Search & fan-out capture 6.0.1

Oct 2026 : Now captures ChatGPT Ads + Gemini capture For each question asked in Gemini while signed in, it records: - the fan-out queries Gemini sends to Google; - the pages consulted, listed per search; - the pages cited in the answer, with the highlighted passage and the answer sentences anchored to each page; - the question classifier scores (topic, task, follow-up type, sentiment); - places, YouTube videos, products, flights, hotels, images and generated media shown in the answer August 2026: V5.4 - Fix - window.fetch interception is now accessor-based with a re-entrancy guard, so SearchGPT is never silently evicted from the chain when another MAIN-world extension hooks fetch first. A man whose name I shall not disclose, but whose first name is Suganthan 🦊, broke our plugin - window.fetch is now under shared custody, and no fan-out was harmed 🤝 V5.3 - Now captures logged-out ChatGPT conversations (sources, products, places) alongside logged-in ones. v5.2.0 - New "Thinking mode" selector in the Prompt Loader: force Think (free tier), Instant, Thinking medium or high (paid) before each prompt, with official ChatGPT UI labels supported across 36 languages. If the mode cannot be applied, the prompt is still sent using the interface's default mode, flagged with a diagnostic, and the actually served model slug and thinking effort are now shown under each analyzed conversation. July 2026 : Version 5.1.2 : - "Listed → promoted → cited" funnel per conversation: side panel URLs → top section (main + supporting sources) → citation-pill lead source; absolute + % of panel; carousels counted separately. - Aggregated pyramid on Dashboard: cumulated over filtered conversations, "Result nature" selector; panel/top/lead/opened per nature + within-nature promotion rates. - 11 named, colored result natures replacing raw ref_type (web search, news, scientific, forum, video, local, product, image, weather, opened page, no ref_type). - Explicit link-type labels: citations (sidebar top), more (sidebar), footnote (strip), widgets, opened (fetched), other (residual). - "Pages opened by the model" card: fetch marker Total lines: <n> + view refs, deduped by normalized URL, fetch signature kept. - Missing-snippet diagnosis: says why empty (opened page / pipeline serves none / rarely serves / served empty). - 🐕 labrador badge: OpenAI own-index results now inferred from format signatures (snippet length ≥186 chars threshold, all-caps snippet header) — needed since result_source vanished from SSE (July 20–22); signals shown in tooltip. Version 3.4.5 : regression : no more Sourcing pipeline as Openai dumped it from source code.. Version 3.4.4 Dashboard filters & richer exports • New global Dashboard filters: Model type, Model slug, Thinking effort, Sourcing pipeline — applied across all widgets and exports. • New "All Fan-outs" copy box: one fan-out per line, filterable by type (Search / Images / Shopping), one-click copy. Reacts to all active filters. • New "Fetched" filter in the Top URLs table: shows only URLs actually opened by the model. • Exports now respect the active filters, with new columns (Model, Model slug, Thinking effort, Pipelines, Fetched). • "Export All Data (JSON)" enriched: full model/reasoning/pipeline metadata, detailed fan-outs, fetched URLs, plus citations/sources/products/images previously missing. + fixes several parsing bugs, notably for Thinking-mode conversations. June 2026 : V3.4 Following the research by Suganthan Mohanadasan, we've extended the plugin to capture three new signals from ChatGPT's search conversations. We also took the opportunity to fix a few bugs along the way. What's new: - Turn Use Case = We now surface the server-side turn_use_case field, which reveals the intent category ChatGPT assigns to each turn (e.g. shopping). It tells you how the model framed the request before answering, a useful signal for understanding when a query triggers commercial vs. informational behavior. - Reference Types = A breakdown of the inline citation markers embedded in the response (e.g. product (9), search (3), news (1)). Parsed directly from the answer's reference tokens, it shows the mix of source types ChatGPT actually cited - products, web search results, news, images, and so on - and in what proportion. - Sourcing pipelines = Each cited web result carries a result_source field identifying the data provider that fetched it. [deprecated, ChatGPT no longer shows these flags] Mars 2026 : V3.3 - This release brings full support for GPT-5.3 model, which no longer show fan-out search queries but still leverage image and shopping fan-outs, and GPT-5.4 thinking & extended thinking. The analyzer now correctly handles this new behavior and adjusts statistics accordingly. A brand-new Map Carousel block has been added, fully recovering and displaying map/local results returned by ChatGPT's browsing engine. Under the hood, Chain-of-Thought detection and Thinking Summaries have been significantly improved, with better extraction of reasoning steps, summaries, and thought content. Entity identification has been refined, generic placeholders are now filtered out, and entity types (products, places, organizations…) are more accurately classified. The Technical Information panel now reports all turn reference types (search, cite, product, forecast…) with counts, and the Top URLs table includes a new Ref column showing each link's role in the search result groups. Jan 2026 : V3.1 new features: • Enhanced Entity Recognition - Entities now include disambiguation context (e.g., "Paris, Texas" vs "Paris, France"), providing richer semantic understanding of ChatGPT's content interpretation • Raw Conversation Debug - New debug tool to view the complete reconstructed response text, perfect for advanced users analyzing SSE streams and troubleshooting data capture • Fan-out Tag Cloud - Visual word cloud displaying the most frequent terms across all fan-out queries with dynamic…