- Replaced `===== ... =====` eval cell headers with `*** Begin ` / `*** End ` markers; legacy format remains renderable in HTML exports.
- Replaced hashline patch grammar with `*** Begin Patch` / `*** End Patch` envelope; old inputs without the envelope are still accepted.
- Extracted `sniffEvalLanguage` into a shared `sniff.ts` module reused by the parser and tool.
- Added `docs/ERRATA-GPT5-HARMONY.md` and `scripts/session-stats/harmony_backtest.py` documenting and backtesting the GPT-5 Harmony-header leak defect.
Anthropic counts sessions by metadata.user_id. Without this fix, OMP
generated fresh random entropy on every API request, inflating the
session count and preventing backend attribution to the authenticated
account.
Changes:
packages/ai:
- resolveAnthropicMetadataUserId() now accepts JSON-format user_id
matching real Claude Code's getAPIMetadata shape
({ session_id, account_uuid, ... }). Previously only the legacy
cloaking format was accepted on OAuth, causing stable caller-supplied
values to be silently discarded.
- AnthropicOAuthFlow.exchangeToken() and refreshAnthropicToken() now
populate OAuthCredentials.{accountId, email} from the token response
account block, removing the need for a separate /api/oauth/profile
round-trip.
- AuthStorage.getOAuthAccountId(provider, sessionId) returns the OAuth
accountId for the session-sticky credential, used to build
account_uuid in metadata.user_id. Guards against misattribution for
API-key, runtime-override, env-key, and fallback-resolver paths that
do not record a session credential.
packages/agent:
- Agent.metadataForProvider(provider) resolves request metadata for
the given provider via the installed resolver, or returns the static
metadata value. The plain metadata getter now returns only the static
value; provider-aware resolution is explicit.
- Agent.setMetadataResolver(fn) installs a (provider: string) resolver
evaluated per LLM request in agent-loop, after getApiKey records the
session-sticky credential, so account_uuid reflects the credential
actually used.
- AgentLoopConfig.metadataResolver is called with config.model.provider
after getApiKey, overriding the static metadata field.
packages/coding-agent:
- AgentSession.#syncAgentSessionId installs a metadata resolver that
builds { user_id: JSON.stringify({ session_id, account_uuid? }) },
matching the Anthropic session attribution format. account_uuid is
only included for provider="anthropic" to avoid leaking the OAuth
identity to third-party Anthropic-format-compatible providers.
- sessionId getter prefers providerSessionId when supplied via
AgentSessionConfig so all API paths (getApiKey, direct calls,
metadata resolver) share the same provider-facing session ID.
- prepareSimpleStreamOptions stamps session metadata on direct calls
(runEphemeralTurn, compaction, branch summary, title generation) so
they share the same session bucket as Agent.prompt requests.
- generateBranchSummary and generateSessionTitle accept a
(provider: string) metadata resolver evaluated after their own
getApiKey call for correct credential attribution.
- Added a boundary normalizer that validated tool output shape and replaced malformed responses with a fallback text-only result.
- Updated tool execution to coerce both streaming partial updates and final tool results through that normalizer.
- Set the tool call error state when a malformed result was detected by the coercer.
- Added `readSseEvents` and `ServerSentEvent` exports in utils for reusable SSE stream parsing.
- Replaced Anthropic's local SSE parser with shared `readSseEvents(response.body, signal)` decoding.
- Updated abort handling in agent stream loop to race an `ABORTED` sentinel with `responseIterator.next()`.
- Expanded stream tests for `readSseEvents` parsing of CRLF, comments, split UTF-8 chunks, and trailing events.
- Added an optional `getReasoning` callback to `AgentLoopConfig` to resolve reasoning effort dynamically for each LLM call.
- Updated the agent loop to resolve reasoning via `getReasoning` and use it in place of static `reasoning` when provided.
- Added a test confirming a run re-reads the thinking level between consecutive model calls when it changes mid-run.
agentLoop/agentLoopContinue IIFEs were fire-and-forget: any error thrown
inside runLoop (e.g. missing API key via getApiKey callback) became an
unhandled rejection and left the EventStream permanently hung, so the
for-await consumer in Agent#runLoop never resumed its catch block.
- Add EventStream.fail(err): rejects all waiting consumers, stores the
error so queue-draining consumers throw after the queue empties, and
rejects finalResultPromise (guarded by a no-op .catch() to prevent a
secondary unhandled rejection when nobody awaits result()).
