- Extracted OAuth identifier logic into public functions extractOAuthCredentialIdentifiers and extractOAuthTokenIdentifiers.
- Replaced single credentialIdentity string with multi-identifier resolveCredentialIdentifiers returning string[] for flexible matching.
- Changed credential deduplication from email-based to accountId-based matching in replaceAuthCredentialsForProvider.
- Updated auth-storage tests to verify accountId-prioritized deduplication behavior across soft-disable and hard-delete scenarios.
- Added documentation comments in coding-agent modules explaining partial JSON preservation for streaming tool previews.
- Documented streaming tool preview requirements and render paths in AGENTS.md.
- Added `env` parameter to bash tool for safe environment variable passing without shell re-parsing.
- Added support for rendering partial environment variable assignments in command preview during streaming.
- Updated bash tool prompt to recommend `env` parameter for multiline, quote-heavy, and untrusted values.
- Refactored tool execution component to conditionally merge partial JSON arguments during streaming.
- Added helper functions for environment variable normalization, escaping, and formatting.
- Fixed WebSocket stream fallback logic to safely replay buffered output over SSE when WebSocket fails after partial content has been streamed.
- Added tracking flag to prevent unsafe replays of tool calls and terminal events during fallback transitions.
- Enhanced error recovery to reset output state when replaying buffered content over SSE connection.
- Added docs.rs scraper for extracting Rust crate documentation from rustdoc JSON, supporting modules, functions, structs, traits, enums, and other Rust items with intelligent caching.
- Implemented rustdoc JSON parsing with type rendering for complex Rust types including generics, lifetimes, trait bounds, and qualified paths.
- Added caching layer for rustdoc JSON with date-based versioning for 'latest' releases to reduce repeated fetches.
- Added fallback search strategies in librarian and explore agent prompts for handling empty results.
- Clarified task completion priority in system prompt by prohibiting premature tool call cessation.
- Consolidated and simplified 'Giving Up' guidance in subagent prompt with clearer uncertainty handling.
- Removed duplicate instructions and redundant phrasing to improve prompt clarity and conciseness.
- Added skipPostPromptRecoveryWait option to HandoffOptions for deferring recovery work in handoff operations.
- Added deferred auto-compaction scheduling for threshold-triggered handoffs via post-prompt task queue.
- Extracted handoff document template to dedicated system prompt file for improved maintainability and reusability.
- Changed handoff prompt generation to use template rendering with custom focus instructions support.
- Refactored prompt-in-flight tracking from boolean flag to counter for proper nested operation handling.
- Moved llms.txt endpoint discovery to fallback strategy when rendered page content is low quality, prioritizing page-specific content over site-wide files.
- Enhanced llms.txt endpoint detection to scope candidates to the requested URL path, searching section-specific files before site-wide ones.
- Replaced getOrigin() with buildLlmEndpointCandidates() to generate path-scoped endpoint candidates with depth-based fallback strategy.
- Updated tryLlmEndpoints() to accept full URL and return endpoint metadata alongside content for better fallback tracking.
- Added 2 integration tests validating section-scoped llms.txt discovery and preference for rendered content over site-wide files.
- Clarified contextual pattern mode behavior in ast-grep and ast-edit documentation to explain that results target the selected node, not the outer wrapper.
- Enhanced TypeScript pattern examples to include class method matching syntax with `class $_ { method(...) }` wrapper pattern.
- Removed redundant class method examples that were superseded by improved documentation of contextual pattern mode.
- Renamed parameter names in ast-grep and ast-edit tools from `patterns`/`selector` to `pat`/`sel` for brevity across schema, implementation, and tests.
- Expanded ast-grep and ast-edit tool documentation with 12+ new usage guidelines, examples, and critical notes on pattern syntax, metavariable placement, and error handling.
- Updated CHANGELOG.md to document parameter renames and expanded tool guidance for AST pattern syntax and metavariable usage.
- Reformatted test assertions and type annotations across ast-edit and ast-grep test files for improved readability.
- Added `glob` parameter to `ast_grep`, `ast_edit`, and `grep` tools for filtering files relative to `path`.
- Implemented `combineSearchGlobs()` utility to merge glob patterns from multiple sources instead of throwing errors.
- Changed `grep` tool to combine glob patterns when both `path` and `glob` parameters are provided.
- Updated tool documentation to recommend pairing `path`, `glob`, and `lang` for language-scoped search in mixed repositories.
- Added comprehensive test coverage for combined path and glob parameter handling across grep, ast_grep, and ast_edit tools.
