Memory makes more room

Portail v1.6.0 is a memory release. It does not change the shape of the chat composer, add another button, or ask people to learn a new workflow. The point is simpler: long-running conversations should keep feeling useful after the first few good turns. When you use Portail as the place where many models share one thread, the memory layer has to stay sharp. It has to remember enough to help, forget enough to make room, and avoid turning yesterday's small detail into today's dead weight.

Enhanced Context already kept two kinds of memory. Active memory stores the compact facts that matter almost every time, like who you are, what you are building, and the preferences you have clearly stated. Passive memory stores a wider trail of useful details from actual conversations. That wider trail is where the pressure builds. A person can talk to Portail for weeks, ask GPT, Claude, Gemini, Qwen, or Llama to take different passes at the same project, and keep adding small facts along the way. If every old passive fact stays in full detail forever, the memory budget starts carrying too much scaffolding and not enough current signal.

v1.6.0 changes how that pressure is handled. Older passive memories can now be compressed into compact summaries while newer passive facts remain intact. The system keeps recent details close to their original wording because those are the details most likely to matter in the next few turns. Older entries still contribute, but in a denser form that leaves more room for the active conversation and the facts that just arrived. The result should feel quiet when it works well: fewer moments where a long account history crowds out what you just said, and fewer moments where a model needs you to restate context that Portail should already know.

This matters most for people who treat Portail as one durable workspace rather than a tab for disposable questions. A product plan can move from rough notes to launch copy. A coding discussion can pass from architecture to implementation to review. A language-learning thread can keep your preferred correction style while still remembering the topic of the current lesson. The memory layer sits underneath all of that. It should not get sentimental about every line it has ever stored. It should make a clean, practical judgment about what deserves space now.

The release also makes the version history itself public. Starting with this entry, Portail releases live at /versions, with one indexed page per version, screenshots from the product, structured data, and sitemap entries. That gives users, search engines, and future contributors a stable record of what changed. Patch releases can stay brief. Minor releases get enough room to explain the reasoning. Major releases will get the longer write-up they deserve. The release modal still handles in-app announcements for major and minor updates, but the version log is now the permanent record.

There is a small operational change behind that public archive too. Future deploys now validate release posts before shipping. A major version must have more than 1,000 words, a minor version must have more than 500, and a patch must have more than 250. Each entry also needs at least one screenshot file that exists in the app's public assets. This is deliberately strict because release history tends to decay when it relies on memory, haste, or a heroic last-minute note. Portail is better served by a small rule that catches omissions before production does.

So v1.6.0 is two things at once: a smarter memory layer and a firmer public record of changes. One helps the product carry context with less waste. The other helps everyone see what changed, when it changed, and why it mattered. Neither should shout. Both should make Portail feel more dependable the next time you come back to a thread and expect the work to still be there.