RAG UI refactor

[Felipedino]

This pull request introduces several improvements to the backend, focusing on localization, configuration resolution, and API enhancements for RAG (Retrieval-Augmented Generation) workflows. The most significant changes include the introduction of a unified localization utility, new endpoints for RAG session configuration and status, and a utility for resolving component schema defaults. Additionally, API responses are now consistently localized, and error messages for unsupported document types have been improved.

Localization improvements:

• Added a new localize function in core/utils.py that recursively replaces every MultilingualString with its localized string based on the Accept-Language header, ensuring all API responses are language-specific and never expose raw language objects. Existing ad-hoc localization logic was replaced with this utility throughout the API endpoints. [1] [2] [3] [4] [5] [6] [7] [8] [9]

RAG API and configuration enhancements:

• Expanded the rag.py endpoints to include chunking presets, session defaults, resolved session configuration, and session index status. All responses are now pre-localized and fully resolved, allowing the frontend to render data directly without additional processing.

• Added filtering by task_name to the generative session listing endpoint, allowing clients to query sessions for a specific task or all tasks. [1] [2]

Component schema defaults resolution:

• Introduced resolve_component_defaults in core/schema_fields/defaults.py, a utility to resolve component schema placeholders into complete parameter dictionaries, including nested components. This ensures the backend can generate ready-to-use configurations for components. [1] [2]

Error handling improvements:

• Improved error messages for unsupported document file types to include a list of supported extensions, enhancing clarity for API consumers.

RAG session creation logic:

• Adjusted the order of RAG session parameter validation to ensure that default values are applied before checking for required component downloads, preventing false negatives in dependency checks. [1] [2]

These changes collectively improve the robustness, internationalization, and usability of the backend APIs, especially for RAG workflows.

[Copilot]

🟡 Changes recommended

There are a few correctness/UX issues in the new/updated flows (notably ambiguous prompt_id handling, undercounted prompt budget math, and an avoidable redirecting sessions API call) that should be addressed before merge.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

This PR refactors RAG (Retrieval-Augmented Generation) backend + frontend integration so the backend returns fully resolved, localized configuration/preset data and the frontend treats RAG as a standalone “entry point” with its own scoped session list and routes.

Changes:

• Added unified backend localization (localize) and expanded RAG backend APIs (presets, session defaults, resolved configuration, index status).
• Updated RAG session validation to apply backend defaults (and generation-model schema defaults) prior to download/dependency checks, and improved unsupported document-type errors.
• Frontend refactor to make RAG a standalone entry point (new pages/routes/providers), consume resolved configuration from the backend, and improve upload/session UX.

