- Pass UI_USERNAME/UI_PASSWORD from .env to LiteLLM container as
  LITELLM_UI_USERNAME/LITELLM_UI_PASSWORD (LiteLLM's expected env vars)
- Derive NEXTAUTH_URL from PUBLIC_BASE_URL instead of hardcoding
  localhost — fixes Langfuse post-login redirect to broken URL

- LiteLLM: /llm-routing/litellm/ui/ (SERVER_ROOT_PATH prefix required)
- Langfuse: / (web UI root)
- Canonical URLs: /litellm/ui/ and /langfuse

…cation

- Router: /visualizer endpoint
- Infrastructure: MinIO /minio/health/live, ClickHouse /ping
- LiteLLM: direct chat completion (bypasses triage router)
- Canonical URLs: /visualizer

- POST agent-simple-core through public URL (dev.vendeuvre.lan)
- Validates full HTTPS path: HAProxy → dev router → LiteLLM → model
- Graceful DNS skip when host unreachable

- All three chat completion parsers now use .get() on choices/message
  instead of direct indexing (Gemini Code Assist review)
- Canonical URL skipped checks are tracked separately and shown
  in summary as 'X skipped' (Sourcery review)

… ports, .env logging

- Extract parse_chat_response() helper to deduplicate chat completion parsing
  across test_e2e_chat, test_litellm_direct_chat, and test_canonical_urls
- Add isinstance guards for defensive JSON response parsing (Gemini suggestion)
- Load MinIO and ClickHouse ports from .env (MINIO_S3_PORT, CLICKHOUSE_HTTP_PORT)
  instead of hardcoding 9002/8123 (Sourcery suggestion)
- Log which .env files were loaded and warn on missing files (Sourcery suggestion)
- Handle export prefix and inline comments in .env parser (Gemini suggestion)

Apply isinstance guards on choices/message across benchmark_classifier.py,
retry_errors.py, and reclassify_all.py — same pattern as the verification
script's parse_chat_response() helper. Prevents unhandled exceptions when
the API returns unexpected response structures.

- Drop fragile inline-comment stripping from env parser (Sourcery + Gemini):
  heuristic truncated '#' in passwords/URLs — only full-line comments now
- Add encoding='utf-8' to open() call (Gemini)
- Add isinstance(data, dict) guards before .get() calls in all 4 classifier
  scripts: benchmark_classifier, reclassify_all, retry_errors, classify_direct
- Use 'or' fallback for UI_USERNAME/UI_PASSWORD (Gemini):
  os.environ.get('X') or 'admin' handles empty-string exports
- Send auth headers on canonical GET endpoints (Sourcery):
  defense-in-depth even though router public endpoints don't require auth
- Fix exit condition: skipped DNS-unreachable endpoints no longer cause
  non-zero exit code (Sourcery high-level)

- Security: UI_PASSWORD defaults to LITELLM_MASTER_KEY (not 'admin')
  when UI_PASSWORD env var is unset (Gemini Code Assist)
- Robustness: strip trailing slash from PUBLIC_BASE_URL to prevent
  double-slash URLs (Gemini Code Assist)
- Deduplication: extract parse_chat_response() into shared
  scripts/chat_helpers.py, imported by all 4 classifier scripts +
  verification script (Sourcery)
- UX: skipped canonical URL tests show '⚠ SKIP' instead of ✓
  (Sourcery)
- Cleanup: remove unused 'base_url' from load_env() return dict
  (Sourcery)

- LiteLLM liveness: /ping (404) → /health/liveness (200)
- Langfuse liveness+readiness: /api/health (404) → /api/public/health (200)

Both containers were crash-looping because Podman's
HealthcheckOnFailureAction=restart kept killing them when the health
checks hit the wrong endpoints (404). RestartCount was 22+ on both
dev and prod before the fix; now 0 and healthy.

Pulls the latest GitHub release tag, clones it, rsyncs runtime files
(pod.yaml, start-stack.sh, litellm/, router/, scripts/) into ~/prod/
without touching .env or data/, then redeploys via start-stack.sh --pull.

