feat(parser): add built-in parser engine with deployment-level fallback and allow list

[nixidexiangjiao]

Description

Add a declarative deployment-level fallback mechanism for parser engine configuration via config/builtin_parser_engine.yaml. When a tenant's parser engine config is absent, the system automatically falls back to the built-in config.

Additionally, introduce PARSER_ENGINE_ALLOW_LIST — a deployment-level env var that lets operators restrict which parser engines are visible and selectable, preventing users from configuring disallowed engines.

Key changes:

• YAML loader (internal/types/builtin_parser_engine_config.go): Loads config/builtin_parser_engine.yaml at startup with ${ENV} and ${ENV:-default} placeholder interpolation, stored in an atomic.Pointer singleton. Supports default_engine, docreader_addr, mineru, mineru_cloud, and weknoracloud blocks.
• Resolver layer (internal/types/parser_engine_resolve.go): ResolveXxxOverrides functions implement tenant → builtin fallback with package-level granularity (if tenant has MinerU endpoint set, the entire tenant MinerU block wins; otherwise the entire builtin MinerU block is used — no field-level mixing). MergeParserEngineOverrides is the one-stop API for handler/service code.
• Allow list (internal/handler/parser_engine_allowlist.go): PARSER_ENGINE_ALLOW_LIST env var controls which engines are permitted. Empty env → all engines allowed (backwards compatible). Disallowed engines are marked Allowed=false in API responses and cannot be configured or tested via the UI.
• Handler refactor (internal/handler/system.go): Replace scattered tenant config extraction with types.MergeParserEngineOverrides(tenant). Apply allow list to all ListParserEngines / CheckParserEngines / ReconnectDocReader responses. Add allowed_providers and default_engine fields to responses.
• Service refactor (internal/application/service/knowledge.go, internal/handler/session/attachment_processor.go): Use types.MergeParserEngineOverrides instead of inline tenant config extraction.
• DocReader address resolution (internal/types/parser_engine_resolve.go): ResolveDocReaderAddr adds a builtin.DocReaderAddr fallback when DOCREADER_ADDR env var is unset.
• Frontend — allow list UI: ParserEngineSettings.vue and KBParserSettings.vue render Allowed=false engines with a distinct "Not allowed" badge and reason hint, disable save/test buttons for disallowed engines, and prevent the "go to settings" link from appearing for allow-list-blocked engines. default_engine from builtin YAML pre-selects the default engine per file-type group when available and allowed.
• i18n: Added notAllowed / notAllowedHint strings for zh-CN, en-US, ko-KR, ru-RU.
• API type (internal/types/docparser.go): Added Allowed bool field to ParserEngineInfo.

Type of Change

• 🐛 Bug fix
• ✨ New feature
• 💥 Breaking change
• 📚 Documentation update
• 🎨 Refactor
• ⚡ Performance improvement
• 🧪 Test
• 🔧 Configuration / Build / CI

Related Issue

Fixes #

Testing

All new test files pass with go test ./internal/types/ ./internal/handler/ ./internal/application/service/knowledge/ ./internal/handler/session/:

• internal/types/builtin_parser_engine_config_test.go — File missing, invalid YAML, env interpolation (string + bool defaults), empty config, env override path, atomic replace
• internal/types/parser_engine_resolve_test.go — Full matrix for ResolveMinerUOverrides / ResolveMinerUCloudOverrides (tenant-wins, builtin-fallback, both-empty, builtin-endpoint-empty-skip), ResolveWeKnoraCloudAppID (creds-win, builtin-fallback, neither-set), ResolveDocReaderAddr (env-wins, builtin-fallback, empty), MergeParserEngineOverrides integration
• internal/handler/parser_engine_allowlist_test.go — Empty env → all allowed, comma list, mixed separators + case insensitivity, unknown names dropped, isParserEngineAllowed, firstAllowedParserEngine, allowedParserEnginesSorted, resolveDefaultParserEngine (no builtin, set+allowed, blocked by allow list, unknown name)
• internal/handler/system_parser_engines_test.go — ListParserEngines with builtin MinerU endpoint enabling engine visibility, allow list blocking engine, default all-allowed behavior

Checklist

• make fmt && make lint && make test pass locally
• Self-reviewed the code
• Added/updated tests covering the change
• Updated related documentation (README, docs/, Swagger annotations, etc.)
• Breaking changes are clearly called out in the description above

Screenshots / Recordings

UI changes: "Not allowed" badge on disallowed parser engines in both KB parser settings and system parser settings; disabled save/test buttons; "default engine" pre-selection respects builtin YAML + allow list.