Spaces:
Running
Running
Commit History
21 days ago
-
Improve author voice evidence, factual warnings, and structural preservation Running
Codex9498b58
30 days ago
-
HumanTouch: authorial-preservation engine, change ledger, Voiceprint, evals (#4)
- HumanTouch: authorial-preservation engine, change ledger, Voiceprint, evals (e13af247dd2413959503eec6fc460e9770af9926) - Remove stale compiled files (1ec16eaa3f7435c246cdb75b6a767c676ebff676) - Deploy script: also clear root __pycache__ (07730280ef4b1496a9144673a7668b7912946273)db3e82c
-
Avoid Space SSR port conflict for domain challenge route
Verified signature1278e90 -
Serve OpenAI domain challenge before Gradio frontend
Verified signature6f90e9b -
Upload mcp_review.py
Verified signature9e8d20d -
Upload mcp_review.py
Verified signature75dee99 -
Upload app.py
Verified signature79b7cd3 -
Delete HumanTouch-server-update.zip
Verified signature061948e -
Upload 2 files
Verified signature06f20bd -
Use interoperable string schemas for MCP voice profiles
Verified signature9f631f5 -
Expose stateless HumanTouch MCP voice and revision tools
Verified signatureAdd analyze_voice, rewrite_in_voice, regenerate_with_notes and compare_versions; enable Gradio MCP and verify existing UI and new callbacks.b0ff352
-
Build preservation-first revision desk and individual Amplify takes
Verified signature60c4292
-
Make the rewrite flow the clear starting point
The first-load hierarchy now tells visitors to begin with Rewrite and makes the selected workflow easy to spot. All four tools remain available in the same tab rail. Constraint: Preserve the existing four workflows and keep each tab directly accessible. Confidence: high Scope-risk: narrow Directive: Keep Rewrite first and selected by default. Tested: 43 unit tests; Python compileall. Not-tested: Gradio interface harness requires Gradio, which is absent locally; fresh browser screenshot unavailable.fdf90f8 -
Add voice amplification workflow
Codexdc3d3a5
-
Keep the deployed studio consistent with its verified preview (#3)
- Keep the deployed studio consistent with its verified preview (bebc6b6a927b30471a80c7f7e2800f28e01803bf)39ed62d -
Keep factual scope and everyday language in natural rewrites (#2)
- Keep factual scope and everyday language in natural rewrites (7d18891ffb7864772d5ea52c90216776e5f7623a)5a1be2f -
Give HumanTouch a considered writing studio and dependable editing (#1)
- Give HumanTouch a considered writing studio and dependable editing (a2766c626ad5d49f686ece3af183cb46bd650ebd)420aaad
-
Update app.py
Verified signature4ca1aa8
-
Update app.py
Verified signature6a70ca9
-
Prevent HF iframe resizing from stretching HumanTouch
Hugging Face Spaces render the Gradio app inside an auto-resized iframe, so viewport-height CSS can feed back into a very tall page. The editor and canvas now use compact fixed responsive heights, and the Humanizer textboxes use fewer Gradio rows. Constraint: Hugging Face Spaces render the app inside an auto-resized iframe Rejected: Only shorten the compose input | did not address vh-driven editor and canvas height Confidence: high Scope-risk: narrow Directive: Do not reintroduce vh units into app CSS without verifying inside the live Space iframe Tested: py -m unittest discover -s tests -v; py -m py_compile app.py tests\test_humantouch.py; local browser measurements at 1280x900 and 375x812 Not-tested: Groq generation behavior, unchanged4f32ec2 -
Make the co-writing input feel appropriately compact
The previous canvas input height was raised to prevent clipped helper text, but it made the typing box feel too tall on mobile. Shortening the placeholder lets the input return to a compact 52px height without clipping or contrast regressions. Constraint: The placeholder must remain readable on narrow mobile widths. Rejected: Keep the 72px textarea | users reported the typing box felt too long. Confidence: high Scope-risk: narrow Directive: Keep canvas input height paired with placeholder length; do not lengthen the placeholder without rechecking mobile. Tested: py -m unittest discover -s tests -v; py -m py_compile app.py tests\\test_humantouch.py; local mobile browser measurement of #chat_input textarea at 52px. Not-tested: Live Groq generation, unchanged by this UI-only adjustment.9b47eb9 -
Restore readable writing controls in the live Space
Gradio 6 applies custom CSS through launch(), so the Space could render nested controls with default colors instead of the intended theme. The UI theme is now generated through a testable builder, passed through launch(), and expanded to cover nested textareas, inputs, tabs, chat surfaces, buttons, and mobile layouts. Constraint: Gradio 6 moved css from Blocks construction to launch(). Rejected: Keep only wrapper-level elem_id styling | nested Gradio controls still inherit unreadable defaults. Confidence: high Scope-risk: moderate Directive: Keep custom CSS launch wiring in sync with the installed Gradio major version before changing theme plumbing. Tested: py -m unittest discover -s tests -v; py -m py_compile app.py tests\\test_humantouch.py; desktop and mobile browser screenshots with computed color samples. Not-tested: Live Groq model response generation, only UI rendering with a dummy local GROQ_API_KEY.408a656
-
Make HumanTouch feel trustworthy enough to use
Verified signatureThe Canvas flow was failing because Gradio 6.15 expects chat history as role/content message dictionaries. The UI also read like a demo, so this tightens the editor prompt, labels, accessible seed actions, metadata, and responsive styling around the actual writing job. Constraint: Hugging Face Space uses Gradio 6.15 and requires message-dict Chatbot history Rejected: Keep tuple/list chat history | caused the deployed Canvas message-format error Rejected: Keep the alchemy-heavy prompt | produced florid rewrites and weaker trust Confidence: high Scope-risk: moderate Directive: Keep co-creation history in {role, content} message format for Gradio 6.x Tested: py -m unittest tests.test_humantouch Tested: py -m py_compile app.py tests\test_humantouch.py Tested: Local browser QA desktop and 390px mobile, no horizontal overflow Tested: Local Canvas submit with dummy key renders graceful assistant message instead of Gradio format error Not-tested: Production Groq secret live generation before …0b7bb00
-
Update app.py
Verified signature67cdd57 -
Update app.py
Verified signatureb4e6a87 -
Update README.md
Verified signature1326d49 -
Update README.md
Verified signature5faa43d -
Update README.md
Verified signature80fbd6c
-
Update app.py
Verified signatureed79b7f
-
Update README.md
Verified signature891f073
-
Update README.md
Verified signature859e650
-
Update README.md
Verified signature84278fa
-
Update README.md
Verified signature465fc9b -
Update app.py
Verified signature0631415 -
Update app.py
Verified signature0ed1add -
Update app.py
Verified signaturedacb0b4
-
Update app.py
Verified signatureea342d7 -
Update app.py
Verified signature4024bea -
Update app.py
Verified signaturedd08002 -
Update app.py
Verified signature7414256 -
Update app.py
Verified signaturef02ad06 -
Update app.py
Verified signaturee457283 -
Update app.py
Verified signatured601d5b -
Update app.py
Verified signaturef2f28b2 -
Update app.py
Verified signature6b704ca -
Update app.py
Verified signature69fdfa8 -
Update app.py
Verified signature285ce64 -
Update app.py
Verified signatureda2ccac -
Update app.py
Verified signature123bfc8
-
Update README.md
Verified signatured134a65