- Added automatic Ollama model capability detection via /api/show endpoint to discover reasoning and input modality support.
- Improved Kagi API error handling with structured error parsing for JSON and plain text response formats.
- Fixed Cerebras streaming compatibility by omitting stream_options.include_usage parameter.
- Simplified API key credential storage to always replace credentials instead of merging for non-minimax providers.
- Updated Kagi Search API key format from 'kagi_...' to 'KG_...' and clarified beta access requirement in provider description.
Fixes#326.
Fixes#321.
Fixes#298.
The context fullness gauge was driven by output token count, causing
erratic jumps between turns (e.g. 84% -> 64%) with no compaction.
Status bar and estimateContextTokens now use calculatePromptTokens()
which returns input + cacheRead + cacheWrite — the actual input context
size. Previously both used a formula that included the final output token
count, which fluctuates with response length and is not part of the
context window for the current request.
isContextOverflow's usage-based fallback (z.ai silent overflow) was
also missing cacheWrite (cache_creation_input_tokens). Per Anthropic
docs the threshold is input + cache_read + cache_creation — all three.
Ref: https://platform.claude.com/docs/en/about-claude/pricing#long-context-pricing
google.ts and google-vertex.ts were double-counting cached tokens.
Gemini's promptTokenCount already includes cachedContentTokenCount, so
assigning input = promptTokenCount and cacheRead = cachedContentTokenCount
overcounted by cachedContentTokenCount on every cached request. Fixed
by subtracting first, matching the OpenAI convention:
input = promptTokenCount - cachedContentTokenCount
cacheRead = cachedContentTokenCount
=> input + cacheRead = promptTokenCount (total prompt, no double-count)
Ref: https://ai.google.dev/api/generate-content#v1beta.GenerateContentResponse.UsageMetadata
All other providers validated: amazon-bedrock (inputTokens is uncached
by API contract), openai-completions/responses/azure (already subtract
cached), kimi/gitlab-duo (delegate to correct implementations), cursor
(API exposes output tokens only — input stays 0 by design).
Co-authored-by: Miroslav Drbal <miroslav.drbal@gendigital.com>
* Add PUPPETEER_PROXY and PUPPETEER_PROXY_IGNORE_CERT_ERRORS env vars
- PUPPETEER_PROXY: routes browser traffic through specified proxy
- PUPPETEER_PROXY_IGNORE_CERT_ERRORS: ignore HTTPS cert errors when set
Made-with: Cursor
* fix(browser): gate PUPPETEER_PROXY_IGNORE_CERT_ERRORS on explicit truthy parse
Previously any non-empty value (including 'false', '0') enabled
--ignore-certificate-errors, silently disabling TLS verification when
operators intended to keep it on. Now only 'true', '1', 'yes', 'on'
(case-insensitive) enable the flag.
Made-with: Cursor
- Added `disabledCause` parameter to credential deletion methods to track reason credentials are disabled.
- Changed credential disabling mechanism from boolean `disabled` flag to `disabled_cause` text field for better auditability.
- Fixed credential purging to respect disabled credentials during email deduplication operations.
- Refactored `replaceAuthCredentialsForProvider()` to update matching credentials instead of deleting all, preserving credential history.
- Added incremental history mode to OpenAI responses .
- Changed OpenAI Codex to exclusively use websockets v2 protocol with fatal error detection for automatic SSE fallback.
- Fixed Gemini model parsing to strip `-preview` suffix for consistent model identification across API calls.
- Improved websocket error handling to extract and report detailed error messages from error events.
- Removed deprecated BETA_RESPONSES_WEBSOCKETS constant and websocket v2 feature flag branching logic.
- Introduced Effort enum and ThinkingConfig metadata for per-model reasoning capabilities with min/max effort levels.
- Migrated thinking level API from string-based ThinkingLevel to structured Effort enum across agent and AI packages.
- Added model-thinking module with effort mapping, policy application, and semantic versioning utilities for provider-specific thinking modes.
- Removed supportsXhigh() function and replaced effort clamping with model-aware validation using ThinkingConfig metadata.
- Expanded models.json with thinking configuration objects for 50+ models including Claude, Gemini, and OpenAI variants.
- Added Python analysis scripts for edit tool usage patterns and tool invocation stream processing.
- Added serviceTier option to OpenAI providers for controlling processing priority and cost across agent, completions, responses, and codex APIs.
- Added providerPayload field to messages for transport-native history reconstruction in OpenAI Responses and Codex APIs.
