v1.0.0

Added (desktop app & first-run setup)

• Double-click desktop bundles so non-technical users can run Gene Lens
  without installing Python: a macOS Gene Lens.app, a single-file
  Windows GeneLens.exe, and a console (Terminal) build for watching
  logs. Built with PyInstaller (packaging/gene_lens.spec plus
  build_macos.sh / build_windows.bat) and published automatically on
  a version tag by .github/workflows/release.yml.
• app_launcher.py: the packaged entry point. Asks the language (a
  native macOS dialog in .app mode, a terminal prompt otherwise),
  starts the local server, and opens the browser.
• Browser first-run setup page (/setup): pick which reference databases
  to download (ClinVar required, PharmGKB optional), each with a plain
  explanation, a progress bar, and a clear "this is the only step that
  uses the internet, everything after is 100% offline" notice. The home
  route gates to it until the core database is present.
• Automatic PharmGKB download (the public clinical-annotations archive,
  no login), with graceful degradation if it is unavailable.
• Manual database update check: a "Check for updates" button in Settings
  and python main.py check-updates. Metadata-only, never auto-downloads,
  and records source freshness (Last-Modified / ETag) in a .db_meta.json
  sidecar so it can tell you when ClinVar or PharmGKB has a newer version.
• New brand logo (a DNA helix inside a lens) as static/img/logo.svg and
  logo-icon.svg, wired into every page header, a favicon on all pages,
  and macOS/Windows app icons (packaging/icon.icns, icon.ico).
• App version shown in the setup and settings footers, stamped into the
  macOS bundle Info.plist, and used in the release asset names.
• Frozen-bundle path handling (config.py): databases, history, logs,
  and I/O are written to a per-user writable directory
  (~/Library/Application Support/GeneLens, %APPDATA%\GeneLens) instead
  of the read-only app folder.
• Tests for the PharmGKB download and update check
  (tests/test_download_databases.py) and the first-run setup flow and
  gate (tests/test_setup_flow.py).

Added

• Per-analysis chat persistence (src/chat_store.py). Conversations
  survive closing the drawer and live under history/<hid>/chat.json
  with atomic writes and secure deletion.
• Streaming generator chat_about_analysis_stream() in src/local_ai.py
  so future UI can render tokens as Ollama produces them instead of
  waiting for the full reply.
• Token-budget estimation (estimate_tokens, estimate_prompt_tokens)
  for sizing prompts against the context window.
• Friendly error translation for common Ollama failures: missing model
  (with the exact ollama pull command), service not running, out of
  memory, plus a fallback that surfaces the raw stderr.
• Context-window guard in the chat prompt — caps the analysis payload
  at 20k characters and tells the model when content was omitted, so it
  won't fabricate details about the truncated portion.
• Suggested starter questions in the chat empty state (4 prompts, EN+PT).
• Animated three-dot typing indicator while the AI is responding.
• Markdown rendering CSS for assistant replies (lists, code, headers,
  inline code, bold/italic, rsID chip styling).
• 39 new pytest tests covering privacy, parsers, panels, system status,
  the chat function, the streaming generator, the persistence layer,
  and token estimation.

Changed

• Renamed product from "Genetic Health Analyzer" / "Analisador Genético
  Local" to Gene Lens across UI, CLI banner, report metadata, README
  titles, CI workflow name, and the i18n strings.
• Replaced 10 emoji glyphs in consent.html and dashboard.html with
  Tabler Icons served locally (font + CSS in static/), keeping the
  dashboard 100% offline.
• Full diacritic pass on the Portuguese README and i18n strings — every
  Configuracoes, Saude, NAO, Conclusoes now properly accented.
• "Why I Built This" README section rewritten to reflect the real
  motivation (family history of inherited conditions, the Nick Saraev
  post, findings that matched prior diagnoses).
• Install instructions updated for PEP 668 (mandatory python -m venv
  on modern macOS/Linux), both EN and PT.
• License switched from AGPL v3 to MIT — permissive use, copyright
  notice retention only, no service-disclosure obligation.
• CI workflow renamed Gene Lens CI, now runs the pytest suite on
  Python 3.10 / 3.11 / 3.12 instead of inline python -c smoke tests,
  with a separate i18n parity step.
• Reduced top spacing on the consent and upload screens so the warning
  and the file picker sit above the fold on shorter viewports.

Added (infrastructure)

• pyproject.toml with package metadata, classifiers, pinned core deps,
  and optional translation / local-ai / dev extras.
• CODE_OF_CONDUCT.md (Contributor Covenant 2.1).
• sample/README.md documenting the synthetic-only policy.
• .github/workflows/db-canary.yml — weekly Sunday check that the
  ClinVar and PharmGKB URLs still resolve and the ClinVar header
  schema hasn't shifted, with auto-issue on failure.
• Gitleaks secret scan in CI (with the fetch-depth: 0 fix so the
  PARENT..HEAD diff actually resolves).
• Sample-fixture marker check: every file in sample/ must declare
  SYNTHETIC in its first 5 lines.
• Dependabot grouped pip + github-actions, with ollama and
  argostranslate ignored so optional extras don't generate weekly PRs.

Security

• _csrf_check is now a before_request hook only; the buggy inline
  call inside /api/chat/ask that turned every valid POST into a 403
  was removed.
• /api/check-updates is POST-only so the global CSRF/Origin guard
  covers it. As a GET it could have been fired from any site open in the
  browser, triggering outbound pings to NCBI/PharmGKB and leaking the
  user's IP.
• First-run setup download errors return an opaque code to the browser;
  the detailed exception (which can include filesystem paths) stays in
  the local log only.
• The GENE_LENS_DATA_HOME override is resolved with .resolve() so a
  .. path cannot redirect where databases and downloads are written.