Prevents temptation to patch prod directly — all config changes flow
through git releases.

- LiteLLM liveness: /ping → /health/liveness
- Langfuse liveness+readiness: /api/health → /api/public/health
- Add upgrade-prod.sh to directory layout

## Gemini Code Assist fixes
- load_env() now checks os.environ before .env values for config resolution
- test_canonical_urls() prints skip message when PUBLIC_BASE_URL not set
- parse_chat_response() uses isinstance guards before .strip() to prevent
  AttributeError on non-string content/reasoning_content values

## Sourcery fixes
- Created scripts/__init__.py and scripts/verification/__init__.py to
  turn scripts/ into a proper Python package
- All scripts now import via
  with a single WORKDIR sys.path.insert instead of per-script hacks
- .env value parser now strips trailing whitespace after quote removal

## Global consistency
- verification_helpers.py: replaced unsafe inline content extraction
  with shared parse_chat_response()
- verify_ollama_routing.py: same — uses parse_chat_response() instead
  of direct choices[0]["message"].get("content") indexing
- scripts/README.md: added chat_helpers.py module documentation

verify_ollama_routing.py, verify_ollama_cooldown.py, and
verify_direct_ollama_cooldown.py now try relative imports first
(from .verification_helpers) and fall back to absolute imports.
This allows them to work both as package imports and when run
directly from the command line.

…d, f-string typo

- upgrade-prod.sh: drop trailing slashes on rsync directory sources to prevent
  --delete from wiping .env/data/backups in PROD_DIR root (Gemini + CodeRabbit)
- upgrade-prod.sh: add [ -t 0 ] guard around interactive read prompt so the
  script doesn't fail with set -e in CI/cron/non-TTY environments (Gemini)
- verify_canonical_endpoints.py: fix \\n literal to actual newline escape
  in summary separator f-string (CodeRabbit)

…empty values

- verify_canonical_endpoints.py: change env.get(key, default) to env.get(key) or default
  so that empty .env values (e.g. ROUTER_PORT=) fall through to defaults instead of
  producing empty-string ports. (Gemini Code Assist review)

- verify_canonical_endpoints.py: check r.status_code == 200 before calling r.json()
  in three locations (router /v1/models, router /api/dashboard-stats, litellm /v1/models).
  Use conditional .json() + isinstance guard to prevent JSONDecodeError on non-200
  responses and improve error reporting with HTTP status codes. (Gemini Code Assist review)

- upgrade-prod.sh: remove risky trailing backslash after command
  substitution; split single rsync into individual calls for clarity
  (--delete only on directories, not on top-level files)
- verify_canonical_endpoints.py: add isinstance(data, list) guards
  for models.get('data') in both router and LiteLLM /v1/models
  parsing; fix type mismatch in Langfuse health check detail param
  (use r.text[:80] instead of r.json() dict)

- chat_helpers.py: add _normalize_chat_content for structured payloads
  (list-based content, dict content, nested shapes from alt providers)
- upgrade-prod.sh: pre-flight .env check before pod stop (CodeRabbit);
  exclude upgrade-prod.sh from rsync to prevent self-overwrite (Gemini);
  add trailing newline (Gemini)
- verify_canonical_endpoints.py: catch httpx.RequestError instead of
  ConnectError for DNS/timeout robustness (Gemini, 2 sites);
  remove extraneous f-string prefix (CodeRabbit / Ruff F541)
- verify_ollama_routing.py: reuse WORKDIR Path for workspace_dir
  (CodeRabbit); drop dead import os; simplify exception handler
  (dead KeyError/JSONDecodeError)
- verification_helpers.py: simplify exception handler (dead
  KeyError/IndexError)

- upgrade-prod.sh: self-copy to /tmp pattern so rsync can safely update
  the script itself; centralized cleanup trap; removed --exclude guard
- chat_helpers.py: strip all _normalize_chat_content return paths
  (list-join and dict-text branches were missing .strip())