual review required.
+The verdict is a heuristic — a well-reasoned PIO order can pass all 18 checks and still be factually wrong. Conversely, the tool may flag a reply that is actually valid. Always read the findings and use your judgement.
+===== Three-tool workflow =====
+  - **[[:tools:question-builder|RTI Question Builder]]** → map your problem to records.
+  - **[[:tools:generator|RTI Application Generator]]** → produce the RTI letter.
+  - **PIO reply arrives →** paste it into this tool.
+  - **[[:tools:first-appeal|First Appeal Builder]]** → draft the §19(1) appeal.
+All four tools are client-side and privacy-safe.
+===== Optional deep analysis with open-source AI — free =====
+Below the static findings, the tool offers an **opt-in AI deep analysis** — **free to use as of now, no sign-up, no API key.** It runs **entirely in your browser** via [[https://github.com/mlc-ai/web-llm|WebLLM]] — nothing is sent to any server.
+==== Choose a model size ====
+The tool now lets you pick one of three open-source Large Language Models:
+  * **Qwen 2.5 0.5B** — ~330 MB first-time download. Fastest. Basic quality; good for a quick second opinion.
+  * **Llama 3.2 1B** (default) — ~720 MB. Balanced speed and quality. The recommended starting point.
+  * **Llama 3.2 3B** — ~1.8 GB. Slower first download but produces the most detailed analysis.
+After the one-time download, the browser caches the model. Subsequent runs are instant.
+==== What the AI does ====
+Reads the PIO reply, identifies §8(1) clauses invoked, assesses validity of each, drafts specific appeal grounds with statutory and case-law authorities, and recommends the next step.
+==== Requirements and caveats ====
+  * **WebGPU** — Chrome 113+, Edge, or a WebGPU-capable browser. Available on most computers made after 2017. Safari / Firefox support is experimental.
+  * **Not legal advice.** A small LLM can miss subtleties or hallucinate. Treat the output as a second opinion and cross-check against [[:explanations:grounds-for-rejection|Grounds for RTI rejection]] and the static findings above.
+  * **Privacy.** The pasted reply stays on your device. The model's weights come from a public CDN; your text never leaves the browser.
+  * **"Free as of now"** means exactly that — the model is free open source, the inference is on your GPU so it costs us nothing, and we reserve the right to revisit if this changes.
+==== 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 the [[https://mlc.ai|MLC AI]] team.
+  * Runtime — [[https://github.com/mlc-ai/web-llm|WebLLM]] (Apache-2.0).
+===== Caveats =====
+  * The static checks run on pattern matching, not full legal reasoning. A well-drafted PIO order may defeat a weak regex; a sloppy order may contain valid grounds the tool misses.
+  * The AI deep analysis is a convenience, not a substitute for legal expertise. A small LLM can hallucinate or miss subtle points.
+  * The tool does not carry over state to the First Appeal Builder — you'll paste the reply again there (a future upgrade will wire them up).
+  * For sophisticated cases (complex §8(2) balancing, multi-party §11 scenarios, Schedule-II §24 agencies), consult a qualified RTI advocate. See our **[[:pio-faa-knowledge-base|PIO / FAA knowledge base]]** for the underlying framework.
+===== Related reading =====
+  * [[:tools:first-appeal|First Appeal Builder]]
+  * [[:tools:generator|RTI Application Generator]]
+  * [[:tools:question-builder|RTI Question Builder]]
+  * [[:explanations:grounds-for-rejection|Grounds for RTI rejection]]
+  * [[:pio-section-8-1-j-framework|PIO Framework — §8(1)(j) after DPDP]]
+  * [[:pio-section-7-9-alternative-form|§7(9) alternative form — not a ground]]
+  * [[:faa-appellate-review-checklist|FAA appellate-review checklist]]
+  * [[:faa-first-appeal-timelines|First Appeal timelines]]
+  * [[:faa-privacy-public-interest-balancing|Privacy vs public-interest balancing]]
+  * [[:pio-rti-reply-guide|PIO RTI Reply Guide]]
+===== Sources =====
+  * Right to Information Act, 2005 — §§ 6(2), 7(1), 7(2), 7(8)(i), 7(9), 8(1), 8(2), 10, 11, 19, 20
+  * RTI (Regulation of Fee and Cost) Rules, 2012
+  * Digital Personal Data Protection Act, 2023 — §44(3) amending RTI §8(1)(j)
+  * Supreme Court: //S.P. Gupta// (1982), //Aditya Bandopadhyay// (2011), //ICAI v. Shaunak Satya// (2011), //R.K. Jain// (2013), //Namit Sharma// (2013), //RBI v. Jayantilal Mistry// (2016), //Puttaswamy// (2017)
+  * High Courts: //Carbon Resources// (Kerala), //Muniyappan// (Madras), //S. Venkatesan// (Madras), //S. Mukherjee// (Calcutta), //Treesa Irish// (Kerala), //Shailesh Gandhi// (Bombay), //GMC Thiruvananthapuram// (Kerala), //C.V. Srinivasa// (Karnataka)
+----
+//Last reviewed: 21 April 2026.//
+{{tag>tool pio-reply-checker tools analysis drafting}}

Was this helpful? — views