FLE Studio 0.3.0

Plugins

• Plugins: installed Python packages can add a page to the sidebar, static files under /plugins/<id>/ and JSON
  routes under /api/p/<id>/ (entry point group fle_studio.plugins, see docs/PLUGINS.md). Every
  plugin is off until it is turned on in Settings → Plugins; the list lives in plugins.json in the data folder.
• Plugins belong to the machine they are installed on: a paired client of a host neither sees nor uses them, and the
  page always calls them on its own Studio, also while another host is chosen.
• A plugin that fails to load is shown in Settings and does not stop FLE Studio.

Fixes

• Engine → Best for my machine: "What to try and for how long" stays open while the page refreshes; before, a refresh
  folded it shut under the cursor.

FLE Studio 0.2.4

[0.2.4] - 2026-10-14

Fixes

• Remote: the block on pairing and host administration now also covers the bare /api/remote path and dot segments (.., %2e).
• Remote: large known-length answers are streamed instead of being read into memory whole (buffered up to 8 MiB).
• Tests: the fake forge tools no longer need an installed package.

Cleanup

• Removed four unused functions.

FLE Studio 0.2.3

Model menu

• Switch the model by hand from the model menu at the top right, for a service that runs next to a model switcher
  (setting switch_url, or the environment variable FLE_STUDIO_SWITCH_URL; for example Rhea and Katy on one GPU).
  The menu lists the models, marks the loaded one and switches on a click; the model that is loaded stays until you
  switch again. FLE Studio only asks: the switcher decides when it is safe (free slots, no measurement running) and
  FLE Studio shows the progress and, if it cannot switch, the reason. Without switch_url the menu is as before.

FLE Studio 0.2.2

Live chat

• Live: a fixed entry at the top of the chat list. Everything that goes through the /v1 address of FLE Studio (this chat page and any other program) shows up there as a temporary chat. Requests that repeat the earlier messages are joined into one conversation. It lives in memory only, keeps the newest 100 messages, and writes nothing to disk. Pictures are left out. The star on a conversation, or on one message, copies it (up to that message) into your normal chats as a favorite; chats that are favorites show a star in the list.

Capture (Settings → Capture) is unchanged and still off by default. The example host names in the remote-host docs and tests are now neutral.

Install or update as described in the README.

FLE Studio 0.2.1

Capture

• Capture (Settings → Capture, off by default): keeps every chat answer that goes through FLE Studio as one line
  of JSON in <data folder>/capture/, with the request, the answer, timings and accepted draft tokens, for building your
  own corpus. Streams are put back together, pictures are left out, files are private, a size limit stops the recording.

Fixed

• Forge: the folder typed into the tools box is no longer lost when the page redraws while the first search is still
  running.

FLE Studio 0.2.0

Not released yet. Studio 0.2 adds Forge, a chat with attachments, tools, MCP and apps, tuning and quality
measurements, an expert cache page, an update check and host/client mode.

Forge: build your own quant

• Forge accepts the named quantization mixtures of llama-quantize (q3_k_m, q4_k_m, ...) as the default type.
• Forge, a new page for building your own quant (see docs/FORGE.md):
  • a hybrid builder that takes whole expert layers from several quants of the same model, with a size-budget
    calculator (exact knapsack over the layers), the exact file size and a RAM/VRAM preview before anything is written;
    the file is written without requantizing, with a correct header, and can be resumed after a stop;
  • quantizing a full-precision model with llama-quantize (per-tensor rules, optional importance matrix from
    llama-imatrix);
  • optional REAP pruning: the best experts per layer are kept, the router and the metadata are made consistent;
  • a check with llama-perplexity (KL divergence, top-1 agreement, perplexity ratio) and a result card;
  • jobs with progress and a log, cancel, resume and a record that survives a restart of Studio;
  • recipes (JSON schema 1) that carry no folders and no commands, to save, share and load;
  • publishing a result to the library with a start profile.
• A calibration text that ships with Studio (about 1 MB: English and German prose, code, dialogue, numbers; public
  domain, CC0 or MIT, with the origin, license and checksum of every file in forge/corpus/SOURCES.md), used by Forge
  as the default, split into a calibration and a held-out comparison part.
• Tests with synthetic mini GGUF files and stand-in llama.cpp tools; the browser test walks through Forge.

Chat, tools and MCP

Details in docs/CHAT.md and docs/MCP.md.

