- Added a task.maxRuntimeMs setting with a default disabled state for per-subagent runtime limits.
- Updated subprocess execution to enforce the configured wall-clock timeout, mark timeouts as aborts, and include timeout-specific abort messaging.
- Tracked per-turn context size and context window through task progress/results and updated UI renderers to display current context against window with cumulative tokens rendered separately.
Eliminates the three-way duplication of toolCount/tokens/cost stat
appending across renderAgentProgress (running), renderAgentProgress
(completed), and renderAgentResult.
renderAgentResult (rendered after the task tool resolves) was missing the
cost display present in renderAgentProgress. Read cost directly from
result.usage?.cost.total which is already available on SingleResult.
Token counter (token_total status-line segment, subagent progress tree,
session-observer stats line) previously included cacheRead in its cumulative
sum. With Anthropic prompt caching, cacheRead per turn equals the full cached
context, so summing across N turns gives N*context_size -- a session with a 1M
context and 5 turns showed ~5M tokens despite no compaction occurring.
Fix: display shows input + output + cacheWrite per turn. cacheWrite is kept
because each byte is written once; cacheRead re-reads the same context every
turn. Dedicated cache_read/cache_write status-line segments still show cache
activity; billing cost is unaffected.
Also adds per-subagent cost display (dollar amount, statusLineCost color)
accumulated incrementally from message_end events. Hidden when cost is zero
(subscription/OAuth providers). Brings token and cost display in line with
what Claude Code shows per-agent.
- Flattened newlines and tabs in summary explanations before generating the review preview text, and trimmed the extracted sentence before truncating.
- Sanitized finding titles during render by normalizing tabs/newlines to spaces after stripping priority prefixes.
- Updated task call rendering to use `args.tasks?.length ?? 0` when displaying agent counts.
- Prevented runtime warnings in streaming task calls when the `tasks` array was still undefined.
- Renamed subagent completion flow from `submit_result` to `yield` across SDK tools, prompts, and docs.
- Updated executor/task handling to require and parse `yield` calls, replacing legacy submit-result extraction and state flags.
- Added `subagent-yield-reminder` and updated system prompts to require `yield` with `result.data` or `result.error`.
- Renamed hidden-tool and registration plumbing to `yield`, including discovery helpers and renderer/test surface.
- Added optional `assignment` field to task result and progress interfaces to track raw per-task assignment text separately from full templated task.
- Updated task rendering to display assignment text instead of full task template when available, improving clarity of task display.
- Modified task section rendering to show trimmed assignment text with fallback to task field if assignment is not available.
- Propagated assignment field through task executor, template renderer, and result objects to maintain consistency across task processing pipeline.
- Add a dedicated abortReason field to SingleResult and thread it through task execution paths so aborted subagents show actionable context instead of a generic badge.
- Populate abortReason for signal cancellation, pre-start cancellation, submit_result aborted status, and missing submit_result after reminders
Fixes#248
- Added topP, topK, minP, presencePenalty, and repetitionPenalty sampling control options to StreamOptions and AgentOptions interfaces.
- Implemented getter and setter properties on Agent class for runtime configuration of model sampling parameters.
- Added UI configuration and preset value providers for five new sampling parameters in coding-agent settings schema and components.
- Integrated sampling control parameters through proxy layer and agent session configuration for end-to-end provider support.
Validate report_finding details before extraction and normalize findings before task rendering to avoid crashes when tool errors emit empty details.
Fixes#173
- Renamed XML tags from underscore to kebab-case format for consistency across prompts and system messages.
- Updated context tag from `swarm_context` to `context` in render logic and test assertions.
- Consolidated conditional logic in subagent user prompt by removing duplicate assignment blocks.
- Updated system prompt documentation to reflect kebab-case naming convention for XML tags.
- Added poll_jobs tool for blocking until background jobs complete without manual polling loops.
- Added task.maxConcurrency setting to limit concurrent subagent task execution with semaphore-based control.
- Enhanced task progress tracking to report per-task status with individual timing and token metrics.
- Improved parallel task execution to schedule multiple background jobs independently for true concurrent execution.