- Wrap runLoop calls in both agentLoop and agentLoopContinue IIFEs with
try/catch that routes errors into stream.fail(err), letting the
existing Agent#runLoop catch block convert them to error assistant
messages.
- Fixed abort-source handling so caller aborts always win and local reasons only attach to matching request signals.
- Fixed agent-loop streaming to race event reads against abort signals and emit an aborted assistant message.
- Fixed Anthropic request construction to honor thinkingEnabled=false and omit temperature/top_p/top_k for Opus non-thinking.
- Fixed OpenAI Codex request handling by normalizing response URLs, decoding non-string websocket frames, and cleaning handshake headers.
- Added regression tests for abort precedence, Anthropic alignment cases, and Codex stream/header normalization.
- Documented the cancellation and provider behavior fixes in package changelogs.
- Replaced `AgentTool` `nointent` and `deriveIntent` with a unified `intent` option that supports `omit`, `optional`, `require`, or a callback function.
- Updated agent tool schema injection and execution paths to inject `_i` as required/optional/omitted and to derive intent through the new `intent` callback.
- Aligned atom-edit tests with the updated default sed behavior (`g` now off by default and explicit `g: true` for global replacement).
- Added `nointent` and `deriveIntent` to `AgentTool`, and made intent-schema injection honor `PI_NO_INTENT` plus per-tool opt-outs.
- Updated tool execution and streaming UI handling to derive a fallback intent when `_i` is absent, without aborting on derivation failures.
- Marked several coding tools as `nointent` and added `deriveIntent` callbacks where needed, then relaxed prompt language to indicate intent is present on most tools.
- Changed the intent marker from "i" to "_i" in the agent loop path.
- Updated intent extraction to always destructure out the intent key and return stripped arguments even when intent is not a string.
- Updated the agent loop intent marker constant to use `i` instead of `_i` for tracing fields.
- Updated intent tracing documentation to describe a generic string marker field and cleanup behavior.
- Adjusted related Anthropic and coding-agent tests to match the renamed intent field via shared `INTENT_FIELD` usage.
Slots a new "apply_patch" variant alongside the existing edit modes
(replace, patch, hashline, chunk, vim). The mode accepts a single input
string containing a Codex *** Begin Patch / *** End Patch envelope,
parses it with a new lenient parser (heredoc-tolerant), and fans each
file-op out to the existing executePatchSingle so LSP writethrough,
plan-mode guards, fs-cache invalidation and diagnostics are shared
with the patch mode.
Exposes both tool shapes from the spec: the JSON function-tool variant
(§1.2, {input: string}) and the OpenAI custom-tool / Lark-grammar
"freeform" variant (§1.1, raw patch string). The edit tool advertises
a Lark grammar via customFormat and a wire name via customWireName;
openai-responses emits it as a grammar-constrained custom tool when a
model opts in with applyPatchToolType: "freeform" in models.json.
custom_tool_call / custom_tool_call_output are plumbed end-to-end
through the shared responses code (emission, streaming, history
replay), and the agent-loop dispatcher matches tool calls by either
name or customWireName so returned calls route correctly.
Also threads preview/diff rendering for apply_patch through the TUI
(tool-execution + edit renderer) so streaming patches show per-file
diffs like the other edit modes.
Default edit mode is unchanged (hashline); opt in via edit.mode or
PI_EDIT_VARIANT=apply_patch.
- agent-loop: sanitize text content in tool_execution result/partialResult
- ai/cursor: fix ANSI escape handling, add incomplete escape detection
- coding-agent/cursor: fix per-delta sanitization with tracked state
- print-mode: flush stderr before exit to prevent data loss
- add unit tests for bash execution clamp display line
- Added `onAssistantMessageEvent` callback to Agent API for inspecting and aborting assistant streaming events.
- Added `setAssistantMessageEventInterceptor()` method to dynamically update assistant message event handlers.
- Converted `checkAutoGeneratedFileContent()` from async to synchronous for improved streaming edit abort detection performance.
- Implemented LRU caching in auto-generated file detection with early path-based checks to prevent unnecessary edits.
- Refactored streaming edit pre-caching to use assistant message event interception for real-time abort capability.
- Extracted `peekFile()` utility for efficient file prefix reading with pooled buffer reuse strategy.
* Add MCP tool discovery search and live refresh
* Fix MCP discovery review feedback
* Address remaining MCP discovery review comments
* feat: compact MCP discovery search results
* fix: align MCP discovery search contract
* feat: add MCP server tool counts to discovery hints
* fix(agent): corrected stale toolChoice validation against active tools
- Fixed stale forced toolChoice passed to provider after mid-turn tool refresh by validating against active tools.