• Attachments in the chat: text and code files, PDFs (optional pypdf, pip install "fle-studio[pdf]") and pictures
  (only when the loaded model reports vision; otherwise a clear message). Choose, drop or paste files. A token
  estimate shows against the context window and warns from 80 %. The upload limit is a setting (Settings → Chat →
  Largest upload, 25 MB by default) instead of a fixed refusal.
• Tool calling through llama-server's OpenAI-style tools field, with the model, tool, model loop in the page and
  an openable card for every call (arguments and result). Built-in tools, each one switchable: calculator, current
  time, read files in one folder you choose (read only, nothing outside it), and web search (off by default, asks
  before each search, DuckDuckGo or your own SearXNG address).
• An MCP client for stdio and Streamable HTTP servers, configured in Settings → Tools with the usual mcpServers
  JSON. Their tools appear in the chat. A tool of an MCP server asks for approval before every call unless you allow
  that server; the server side enforces this too. A autoApprove entry in the JSON is ignored on purpose.
• MCP, complete (see docs/MCP.md): web servers that need a sign-in get a Sign in button (OAuth per
  the MCP authorization specification 2025-11-25: 401 challenge, protected resource metadata, authorization server
  metadata, dynamic client registration or a client id you enter, PKCE S256, resource indicators, refresh, sign out).
  Tokens live in the data folder with mode 0600 and never in settings, logs or exports. The older HTTP+SSE transport is
  the fallback when Streamable HTTP is refused with a 4xx (or "type": "sse"). MCP Apps (SEP-1865, 2026-01-26): a tool
  with a ui:// resource shows an app under its card in a sandboxed iframe (no allow-same-origin, own CSP from the
  resource metadata) with the postMessage bridge: tool input and result, light/dark, resize, tools/call to its own
  server with the usual confirmation, a message into the chat and links only after you agree.
• Edit your own messages: the edit becomes another version and the model answers it. "Answer again" keeps the
  previous answer as a version too. A < 2 / 3 > switch moves between versions.
• Search over the titles and the text of all chats (every version, attachment names and attachment text).
• Export a chat as JSON (every version, pictures inline) or Markdown (the version on screen), and import both again,
  plus plain OpenAI-style message lists.
• GET /api/chat/capabilities, /api/tools, /api/attachments, /api/chats/search, /api/chats/<id>/export and
  /api/chats/import.
• Tests: the stand-in engine can return tool_calls and accepts pictures with FAKE_VISION=1, a stand-in MCP server
  (tests/fake_mcp_server.py), unit and integration tests for all of it, a Node check of the request builder, and
  browser steps for attachments, versions, export and import, search, tool cards and MCP approval.

Start, tuning and measuring

Details in docs/TUNING-AND-QUALITY.md.

• Simple and advanced Engine page. Simple (the default) shows the model, "Best for my machine" and Start; Advanced
  shows everything as before. Under the command, every flag and environment variable has a one-line explanation (also as a
  tooltip); the full command stays.
• Best for my machine: an auto-tuner that tries threads, expert cache or --n-cpu-moe, batch and context, measures
  each several times, keeps only what is clearly faster, and saves the winner as a launch profile. Probe run first,
  adjustable time budget, optional llama-bench screening for stock llama.cpp.
• Repeats with median and 95 % interval in Experiments, a comparison that says "within noise" when it is, and an
  A/B/A comparison of two profiles.
• Quality next to speed: KL divergence via llama-perplexity, and a knowledge probe with original CC0 questions
  (optional MKQA download). The speed test can ask the questions too.
• Expert cache page from the engine's log lines; docs/EXPERT-CACHE-DATA.md says what the
  engine would have to print for the per-layer view.
• Catalog: gpt-oss-20b and gpt-oss-120b (a second mixture-of-experts family, unmeasured).
• docs/HARDWARE-TESTS.md, docs/TUNING-AND-QUALITY.md.

Updates

• Update check: about once a day (and with Check now) FLE Studio asks GitHub for the latest release, without
  identifiers or telemetry. A discreet card in the sidebar says "FLE Studio X.Y.Z is available", with the release notes
  as plain text, a link and Later. Switchable in Settings → Updates, with a switch for pre-releases. Offline it stays
  silent.
• One-click update for installs made with uv tool (not in service mode): installs the release, then restarts with the
  same command line. A failed install keeps the old version running and shows the log. On Windows a helper script does
  the install after Studio has ended. Other installs get the fitting command to copy.
• The same notice reports a newer FLE engine build for the engines you installed.

Host and client

Details in docs/REMOTE.md.