- Updated bash and task tool documentation to recommend poll_jobs instead of polling read jobs:// in loops.
- Fixed task progress display to hide tool count and token metrics when zero, reducing visual clutter in status output.
- Refactored async task execution to use job progress reporting API instead of direct update callbacks, improving progress tracking accuracy.
- Implemented pre-allocation of unique IDs for batch tasks to prevent artifact collisions across concurrent executions.
- Extracted async details builder into reusable function to consolidate state construction logic.
- Added null check for details.results before accessing length property to prevent potential crash when rendering results with undefined details.results.
- Apply replaceTabs to all text rendering paths in task tool (output,
task section, context, review explanations, finding bodies)
- Strip <swarm_context> block from expanded task view since it is the
shared context duplicated across every task
- Implemented width-aware text truncation across all tool renderers to respect terminal width constraints and prevent output overflow.
- Enhanced render caching to include width parameter for cache invalidation when terminal width changes, improving cache accuracy.
- Migrated truncateToWidth function imports from render-utils to centralized tui module for better code organization.
- Introduced Hasher utility for generating width-aware cache keys in calculator, find, grep, todo-write, and write tools.
- Optimized box component rendering to only truncate lines that exceed content width, reducing unnecessary truncation operations.
- Implemented caching for tool output block rendering to avoid redundant computations on re-renders.
- Enhanced read tool grouping to prevent coalescing across narrative boundaries (text/thinking blocks between tool calls).
- Improved string preview formatting to detect multi-line strings and display line counts with ellipsis indicators.
- Refactored tool execution component to use reactive render state tracking for better state management.
- Enhanced error handling in tool renderers with logging for failures.
- Fixed truncation indicator to use ellipsis character instead of verbose suffix.
- Added task isolation setting to enable running subagents in isolated git worktrees.
- Made task schema dynamically include isolated parameter only when isolation is enabled.
- Added validation to reject isolated parameter when task isolation is disabled.
- Updated task tool to conditionally render isolation documentation based on feature flag.
- Refactored task schema into factory function to support dynamic isolation configuration.
- Refactored task API to use 'assignment' field instead of 'args' for per-task instructions, enabling clearer separation between shared context and task-specific work.
- Introduced structured context/assignment separation pattern with '<swarm_context>' wrapper for template rendering, replacing placeholder-based substitution.
- Changed agent frontmatter field from 'thinkingLevel' to 'thinking-level' (kebab-case) for consistency with YAML conventions.
- Removed 'context' parameter from ExecutorOptions as context is now prepended at template level rather than executor level.
- Removed 'args' field from AgentProgress and SingleResult interfaces, simplifying task result tracking.
- Updated task rendering to display full task text instead of formatted args, improving clarity in progress output.
- Added task.maxRecursionDepth setting to control how many levels deep subagents can spawn their own subagents (0=none, 1=one level, 2=two levels, -1=unlimited).
- Added nested task artifact naming with parent task prefixes (e.g., '0-Auth.1-Subtask') to support hierarchical task identification.
- Added taskDepth and parentTaskPrefix options to CreateAgentSessionOptions for supporting nested/subagent sessions.
- Changed task tool spawns configuration from 'explore' to '*' to allow spawning any type of subagent.
- Updated system prompt to unconditionally include parallel delegation guidance for all agent types instead of only coordinators.
- Implemented automatic task tool disabling at maximum recursion depth to prevent excessive task nesting.
- Changed TextInput type parameter to string in truncateToWidth(), visibleWidth(), sliceWithWidth(), and extractSegments() functions in native bindings.
- Removed TextInput type export from public API of pi-natives package.
- Removed format.ellipsis symbol from theme configuration, replacing all theme-based ellipsis references with hardcoded Unicode ellipsis character (...).
- Removed ellipsis property from SymbolTheme interface in pi-tui package.
- Removed truncateToWidth, sliceWithWidth, and extractSegments wrapper functions from pi-tui public API, now re-exporting directly from pi-natives.
- Added early return optimizations in truncateToWidth() and sliceWithWidth() native functions to skip native calls when possible.