- Added /fast slash command and serviceTier setting to coding-agent for toggling OpenAI priority mode with fast mode indicator.
- Added remote compaction support with encrypted reasoning preservation for OpenAI models in coding-agent.
- Removed usage caching layer across all providers and refactored UsageFetchContext to eliminate cache and now dependencies.
- Fixed OpenAI Codex streaming service_tier inclusion, provider retry logic with exponential backoff, and email-based credential deduplication.
* idiomatic rust fixes
* idiomatic rust fixes
* display an image if we are fetching an image
* MIME type strictness
* codex nagging me
* codex nagging
* handoff instead of compaction as context filled strategy and surfacing
* handoff instead of compaction as context filled strategy and surfacing p2
* handoff instead of compaction as context filled strategy and surfacing p3
* handoff instead of compaction as context filled strategy and surfacing p4
* handoff instead of compaction as context filled strategy and surfacing p5
* handoff instead of compaction as context filled strategy and surfacing, fixes
* failing fetch test from the fetch tool updates
* handoff focus prompt skeleton
* handoff focus prompt skeleton p2
* fetch bugs
* further codex improvements
* further codex improvements
---------
Co-authored-by: Brit <lol@no.com>
- Added auto-correction logic for off-by-one range start errors in hashline edits.
- Added safety check to prevent false-positive auto-correction when position already includes boundary.
- Added test coverage for off-by-one range correction and duplicate leading line handling.
- Extracted turn-aborted-guidance prompt to external markdown file for better maintainability.
- Corrected provider selection case labels from flat names to namespaced keys (webSearchProvider -> providers.webSearch, imageProvider -> providers.image).
- Expanded web search provider options to include brave, kimi, kagi, and synthetic providers.
- Added RedactedThinkingContent type to support secure encrypted reasoning blocks in Anthropic messages.
- Updated message transformation logic to preserve signed thinking blocks and redacted thinking for latest assistant messages in Anthropic conversations.
- Added handling for redacted_thinking events in Anthropic stream processing with type, data, and index fields.
- Added comprehensive tests validating redacted thinking block preservation and Anthropic thinking immutability during message transformation.
- Removed static thinking mode constant exports (THINKING_LEVELS, ALL_THINKING_LEVELS, ALL_THINKING_MODES, THINKING_MODE_DESCRIPTIONS, THINKING_MODE_LABELS) in favor of dynamic function-based API.
- Renamed formatThinking() to getThinkingMetadata() with return type changed from string to structured ThinkingMetadata object containing value, label, and description.
- Renamed getAvailableThinkingLevel() to getAvailableThinkingLevels() and getAvailableThinkingEffort() to getAvailableThinkingEfforts() with added default parameters for runtime flexibility.
- Updated all consumer modules to use new function-based API instead of static constants, enabling dynamic thinking mode configuration.
- Added `read.defaultLimit` setting to configure default line count for read tool output (default 300 lines).
- Added preset options (200, 300, 500, 1000, 5000 lines) for read default limit in settings UI.
- Updated read tool to distinguish between default and maximum limits per call in prompt documentation.
- Refactored read tool limit logic to use configurable default limit with bounds validation.
- Fixed provider session state not being cleared when branching or navigating tree history, preventing resource leaks with codex provider sessions.
- Added calls to `#closeCodexProviderSessionsForHistoryRewrite()` in branch and navigateTree methods to ensure proper cleanup.
- Added test coverage for provider session cleanup during history branching and tree navigation.
- Removed complex fuzzy matching logic including Levenshtein distance calculation and similarity scoring in favor of a simpler glob-based suffix pattern approach. Replaced findReadPathSuggestions with findUniqueSuffixMatch that returns a path only when exactly one candidate matches, eliminating ambiguous suggestions. Removed legacy ttsr_trigger and ttsrTrigger fields from RuleFrontmatter interface.
- Added fallback to parse the legacy 'thinking' field when 'thinkingLevel' is not provided. The 'thinkingLevel' field takes precedence when both are present. Added tests to verify backward compatibility and field precedence.
- Added kebabToCamel and normalizeKeys utility functions to convert kebab-case keys to camelCase recursively. Updated parseFrontmatter to normalize all parsed keys, ensuring consistent camelCase property access throughout the codebase. Updated all frontmatter key accesses to use camelCase notation (e.g., thinkingLevel, spdxId).
- Updated option interfaces, param builders, and stream functions to use unified `reasoning` field.
- Added `resolveOpenAiReasoningEffort()` to centralize xhigh clamping logic.
