Merged
Conversation
Keep Web sessions and subagents on the active runtime backend so Axon Cloud and OpenAI-compatible flows do not fall back to Claude models or stale backend state. Also surface the new chat action in the primary input controls.
… loop - LeadAgent self-heal now restores old Session (cost, messages, permissions) via setSession() instead of discarding it on loop rebuild - Worker finally block now calls loop.abort() to prevent dangling autoMemorize promises (matching LeadAgent.stop() pattern) - Remove unused totalCostUSD dead code field from ConversationLoop (all cost tracking is in Session.state.totalCostUSD) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add a "layer zero" context optimization that runs before microcompact and autoCompact. Instead of irreversibly deleting old messages, it reduces old tool_result content to brief summaries while keeping the original data intact in the session. Key behaviors: - Only triggers when tokens exceed 70% of autoCompact threshold - Always preserves the 5 most recent tool results in full - Never reduces Edit/Write results (state changes) - Generates tool-specific summaries (Read → re-read hint, Bash → exit code + last line, Grep → match count, etc.) - Does not modify session messages - only affects the API request copy Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
SmartPlanner now validates AI-generated task decomposition before building the execution plan: - Detect and rename duplicate task IDs - Remove invalid/self dependency references - Detect and break circular dependencies (iterative until DAG) - Ensure all tasks are covered in parallelGroups - Split parallel groups when tasks have in-group dependencies - Split parallel groups when tasks modify the same files (prevents concurrent edit conflicts that waste Worker tokens) Also improves LeadAgent brief quality with a structured checklist (entry files, data models, naming conventions, error handling, API contracts, test requirements) replacing the single good/bad example. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Tool consolidation (40 → 27 tools): - Merge Eye+Ear → Sense, CreateTool+CreateAgent → Create, KnowledgeIngest+KnowledgeCompile → Knowledge - Absorb WebFetch → urlReader, MemorySearch → memoryReader, Database → sqlReader, NotebookWrite → notebookWriter - Remove: GoalManage, Network, LSP, REPL, MCP tools HTTP Native (third wall): - urlReader: custom headers, auth-aware caching, redirect auth stripping, error response bodies - urlWriter (new): POST/PUT/PATCH/DELETE to any URL via Write tool - Read/Write schemas extended with headers/method params - System prompt updated: model knows to use Read/Write for HTTP APIs directly IORuntime pipeline: Read: url(1) → memory(1) → sql(1) → image → pdf → office → svg → notebook → binary Write: url(1) → notebook(0) → image(3) → spreadsheet(4) → presentation(4) → document(5) → binary(10) Security: stripAuthOnRedirect shared util, metadata endpoint blocking, no auth response caching Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
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.
Summary
Changes
Related Issue
Testing
npx tsc --noEmitpassesnpm test -- --runpassesScreenshots