- _PinnedResolver raises on unpinned hosts instead of falling back to
live DNS: nothing legitimate resolves other hosts on a pinned session
(redirects are off), so fail closed
- dedicated session is also closed on EVENT_HOMEASSISTANT_CLOSE: core
stop does not unload entries, and the raw ClientSession has no HA
auto-cleanup, which left 'Unclosed client session' logs at shutdown
- hassfest.yaml: remove the version-print step, hassfest runs inside
the action's docker image and is never on the runner PATH
- create_pinned_session built directly on aiohttp: the HA helper injects
its own connector and raised TypeError on every setup and config flow
- session lifecycle: closed in APIClient.shutdown, on failed setup, and
in config flow validators; allow_redirects=False everywhere so 3xx
cannot route past the pinned resolver
- defaults: gemini-2.0-flash (shut down 2026-06-01) -> gemini-3.5-flash,
deepseek-chat (retires 2026-07-24) -> deepseek-v4-flash
- DeepSeek V4 disable_thinking via thinking request parameter
- Gemini 3.x disable_thinking via thinking_level (MINIMAL, LOW for Pro)
- get_history limit: no forced default/max in schema, storage layer clamps
- translations: 6 locales caught up with strings.json, 2 orphan keys dropped
- README: manual install path matches zip layout, model sections updated
Security:
- H1 DNS rebinding TOCTOU: validate_endpoint returns (endpoint, resolved_ips);
create_pinned_session builds an aiohttp session with a custom resolver that
returns only the pre-validated IPs, closing the re-resolve gap.
- H2/M3 Shared session cookie pollution: integration no longer uses HA shared
clientsession; isolated session with DummyCookieJar prevents cross-integration
cookie leaks.
- Cloud metadata / link-local block added to allow_local_network mode to
prevent IMDS exfiltration on cloud VMs.
- L3 hard cap extended to history.async_get_history (not only service schema).
ML/LLM correctness:
- _is_openai_reasoning_model uses regex with gpt-5-chat* blacklist and handles
OpenRouter-style openai/ prefix. Future o5/gpt-6 auto-recognized.
- reasoning_effort=minimal for gpt-5 family, low for o-series.
- Gemini 2.5 Pro gets thinking_budget=128 (Pro rejects 0, flash accepts 0).
- Anthropic extracts first type=text block instead of hardcoded content[0].
- /no_think dedup uses word-boundary regex instead of substring.
- DeepSeek-reasoner detected: skips /no_think; preserves reasoning_content.
Reliability:
- Exception chaining added to Gemini-block re-raises.
- Top-level imports for json/re/hashlib (no more lazy stdlib imports).
- allow_local_network logged at INFO, not WARNING on every setup.
Code style:
- PEP 604 type hints across all .py files (dict/list/| None).
No breaking changes for users. Internal API: validate_endpoint return type
changed from str to tuple[str, list[str]].
Add per-instance boolean option to allow private IP endpoints and HTTP
scheme for self-hosted LLM proxies (LiteLLM, Ollama, vLLM, etc.).
- New CONF_ALLOW_LOCAL_NETWORK config option (default: false)
- When enabled: allows RFC1918 private IPs and HTTP endpoints
- When disabled: full SSRF protection preserved (HTTPS + public IPs only)
- Multicast and unspecified addresses blocked regardless of setting
- Warning logged when local network mode is active
- Checkbox added to ConfigFlow and OptionsFlow UI
- Translations for all 8 languages
Closes#9
Replace Creative Commons Attribution-NonCommercial-ShareAlike 4.0
with PolyForm Noncommercial License 1.0.0 across all files:
LICENSE, README badge/footer, and all Python module headers.
Phase A — Critical:
- Remove dual timeout stacking in coordinator._send_to_api
- Add Gemini-specific asyncio.timeout (sync SDK via to_thread)
- Store full text in history for context; cap per-field at 32KB on disk
- Fix instance lookup to match by normalized_name
- Add Bearer/sk-/x-api-key credential sanitization patterns
Phase B — Dead code removal:
- Merge async_ask_question/async_process_question into single method
- Remove dead is_anthropic flag from coordinator and __init__
- Remove unused DEFAULT_TIMEOUT and API_TIMEOUT constants
- Remove redundant _create_history_dir calls
- Consolidate async_check_api to use provider registry
Phase C — Config flow correctness:
- Truncate name before uniqueness check (prevent post-truncation collisions)
- Add async_set_unique_id + _abort_if_unique_id_configured
- Extract shared _build_parameter_schema for ConfigFlow/OptionsFlow dedup
Phase D — UX improvements:
- Optimize history write: serialize from memory, single file write
- Show last 5 history entries in sensor attributes (was 1)
- Return actual error type in ask_question service response
- Add dedicated api_key_required error for provider/endpoint changes
- Pass config_entry to DataUpdateCoordinator (HA 2024.8+)
Phase E — Cleanup:
- Extract _apply_structured_output for OpenAI/DeepSeek dedup
- Reduce ABSOLUTE_MAX_HISTORY_SIZE to 200 with Final annotation
- Remove dead translation keys (queued, invalid_characters)
- Migrate to _attr_has_entity_name = True
- Add from __future__ import annotations to all modules
- Remove redundant api_status sensor attribute
- Add missing translation keys (last_model, last_timestamp, etc.)