- Replaced type casts with `castApi()` helper and fixed test option names.
- Updated coding-agent and benchmark references for consistency.
- Extracted thinking module with ThinkingEffort, ThinkingLevel, and ThinkingMode types to centralize reasoning configuration across packages.
- Migrated ThinkingLevel type from pi-agent-core to pi-ai package with new validation functions parseThinkingLevel() and getAvailableThinkingLevel().
- Consolidated thinking level constants and descriptions into reusable exports (ALL_THINKING_LEVELS, THINKING_MODE_DESCRIPTIONS) for consistent UI display.
- Removed local thinking-effort-label utility and replaced formatThinkingEffortLabel() with centralized formatThinking() function from pi-ai.
- Refactored thinking mode handling to distinguish ThinkingSelector (user-facing with 'off' option) from ThinkingEffort (provider-level).
- Added formatRoleThinkingModeLabel helper to display 'inherit' for default thinking mode, preventing badge ambiguity when multiple roles share the same model. Enhanced role menu labels to include role tags for clarity. Fixed model resolver to avoid substring matching that could incorrectly resolve exact model IDs to similar variants.
- Removed ancestor directory walking logic from codex and opencode skill loading. Simplified project skills discovery to use single project directory instead of scanning up the filesystem tree. Fixed variable shadowing issues in gemini and github context file loading.
NanoGPT exposes per-thinking-level model IDs (e.g.
anthropic/claude-opus-4.6:thinking:low) as separate entries. These
pollute the model list and collide with parseModelString's thinking
level suffix stripping.
Filter out all :thinking and :thinking:level suffixed nanogpt models
during discovery. When a :thinking variant existed, mark the
corresponding base model as reasoning-capable. NanoGPT already supports
reasoning_effort via the OpenAI-compat path so no provider-side model
ID suffixing is needed.
- Added buildCompactHashlineDiffPreview() function to generate collapsed diff previews for tool responses.
- Enhanced EditTool response to include diff summary with added/removed line counts and compact preview.
- Added project-level discovery for .agent/ and .agents/ directories walking up to repository root.
- Implemented helper functions for truncating long diff runs (collapseFromStart, collapseFromEnd, collapseFromMiddle).
- Clarified pos semantics for replace, prepend, and append operations with concise bullet format.
- Simplified lines parameter documentation to focus on practical usage patterns.
- Rewrote rules section to emphasize structural anchoring and when to use prepend/append versus range replace.
* feat: monorepo-friendly discovery for AGENTS.md and skills
Support hierarchical config discovery in monorepos by walking up from
cwd through ancestor directories, combining files from all levels
instead of only checking the immediate working directory.
AGENTS.md context files:
- Change dedup key from file.level to file.level + depth, so context
files at different directory levels coexist instead of shadowing
- Root AGENTS.md and sub-project AGENTS.md are both included in the
system prompt, ordered from least specific to most specific
Skills:
- Modify all 4 providers with project-level skill directories (native,
claude, codex, opencode) to walk up from cwd through ancestors,
scanning for skills at each level
- Skills at closer directories win on name conflicts via existing
name-based dedup
Repo root boundary:
- Add findRepoRoot() utility that detects .git to identify repo root
- Add repoRoot field to LoadContext, computed once in loadCapability()
- All walk-up traversals (AGENTS.md, skills, nearest config dir) stop
at the repo root, preventing discovery from leaking outside the repo
- When not in a git repo, falls back to walking to filesystem root
Files changed:
- capability/fs.ts: add findRepoRoot()
- capability/types.ts: add repoRoot to LoadContext
- capability/index.ts: compute repoRoot in loadCapability()
- capability/context-file.ts: depth-aware dedup key
- discovery/agents-md.ts: bound walk-up at repo root
- discovery/builtin.ts: getAncestorDirs stopAt param, skill walk-up
- discovery/claude.ts: skill walk-up with repo root bound
- discovery/codex.ts: skill walk-up with repo root bound
- discovery/opencode.ts: skill walk-up with repo root bound
- extensibility/skills.ts: add repoRoot to inline LoadContext
* feat(agents-provider): add project-level discovery with ancestor walk-up for all capability types
The agents provider (.agent/.agents directories) previously only loaded
capabilities from the user home directory (~/.agent/, ~/.agents/). This
meant project-level .agents/ directories in monorepo roots were not
discovered when sessions started from subdirectories.