- Added refreshToolChoiceForActiveTools() to filter invalid tool choices when available tools change.
- Changed getToolChoice config to use computed function instead of static property for dynamic validation.
- Fixed MCP tool selection tracking in coding-agent to distinguish between discovery-enabled and non-discovery sessions.
- Updated search_tool_bm25 to filter already-selected tools before applying limit parameter.
---------
Co-authored-by: can1357 <me@can.ac>
- Extracted tool result emission logic into dedicated emitToolResult function to eliminate duplication.
- Consolidated tool execution result handling to emit results immediately after execution rather than deferring to post-processing loop.
- Simplified post-execution loop by delegating result emission to emitToolResult and removing redundant message construction.
- Renamed intent field from `agent__intent` to `intent` in tool schemas for cleaner API contracts.
- Refined intent parameter guidance to require concise 2-6 word present participle sentences.
- Updated intentTracing documentation and test fixtures to reflect the new field naming convention.
- Added lenientArgValidation option to tools for graceful handling of argument validation errors.
- Refactored schema reference resolution to inline all $ref definitions instead of preserving them at root level.
- Added circular reference detection during schema resolution to prevent infinite loops.
- Added AJV compilation verification to catch unresolved $ref references before tool execution.
- Added fallback empty string for undefined tool descriptions across all AI providers to prevent runtime errors.
- Made `description` field required in CustomTool interface and normalized tool descriptions in agent-loop.
- Refactored tool normalization in agent-loop by renaming `injectIntentIntoTools()` to `normalizeTools()` with conditional intent injection.
- Added comprehensive description metadata to todo-write tool schema fields for improved clarity.
Move agent__intent description from per-tool JSON schema injection (repeated
once per tool) into a single conditional block in the system prompt template.
Token savings scale with tool count.
- agent-loop: remove description from intent schema injection
- system-prompt.md: add intentTracing/intentField template block
- sdk: compute intentField once, pass through to prompt builder
- system-prompt.ts: thread intentField option into template context
- Renamed `file` parameter to `path` in edit tool schemas and all test cases for consistency.
- Removed `file` property fallback from EditRenderArgs interface and path resolution methods.
- Refactored browser viewport schema to use snake_case `device_scale_factor` with explicit camelCase mapping for Puppeteer API.
- Added hashline edit rule requiring `end` tag to include closing braces/brackets when replacing blocks.
- Refactored INTENT_FIELD injection logic to conditionally reorder schema properties and update required array.
- Updated tool result messages to include error details when tool execution fails.
- Modified createAbortedToolResult() to accept optional errorMessage parameter for enhanced error reporting.
- Changed hashline format separator from pipe (|) to colon (:) for improved readability across all tools and output formats.
- Refactored hashline edit API with operation-based structure: renamed delete->rm, rename->mv, set->target/new_content, and added explicit op field for operation types.
- Updated hashline hash encoding from 4-character base36 to 2-character hexadecimal for more compact representation.
- Replaced anchor terminology with tags throughout hashline documentation and API for clearer semantics.
- Added streamed tool intent display in working message to show real-time intent tracking during agent execution.
- Changed intent tracing field name from `$intent` to `_intent` across tool schemas and agent core for consistency.
- Added support for file deletion and renaming operations in hashline edit mode.
- Renamed hashline edit operation fields: `set` to `target`/`new_content`, `set_range` to `first`/`last`/`new_content`, `insert` to `inserted_lines`.
- Added optional `intent` field to `ToolCall` interface for capturing harness-level intent metadata.
- Added `intentTracing` configuration option to enable intent goal extraction from tool calls with automatic `$intent` field injection and argument stripping.
- Implemented intent injection and extraction logic in agent-loop to populate tool call intent metadata when intentTracing is enabled.
- Added `tools.intentTracing` setting to coding-agent configuration schema with environment variable override support.
- Migrated from Node.js to Bun runtime by replacing @types/node with @types/bun and bun-types dependencies.
- Downgraded @types/diff from ^8.0.0 to ^7.0.2 in coding-agent package.
- Updated tsconfig.base.json to include DOM.AsyncIterable in lib configuration for async iterable DOM API support.
- Simplified stream handling in update-cli.ts by removing Readable.fromWeb() conversions and using response.body directly with pipeline().
- Simplified TextDecoderStream instantiation in stream.ts to use default parameters instead of explicit UTF-8 encoding configuration.
