Show what Cortex actually looks like, and stop describing it as Ollama-only - #119
Merged
Conversation
…a-only The README opened by claiming screenshots were "intentionally omitted while the curated capture set is being finalized" and then described an application that only runs Ollama models. Both had drifted from the code. Adds three captures of the real build -- the workspace, the AI Model settings section, and the command palette -- taken at 1440x900 on a 2x device scale. The workspace shot is framed and hovered so the per-message footer is visible: timestamp, token count, tokens/sec, and the copy, regenerate, and fork controls that are opacity:0 until a card is hovered. The conversations behind them are a fixed demo workspace, not live output. No model is contacted during a capture and every input is pinned -- installed inventory, settings, message timestamps, locale, and timezone -- so re-running ./tools/screenshots/capture.ps1 reproduces the same images instead of a fresh set that has to be eyeballed. The images are the real UI; only the conversation content is fixture data, and the README says so rather than implying otherwise. Also documents the local GGUF runtime, which the README had never mentioned: llama.cpp models served by a fetched and SHA-256-verified llama-server, GPU backend selection, downloading a GGUF by URL or Hugging Face repo, and the fact that Ollama is no longer a hard requirement. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Rebasing onto main pulled in the system-prompt bypass and the local quality script, which made two parts of this change stale before it merged. The AI Model capture is reframed onto the lower half of the panel -- generation parameters, system instructions, and the new bypass toggle. Model selection is already visible in the composer of the workspace shot, so nothing is lost by scrolling past the picker here. Development checks now point at ./scripts/check.ps1 and the pre-push hook rather than listing the individual commands by hand. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
dovvnloading
force-pushed
the
docs/readme-screenshots
branch
from
August 10, 2026 14:29
648b5ce to
fb6d163
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Why
The README opened by saying screenshots were "intentionally omitted while the curated capture set is being finalized", and then described an app that only runs Ollama models. Both had drifted from the code.
Screenshots
Three captures of the real build at 1440x900, 2x device scale:
workspace.pngsettings.pngcommand-palette.pngCtrl/Cmd+Kover the workspaceThe workspace shot is deliberately framed and hovered so the per-message footer is visible — timestamp, token count, tokens/sec, and the copy / regenerate / fork controls. Those are
opacity: 0until a card is hovered, so a naive screenshot silently omits them.Deterministic by construction
The conversations are a fixed demo workspace (
tools/screenshots/demo_server.py), not live model output. No model is contacted during a capture, and every input that could drift is pinned:statslocale: en-USandtimezoneId: UTC— message times go throughIntl, so an unpinned zone would change the images per machineRe-running
./tools/screenshots/capture.ps1reproduces the same images rather than a fresh set that has to be re-reviewed.The README states this outright rather than implying the transcripts are genuine model output: the UI is the real build; the conversation content is fixture data.
Corrections to the text
The README had never mentioned the local GGUF runtime. Now documented:
.ggufmodels served by a managed llama.cpp runtime, withllama-serverfetched once and verified against a pinned SHA-256auto/vulkan/cpu)Verification
python -m ruff check backend tests tools main.pypasses (CI lintstools/). All three image paths resolve from the README; 632 KB total.🤖 Generated with Claude Code