Add getProjectPathCandidates() helper that walks from cwd up to repoRoot,
scanning both .agent/ and .agents/ at each ancestor level. Apply this to
all six capability types: skills, rules, prompts, commands, context files
(AGENTS.md), and system prompts (SYSTEM.md). This matches the ancestor
walk-up behavior already present in the builtin (.omp), claude, codex,
and opencode providers.
All loaders now parallelize project-level and user-level scans via
Promise.all. Project-level results appear closest-first so dedup at the
capability layer picks the nearest override.
* fix(discovery): remove hard cap of 20 on ancestor directory walking
All walk-up loops already terminate naturally at repoRoot or filesystem
root. The depth < 20 / .slice(0, 20) / MAX_DEPTH caps were redundant
safety guards that would silently stop discovery in deeply nested
projects.
Removed from: agents.ts, agents-md.ts, builtin.ts, claude.ts, codex.ts,
opencode.ts, and corresponding test files.
* fix(agents-provider): set depth on project-level ContextFile entries for correct dedup
loadContextFiles was creating project-level ContextFile entries without
a depth field. Since contextFileCapability.key uses
`project:${file.depth ?? 0}`, all ancestor-level AGENTS.md files
collapsed to the same key and only the first survived dedup.
Compute depth via calculateDepth(cwd, ancestorDir) where ancestorDir is
two levels up from the file path (past the .agent/.agents config dir).
This gives each ancestor level a distinct dedup key.
* fix(discovery): stop ancestor walk-up at $HOME when not in a git repo
When repoRoot is null (no .git found), walk-up loops traversed all the
way to filesystem root. This caused $HOME to be scanned as a project-
level directory and then again as user-level, producing duplicates.
All providers now use `ctx.repoRoot ?? ctx.home` as the stop boundary:
stop at repoRoot if in a repo, otherwise stop at home. Applied to
agents.ts, agents-md.ts, builtin.ts, claude.ts, codex.ts, and
opencode.ts.
* fix(context-files): clamp depth >= 0 in dedup key and fix provider depth computations
The dedup key `project:${file.depth}` used raw depth, which could be
negative when providers computed it from config subdirectories (e.g.
.claude/, .github/, .gemini/) rather than the ancestor directory. This
caused same-scope cwd-level files to get distinct keys like project:-1
and project:0, bypassing dedup and injecting conflicting instructions.
Two-layer fix:
1. Key function: clamp to Math.max(0, depth) so any file at or below
cwd is treated as cwd-scope (depth 0). Defensive against future
providers.
2. Providers: fix root cause in claude.ts, gemini.ts, github.ts to
compute depth from the ancestor directory (parent of the config
subdir), not the config subdir itself.
---------
Co-authored-by: Can Bölük <can1357@users.noreply.github.com>
parseModelString now extracts valid thinking level suffixes (e.g.,
"anthropic/claude-opus-4-6:high") instead of treating them as part of
the model ID. This enables per-role thinking levels in config:
modelRoles:
slow: anthropic/claude-opus-4-6:high
default: anthropic/claude-opus-4-6:low
smol: google/gemini-3-flash:medium
The thinking level is applied at startup, in SDK fallback resolution,
and during Ctrl+P role cycling. The original config string is preserved
on role cycle so the suffix round-trips correctly.
* idiomatic rust fixes
* idiomatic rust fixes
* display an image if we are fetching an image
* MIME type strictness
* codex nagging me
* resize so we do not blow up the terminal we are in
* codex nagging
* codex nagging
* codex nagging
* codex nagging
* codex nagging
---------
Co-authored-by: Brit <lol@no.com>
- Added language module with 39 supported languages and tree-sitter parser bindings for ast-grep integration.
- Exported dedupeParseErrors utility function to remove duplicate parse error messages in tool output.
- Fixed duplicate parse error messages when multiple patterns fail on the same file.
- Replaced language alias lookup with O(1) phf::Map and added 30+ new language aliases (golang, julia, toml, zig, etc.).
- Migrated from ast-grep-language to direct tree-sitter language dependencies for improved language support.
- Added `fetch.useKagiSummarizer` configuration setting to toggle Kagi Universal Summarizer usage in fetch tool.
- Updated fetch tool to conditionally apply Kagi summarization based on configuration setting.
- Added comprehensive test coverage for Kagi summarizer toggle behavior with mocked dependencies.
- Added Bun version validation at startup to enforce minimum version 1.3.7 for JSONL session parsing.
- Fixed `/resume` and `--resume` flags silently failing on older Bun runtimes by adding version check.
Fixes#277
can1357