Conversation
- Replace NavigationStep with ReasoningStep that includes detailed decision reasoning, stage information, and candidate alternatives - Add reasoning_chain field to PipelineContext and RetrieveResponse - Implement reasoning recording for each retrieval stage (Analyze, Plan, Search, Evaluate) - Include decision rationale, strategy used, and LLM call summaries in reasoning steps - Enable auditable and explainable retrieval decisions instead of opaque navigation traces docs: add function specification and scenario documentation - Create design document outlining core differentiation features including structured reasoning chain, query planner, multi-hop reasoning - Add comprehensive scenario documentation covering various use cases like AI coding assistant, enterprise knowledge base, legal contract review, academic research - Define API design principles with two-method interface approach - Document implementation priorities from P0 to P6 with clear dependencies refactor(retrieval): enhance pipeline context with reasoning capabilities - Update PipelineContext to use ReasoningChain instead of navigation_trace - Add push_reasoning_step and record_reasoning helper methods - Modify stage implementations to populate reasoning information - Update type exports to include new reasoning-related types
…nagement - Introduce RetrievalBudgetController to track token consumption across all pipeline stages (Plan, Search, Evaluate) - Add BudgetStatus enum with Healthy, Constrained, and Exhausted states - Implement budget-aware strategy selection in Plan stage - Apply dynamic beam width adjustment based on budget status in Search stage - Enable graceful degradation when budget is exhausted - Track and report budget utilization throughout the pipeline - Add comprehensive tests for budget controller functionality
Add optional path_constraint field to SubQuery to store structural path constraints extracted from queries (e.g. "3.2", "Chapter 5"). Add resolved_path_hints field to PipelineContext to store node IDs extracted from structural queries. Implement Chinese number parsing utility function to convert Chinese numerals to integers (e.g. "三" → 3, "二十一" → 21). Add extract_structure_hints method to recognize and map various structural patterns including: - Chinese chapter/section patterns: "第X章", "第X节", "第一章" - English section numbers: "Section 3.2", "section 4.1.2" - Chapter references: "Chapter 5", "chapter 10" - Bare section numbers: "3.2.1", "2.1" - Table/Figure references: "表3", "Table 5", "图2", "Figure 4" Integrate structure hints into search stage by injecting them as high-priority search cues with confidence score of 1.0. Update all relevant initialization code to include the new fields in tests and implementation.
- introduce ReasoningIndex and related types (TopicEntry, SummaryShortcut, SectionSummary, HotNodeEntry, ReasoningIndexConfig) for pre-computed topic-to-path mappings - create ReasoningIndexStage that builds the reasoning index after enrich stage and before optimize stage - add reasoning_index field to PipelineOptions and IndexContext to configure and store the reasoning index - implement hot node tracking functionality in HotNodeTracker for identifying frequently retrieved nodes - integrate reasoning index into persist stage to save it with documents - update metrics to track reasoning index build time and statistics - add reasoning cache integration in retrieval pipeline context
- Introduce DocumentGraph struct with nodes, edges, and keyword indexing - Implement graph builder for computing relationships based on shared keywords - Add graph-aware retrieval strategy with boost capabilities - Support serialization and storage integration in workspace - Include comprehensive test coverage for graph operations
…vent support This commit introduces comprehensive streaming retrieval capabilities to the vectorless engine. Users can now receive incremental results through a channel-based event system that provides real-time feedback during the retrieval process. Key changes include: - Added `retrieve_streaming()` method to `PipelineRetriever` that returns a channel for receiving `RetrieveEvent`s as the pipeline progresses - Implemented `RetrieveEvent` enum with variants for different pipeline milestones including Started, StageCompleted, NodeVisited, ContentFound, Backtracking, SufficiencyCheck, Completed, and Error - Created new streaming module with proper channel handling and event emission throughout the retrieval pipeline - Added streaming example in `examples/rust/streaming.rs` demonstrating how to process events as they arrive - Updated `RetrievalOrchestrator` to support both regular and streaming execution paths without affecting existing functionality - Enhanced documentation with usage examples and comprehensive event descriptions The streaming API maintains backward compatibility - existing `query()` method remains unchanged while `query_stream()` provides the new incremental functionality. This enables use cases like interactive Q&A with real-time feedback, progress monitoring, and responsive UI updates.
Removed the main README.md and design documents for client-module and content-aggregation as they are no longer relevant to the current codebase structure. BREAKING CHANGE: Documentation files removed as part of cleanup process.
…ontent - Replace old logo with new lovable-vectorless.png and update styling - Change title to "Reasoning-native Document Intelligence Engine" - Update crates.io badge to show version instead of downloads - Remove detailed architecture diagram and how-it-works sections - Simplify quick start with streamlined Python and Rust examples - Consolidate API usage patterns into concise code snippets - Remove configuration details and focus on essential setup - Streamline feature descriptions and remove redundant sections - Remove architecture diagram file as it's no longer referenced
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.