• Host and client mode (docs/REMOTE.md). A host (--host) serves TLS with a self-made certificate
  whose fingerprint is shown and pinned by clients; clients pair with a 10-minute single-use code and get a revocable
  token of their own (list with Revoke in Settings → Remote access). --pair-code, --pair, --tls, --no-tls.
• Client switcher at the top of the sidebar ("This machine / host / Add host…"). Hosts are added by address and code,
  or by SSH target; Studio runs and watches ssh -N -L itself (keys only). All pages, streaming chat included, run
  against the chosen host through a generic pass-through, so pages added later work too. The window shows the host's
  name and colour and reports an unreachable host with Reconnect.
• Rights per client: chat, look and measure by default; Admin for settings, installs and models; installs and updates
  ask for a confirmation on the client. Pairing is rate limited; tokens are stored as hashes and never logged.
• Optional read-only local copy of a host's chats for offline reading.
• Everything that changes something on a host needs the Admin right for a paired client (tools and MCP servers, Forge jobs, tuning, quality runs, updates); chat, attachments and speed tests do not.

Changed

• Chats are stored as a tree of messages (version 2 of the file). Older chat files open unchanged. Pictures are kept
  as files next to the chats, not inside them. The size limit of one saved chat is 32 MB (was 4 MB).
• A request that carries tools and a forced answer format (response_format, json_schema or grammar) is refused
  by the proxy with HTTP 400, and the page never sends both: the model would not call the tools and would answer
  with substitute JSON. A test keeps this.
• The tools.py of the tuning work (the programs that ship with llama.cpp) is engine_tools.py now; tools.py
  holds the tools the model can call.
• ruff format is checked in CI, with a pinned ruff version.

FLE Studio 0.1.1

Installers, CI and docs. The app itself does not change in this release.

Added

• Both installers can install from a folder or a wheel file on disk: FLE_STUDIO_SOURCE=<path> sh install.sh, or
  install.ps1 -Source <path>. No GitHub, no git and no PyPI are needed for that.
• CI: a package job on macOS, Linux and Windows builds the wheel and the source archive, installs the wheel into a
  clean environment, starts it and checks that the web files are inside.
• CI: an installer job on macOS, Linux and Windows runs install.sh and install.ps1 from the checkout, so an
  installer change is tested before it is published.
• A test (tests/test_release.py) that keeps the version in the code, both installers, the README and this file
  the same.
• docs/SERVICE-MODE.md: running FLE Studio on a Linux server next to a systemd service
  (journal access, sudoers line, the two log options for the live panel, SSH tunnel, a user unit for Studio itself,
  and what the common messages mean).
• README: how to install from a local folder or wheel, and a pointer to the service-mode page.

Changed

• The installers install v0.1.1 by default (FLE_STUDIO_REF or -Ref still choose another release).

Not in this release

• New speed measurements and screenshots of the Rhea runs. The figures and pictures in the README and the model
  catalog are still the ones from 0.1.0.

FLE Studio 0.1.0

First public release of FLE Studio: a local chat and lab for llama.cpp and the FLE engine, on macOS, Linux and Windows.

What it does

• Chat in your browser with Markdown, code blocks, a collapsible reasoning block, streaming answers and the speed of each answer. History stays on your machine.
• Live inference data while the model answers: generation and prompt speed, draft/MTP acceptance, NVIDIA GPU use and the FLE engine's expert-cache hit rate.
• Model loader for the FLE quants of Qwen3.8-Flash-Next and any GGUF repo on Hugging Face, with a fit check before downloading, pause/resume and SHA-256 checks.
• Launch profiles with the full command line shown before anything starts.
• Picks a suitable engine (FLE CUDA build, or llama.cpp CUDA/Metal/Vulkan/CPU), or uses your own llama-server, a running server, or a systemd service.
• Experiments page with a short, repeatable speed test.

Install

macOS and Linux:

curl -LsSf https://raw.githubusercontent.com/sirfyyn/fle-studio/main/install.sh | sh

Windows (PowerShell):

irm https://raw.githubusercontent.com/sirfyyn/fle-studio/main/install.ps1 | iex

The installers need git, never ask for administrator rights and install this release (v0.1.0) as a uv tool.

Notes

• Early release: expect rough edges and please open an issue when you find one.
• No account, no telemetry; nothing leaves your machine except the downloads you start.
• Tested in CI on macOS, Linux and Windows (Python 3.10 and 3.13, plus a browser end-to-end run against a stand-in engine).
• MIT licensed.