Review agent fixes:
- Add archive file cleanup (max 3 archives) to prevent disk exhaustion
- Per-entry storage cap (32KB per field) for history on disk
Security:
- Make validate_endpoint async with hass.async_add_executor_job for DNS resolution
- Use _RestrictedIPError instead of fragile string matching for IP check flow
- Add is_multicast/is_unspecified to SSRF IP restriction checks
- Remove resolved private IP from error messages (generic message)
- Remove raw endpoint from __init__.py error log
- Sanitize error messages in metrics: strip URLs, API keys, Gemini key patterns
- Truncate API error response bodies before logging (512 chars)
- Use generic error messages in Gemini exception handlers (no str(e) interpolation)
- Pass api_key as explicit APIClient constructor parameter (not from header)
- Add defensive validation: Gemini provider requires api_key at construction
Code quality:
- Add constants: DEFAULT_INSTANCE_NAME, MIN/MAX_CONTEXT_MESSAGES, MIN/MAX_HISTORY_SIZE
- Replace all hardcoded schema ranges with named constants
- Move datetime import to module level in history.py
- Remove unused full_history/full_history_available keys
- Chain socket.gaierror properly with raise...from
- Extract HistoryManager (history.py) and MetricsManager (metrics.py) from coordinator
- Fix Gemini chat: use client.chats.create(history=...) instead of sequential send_message
- Fix float("inf") in metrics breaking json.dump persistence
- Fix _get_current_state checking wrong error key ("error" vs "error_message")
- Fix API key re-entry requirement when endpoint/provider changes in options flow
- Make CONF_API_KEY optional in options schema (stored key used as fallback)
- Add DNS rebinding protection via socket.getaddrinfo in validate_endpoint
- Remove mass assignment vulnerability in config_flow._create_entry
- Add description_placeholders to all error re-show paths
- Fix history migration overwriting existing JSON data
- Return list copy from get_limited_history to prevent reference leaks
- Add history_info to _get_safe_initial_state for data shape consistency
- Add defensive None check in _get_sanitized_last_response
- Move dt_util import to module level in config_flow
- Remove dead fallback branches in coordinator.last_response property
- Fix sensor min_latency display after float("inf") removal
- Fix history directory permissions from 0o777 to 0o755
- Deduplicate schema definitions via _build_provider_schema()
- Use centralized build_auth_headers from providers.py
Security:
- Add SSRF protection via HTTPS-only endpoint validation with private IP blocking
- Mask API keys in all log output via safe_log_data()
- Sanitize error responses to prevent internal detail leakage
- Remove API key pre-population in config flow forms
- Harden service schemas with input length limits and type coercion
Bug fixes:
- Fix temperature=0 rejected due to Python falsy value bug (0 or default)
- Fix fire-and-forget race condition in coordinator init (5 async_create_task → awaited async_initialize)
- Fix rate limit state not resetting after successful API calls
- Fix ConnectionError incorrectly setting is_rate_limited=True
- Fix _rotate_history calling async method via sync executor
- Fix shutil.move blocking event loop in history migration
- Fix async_shutdown removing wrong key from hass.data
- Replace os.rename with shutil.move for cross-device safety
Improvements:
- Add asyncio.Lock for request serialization
- Replace async_timeout with stdlib asyncio.timeout (Python 3.12+)
- Use SupportsResponse.OPTIONAL for ask_question and get_history services
- Use Platform.SENSOR enum instead of string literal
- Add strings.json for HA translation framework
- Update DEFAULT_ANTHROPIC_MODEL to claude-sonnet-4-6
- Retry only transient errors (429, timeout) in API client
Cleanup:
- Remove dead code: unused schemas, imports, sync_write_history, check_memory, check_connection
- Remove icon copying code from async_setup
- Remove unused dependencies from manifest (anthropic, openai, certifi, async-timeout)
- Remove empty manifest arrays (bluetooth, mqtt, ssdp, usb, zeroconf)
- Fix duplicate JSON keys in 5 translation files
- Fix Serbian translation: "Прекључено" → "Искључено" for disconnected state
Bump version to 2.4.0
SMKRV