File summaries

File	Description
tests/back/tasks/test_tasks.py	Updates task metadata expectations to include entry_point.
tests/back/RAG/test_RAG_strict_validation.py	Adjusts strict validation to allow generation-model param defaulting and asserts defaults are applied.
tests/back/RAG/test_RAG_session_validation.py	Updates session validation expectations for defaulted vs required parameter keys.
tests/back/RAG/test_RAG_session_configuration_api.py	Adds tests for resolved configuration + index-status endpoints.
tests/back/RAG/test_document_extractor_api.py	Adds coverage for improved unsupported-upload error details and no-extension rejection.
tests/back/RAG/test_chunking_presets.py	Adds tests for chunking presets, localization, and session-default behavior.
DashAI/front/src/utils/apiError.js	Adds helper to extract user-facing error messages from API failures.
DashAI/front/src/types/retrieverPreset.ts	Extends retriever preset typing with backend-provided display_name.
DashAI/front/src/types/ragConfiguration.ts	Adds types for resolved RAG configuration, presets, defaults, and index status.
DashAI/front/src/pages/generative/SessionRouter.jsx	Redirects RAG sessions to canonical standalone route and localizes “not found”.
DashAI/front/src/pages/generative/RAGSession/sections/PromptSection.jsx	Removes legacy prompt section implementation (backend-resolved config flow).
DashAI/front/src/pages/generative/RAGSession/sections/GeneratorSection.jsx	Removes legacy generator section implementation (backend-resolved config flow).
DashAI/front/src/pages/generative/RAGSession/sections/ChunkingSection.jsx	Removes legacy chunking section implementation (preset-driven backend flow).
DashAI/front/src/pages/generative/RAGSession/components/SectionCard.jsx	Removes legacy section wrapper component.
DashAI/front/src/pages/generative/RAGSession/components/RAGCard.jsx	Removes legacy accordion card component.
DashAI/front/src/pages/generative/RAGSession/components/PresetCard.jsx	Removes legacy preset-card component.
DashAI/front/src/pages/generative/RAGSession/components/GeneratorBody.jsx	Removes legacy generator UI body component.
DashAI/front/src/pages/generative/RAGSession/components/AdvancedConfigCard.jsx	Removes legacy “advanced config applied” indicator card.
DashAI/front/src/pages/generative/RAG/RAGHomePage.jsx	Adds standalone RAG home entry point page.
DashAI/front/src/pages/generative/RAG/RAGCreatePage.jsx	Adds standalone minimal RAG session creation page (name + docs + model).
DashAI/front/src/hooks/generative/useTaskDisplayName.js	Adds hook to show backend-localized task names in the UI.
DashAI/front/src/hooks/generative/useSessions.js	Adds backend-driven session filtering (e.g., by task_name).
DashAI/front/src/components/threeSectionLayout/GroupedCollapsibleList.jsx	Flattens rendering for single-group lists and refactors item rendering helper.
DashAI/front/src/components/generative/standaloneEntryPoints.js	Adds frontend map for standalone tasks (routes + icons).
DashAI/front/src/components/generative/SessionBar.jsx	Routes session clicks to standalone task routes and supports custom title.
DashAI/front/src/components/generative/RAG/RAGDocumentsPanel.jsx	Passes task name constant and index status into documents UI.
DashAI/front/src/components/generative/RAG/RAGBreadcrumbs.jsx	Normalizes RAG routes (lowercase) and updates task selection behavior.
DashAI/front/src/components/generative/RAG/PromptParamsCard.jsx	Updates prompts navigation route casing to canonical /rag/....
DashAI/front/src/components/generative/RAG/GeneratorPicker.jsx	Adds unified generator model picker with download/credential gating.
DashAI/front/src/components/generative/RAG/DocumentTable.jsx	Improves upload error reporting via snackbar and extracted API messages.
DashAI/front/src/components/generative/RAG/DocumentSelector.jsx	Improves upload error reporting via snackbar and extracted API messages.
DashAI/front/src/components/generative/RAG/DocumentsBar.jsx	Canonicalizes routes, adds per-document index badge support, improves upload error messaging.
DashAI/front/src/components/generative/RAG/DocumentListItem.jsx	Adds indexed/not-indexed chip with chunk counts.
DashAI/front/src/components/generative/RAG/DocumentList.jsx	Threads index-state map through to list items.
DashAI/front/src/components/generative/GenerativeContext.jsx	Adds provider-level session filtering and fixes fetch effect dependencies.
DashAI/front/src/components/generative/GenerativeChat.jsx	Adds indexing-in-progress hint when RAG is not yet indexed.
DashAI/front/src/components/generative/CreateSessionRight.jsx	Removes RAG-specific info bar from generic session creation flow.
DashAI/front/src/components/generative/CreateSessionLanding.jsx	Adds standalone task tiles alongside “create session” on landing.
DashAI/front/src/components/generative/CreateSessionContext.jsx	Excludes standalone tasks from the shared model gallery.
DashAI/front/src/components/generative/CreateSessionCenter.jsx	Removes embedded RAG setup from generic creation flow.
DashAI/front/src/App.jsx	Adds canonical standalone RAG routes and scopes them via a filtered provider.
DashAI/front/src/api/session.ts	Adds optional server-side filtering to sessions API call.
DashAI/front/src/api/rag.ts	Adds RAG constants and new APIs for presets/defaults/config/index status.
DashAI/front/src/api/job.ts	Removes dedicated RAG job enqueue call (likely superseded by generic job flow).
DashAI/back/tasks/RAG_task.py	Marks RAG task as ENTRY_POINT = "standalone".
DashAI/back/tasks/base_generative_task.py	Introduces ENTRY_POINT and returns it via task metadata.
DashAI/back/services/RAG/session_validation_service.py	Applies backend defaults, resolves generation-model defaults, and reorders download checks.
DashAI/back/services/RAG/session_defaults_service.py	New service to construct backend-resolved default RAG parameters.
DashAI/back/services/RAG/session_configuration_service.py	New service to return UI-ready resolved configuration + context budget.
DashAI/back/services/RAG/retriever_presets.py	Refactors preset resolution, adds localized names, adds matching + top_k helpers.
DashAI/back/services/RAG/index_status_service.py	New service to report indexed/stale/not-indexed state without mutating.
DashAI/back/services/RAG/chunking_service.py	Exposes chunk-set signature builder publicly for read-only callers.
DashAI/back/services/RAG/chunking_presets.py	New localized backend-owned chunking preset recipes.
DashAI/back/models/RAG/RAG_pipeline.py	Changes default retriever placeholder to BM25.
DashAI/back/models/RAG/documents/file_type.py	Adds supported_extensions() helper for clearer error messages.
DashAI/back/models/RAG/chunking_models/token_chunk_model.py	Adds localized aliases/descriptions plus CHUNK_UNIT="tokens" and model metadata.
DashAI/back/models/RAG/chunking_models/character_chunk_model.py	Updates placeholders + adds localized aliases/descriptions and model metadata.
DashAI/back/models/RAG/chunking_models/base_chunking_model.py	Adds CHUNK_UNIT to identify chunk-size unit.
DashAI/back/core/utils.py	Adds localize() utility to collapse MultilingualString in responses.
DashAI/back/core/schema_fields/defaults.py	Adds resolve_component_defaults() for schema placeholder resolution.
DashAI/back/core/schema_fields/init.py	Exports resolve_component_defaults from schema_fields package.
DashAI/back/api/api_v1/endpoints/rag.py	Expands RAG endpoints and ensures responses are localized + resolved.
DashAI/back/api/api_v1/endpoints/generative_session.py	Adds task_name filtering, localizes task display_name, and reorders RAG validation vs download checks.
DashAI/back/api/api_v1/endpoints/documents.py	Improves unsupported-type error detail with supported extension list.
DashAI/back/api/api_v1/endpoints/components.py	Replaces ad-hoc language filtering with localize().

Review details

• Files reviewed: 78/78 changed files
• Comments generated: 4
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.