- Extracted tool choice mapping logic into dedicated utility module for provider-specific format conversions.
- Refactored task summary generation to use template-based rendering instead of string concatenation.
- Migrated system prompt generation to use Handlebars templates with structured data instead of hardcoded strings.
- Extracted JTD to TypeScript conversion logic into dedicated utility module for schema documentation.
- Consolidated prompt template rendering across multiple modules using unified renderPromptTemplate function.
- Simplified type assertions in stream module to use generic OptionsForApi type instead of provider-specific types.
- Renamed 'complete' tool to 'submit_result' throughout codebase for clarity and consistency.
- Updated all tool references, function names, and variable names from 'complete' to 'submit_result' in executor, render, and tools modules.
- Renamed CompleteTool class to SubmitResultTool and CompleteDetails interface to SubmitResultDetails.
- Updated agent prompts and documentation to reference the new 'submit_result' tool name.
- Reorganized task.md prompt to move critical guidance to the top and restructure instructions for better clarity.
- Added ToolChoice type and toolChoice parameter support across all AI providers (OpenAI, Azure OpenAI, Anthropic, Google) enabling fine-grained control over tool/function selection during LLM calls.
- Added toolChoice override capability to Agent.prompt() method and session prompt options allowing callers to control tool selection behavior per request.
- Added provider-specific tool choice mapping functions (mapAnthropicToolChoice, mapGoogleToolChoice, mapOpenAiToolChoice) to normalize tool choice formats across different LLM APIs.
- Removed kernel heartbeat/ping mechanism from PythonKernel, simplifying health monitoring by relying on direct isAlive() checks instead of periodic HTTP requests.
- Added detection and rendering of missing complete tool warning in agent output.
- Introduced extractMissingCompleteWarning function to parse and separate warning messages from agent output.
- Enhanced renderOutputSection to display warning indicators alongside output content.
- Updated agent result rendering to show warning status icon when subagent exits without calling complete tool.
- Removed Prettier configuration files (.prettierignore and .prettierrc) and migrated formatting to Biome.
- Updated Biome configuration from version 2.3.11 to 2.3.12 and changed arrowParentheses rule from 'always' to 'asNeeded'.
- Pinned @biomejs/biome dependency to exact version 2.3.12 in package.json and bun.lock.
- Applied consistent arrow function formatting across 489 files by removing unnecessary parentheses around single parameters.
- Removed blank lines after comment blocks and reorganized imports for consistency across the codebase.
- Removed WASM generation script; use Bun `wasm?raw` loader for imports.
- Added bunfig.toml with loaders for `.md`, `.py`, and `.wasm?raw` text imports.
- Added types/assets/index.d.ts for global TypeScript module declarations.
- Unified TypeScript configuration with tsgo-based checking across monorepo.
- Removed build and WASM steps from install and publish pipelines.
- Added tsconfig.publish.json files to all packages with optimized publish-time configuration.
- Updated all package.json scripts with prepublishOnly hooks for correct type checking during publish.
- Added @oh-my-pi/omp-stats path mappings to root tsconfig.json for consistent imports.
- Added WASM generation script for photon module and integrated into install:dev script.
- Output block border color now reflects state (error, success, warning) for visual feedback.
- Tools use renderStatusLine, renderTreeList, renderFileList, renderCodeCell for unified formatting.
- Write tool shows streaming preview of content being written with syntax highlighting.
- Read tool displays resolved path information when reading from URLs or symlinks.
- Calculator shows expression and output (e.g., `2+2 = 4`) instead of just the result.
- Empty and error states render consistently with status headers across Find, Grep, Ls, Notebook, Ask.
- Created new TUI component library with renderStatusLine, renderOutputBlock, renderCodeCell, renderFileList, and renderTreeList utilities for consistent tool output rendering.
- Refactored 13+ tools to use centralized TUI components, replacing inline theme formatting and ToolUIKit class usage with declarative component APIs.
- Moved tree rendering utilities (getTreeBranch, getTreeContinuePrefix) from tool render-utils to TUI module for improved code organization.
- Streamlined render function return types across tools from Text to Component objects with render/invalidate methods for better state management.
can1357