- Added defensive null/undefined handling in agent-loop.ts result assignment using nullish coalescing operator.
- Added `concurrency` option to `AgentTool` interface supporting 'shared' (default, parallel execution) and 'exclusive' (solo execution) modes.
- Implemented parallel execution of shared tools within a single agent turn with ordered result emission.
- Refactored tool execution to support concurrent scheduling with interrupt handling via steering messages and AbortSignal.
- Added steering message checking mechanism that interrupts tool execution when user messages arrive during tool processing.
- Removed HTTP proxy setup code from stream.ts that was conditionally setting up undici global dispatcher.
- 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.
- Added nonAbortable option to AgentTool interface for ignoring abort signals during tool execution.
- Added isolated option to task tool for running tasks in isolated git worktrees.
- Simplified tool parameter descriptions across all coding-agent tools for better readability.
- Updated find tool to always sort results by modification time and removed sortByMtime parameter.
- Changed web-fetch timeout default from 20s to 45s maximum and updated bash prompt to use cwd parameter.
- Removed model parameter from task and gemini-image tools to use session/provider defaults.
- Added ToolCallContext interface with batch execution metadata including batchId, index, and total.
- Enhanced getToolContext() to accept optional ToolCallContext parameter for batch-aware tool execution.
- Implemented LSP batching support to coalesce formatting and diagnostics operations on parallel edits.
- Added comprehensive tests for tool call batch context and LSP writethrough batching functionality.
- Added proper tool result messages for tool calls that are aborted or error out.
- Ensured tool_use/tool_result pairing is maintained when tool execution fails.
- Port pi-ai from upstream with Bun-first approach (source files, no dist)
- Add cli_ prefix to tool names in OAuth mode to avoid collisions
- Improve beta header handling with proper deduplication
- Convert codex-instructions.md to Bun text embed
- Strip .js extensions from all imports
- Added environment variable management via settings.json with applyEnvironmentVariables() and related methods.
- Enhanced file mentions to support directories, validate boundaries, and truncate large files.
- Added fallback model candidate logic in title generator to retry on model failures.
- Fixed bash output preview to minify JSON content and skip structural-only lines.
- Added Cerebras provider support with thinking block filtering and zai-glm-4.6 model priority.
- Refactored tree rendering to use consistent theme.tree properties instead of boxSharp.
- Changed isError property from required to optional across ToolResultMessage and tool event interfaces.
- Added SpinnerType variants with distinct animation frames per symbol preset.
- Fixed spinner animation crash when frames array is empty by adding guard clause.
- Refactored agent-loop tool execution to consolidate toolCallId, toolName, and isError into details object.
- Added Time Traveling Streamed Rules (TTSR) feature for real-time pattern matching and rule injection during streaming.
- Added TtsrManager with configurable settings for context mode, repeat mode, and repeat gap.
- Added TTSR trigger support across all rule discovery sources (builtin, cline, cursor, windsurf).
- Added abort signal handling in agent loop to support stream interruption for TTSR.
- Added TtsrNotificationComponent for displaying rule violation notifications in interactive mode.
- Added interruptMode option to control when queued messages interrupt tool execution.
- Implemented "immediate" mode (default) to check queue after each tool and interrupt remaining tools.
- Implemented "wait" mode to defer queue processing until the entire turn completes.
- Exposed interruptMode configuration in interactive settings UI with immediate/wait options.
- Wired interrupt mode through AgentSession and SDK to enable runtime configuration and persistence.
- Renamed npm package scope from @mariozechner to @oh-my-pi across all packages for consistent branding.
- Removed packages/pods directory and all related pod management functionality.
- Updated import paths and module declarations throughout codebase to reflect new package names.
- Reformatted code with consistent trailing comma removal and ternary operator alignment.
- Added MCP (Model Context Protocol) support with stdio and HTTP transports for external tool integration.
- Implemented plugin system with install, enable, configure, and doctor commands for extensibility.
- Added 22 Exa MCP tools for web search, LinkedIn, company research, and websets operations.
- Introduced LSP tool with rust-analyzer support for code intelligence (diagnostics, hover, references, code actions).
- Added task tool for delegating work to specialized subagents with parallel execution support.
- Implemented web_search and web_fetch tools with Anthropic and Perplexity provider support.
- agentLoop now accepts AgentMessage[] instead of single message
- agent.prompt() accepts AgentMessage | AgentMessage[]
- Emits message_start/end for each message in the array
- AgentSession.prompt() builds array with hook message + user message
- TUI now receives events for before_agent_start injected messages
can1357