Trajectories tools #918
Merged
Trajectories tools #918
Conversation
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
Update checkmark positioning to use transform: translateY(-50%) for proper vertical centering instead of fixed top position.
Update knowledge folder naming convention from .refact_knowledge to .refact/knowledge and adjust allowed hidden folders configuration accordingly.
Add support for extracting and searching through chat trajectories: - New trajectory_memos module that periodically extracts insights from abandoned trajectories using LLM analysis - New search_trajectories tool to find relevant trajectory segments - New get_trajectory_context tool to retrieve full context from trajectories - TrajectoryFileSplitter for vectorizing trajectory JSON files - Integration with vecdb for trajectory-based semantic search
Filter out system role messages alongside context_file and cd_instruction messages when extracting and processing trajectory data across multiple modules. Also skip text validation for trajectory files to handle their unique structure appropriately.
- Expand PROMPT_EXPLORATION_TOOLS with knowledge management instructions - Add knowledge and trajectory search tools to subagent documentation - Enhance EXTRACTION_PROMPT with structured overview and title generation - Increase minimum message threshold from 4 to 10 for trajectory processing - Add TrajectoryMeta struct to capture overview and auto-generated titles - Refactor extract_memos to extract_memos_and_meta for dual extraction - Support dynamic title updates for auto-generated trajectory titles - Add title hint when current title is auto-generated
Import memories module and create enriched memory entries from subagent execution results with appropriate tags and metadata. This allows subagent tasks to be persisted and retrieved for future context.
Update create_knowledge tool description to mention "Use it if you need to remember something" for better user understanding.
Implement automatic context enrichment in AGENT chat mode by injecting relevant knowledge and trajectories before the user message. This includes: - New knowledge_enrichment module with signal-based heuristics to determine when enrichment should occur (first message, error keywords, file refs, etc.) - Enhanced memories_search to support separate top_n for knowledge vs trajectories - Score field added to MemoRecord for relevance filtering - Pre-stream messages passed through restream for UI display - Updated tool descriptions to mention trajectory search capability - Removed standalone search_trajectories tool (now integrated into knowledge) Refs #123
Enhance tool output messages by appending knowledge base save confirmation when enriched memories are successfully created. This provides users with visibility into where their tool results are being persisted. Changes: - Capture memory save path from memories_add_enriched result - Append formatted memory note to final tool output messages - Applied to deep_research, strategic_planning, and subagent tools Refs #123
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.