Skip to content

RustyCode v0.4.0

Choose a tag to compare

@github-actions github-actions released this 11 May 23:56
· 85 commits to main since this release

[0.4.0] - 2026-05-11

Features

  • bench: Add token usage tracking to TrialResult and BenchAgent trait
  • bench: Add token/cost fields to BenchmarkResults and TaskResult
  • tools-api: Add InputValidator, TokenAccountant, PrivilegeGate traits
  • bench: Pre-validate required tool args before execution
  • Add GA4 analytics module + tool schema/annotations refactor
  • Bench multi-provider, task_context module, bus/team tools, production hardening status doc
  • llm: Add schema, wire, transport, auth, route modules for provider abstraction
  • llm: Add shared ModelCache with dynamic provider model discovery
  • llm: Wire ModelCache into Gemini and OpenAI providers
  • tui: Async service init with loading screen
  • tui: Async service init with loading screen
  • tui: Show loading screen during service init

Bug Fixes

  • Resolve 22 bugs across 7 crates from systematic bug audit
  • Resolve 12 bugs across orchestration, CLI, and TUI (wave 2)
  • Resolve pre-existing clippy errors and test compilation issues
  • Resolve 3 confirmed bugs from Oracle verification
  • Resolve 30+ bugs across 10 crates (waves 5-6)
  • Resolve 10 critical/high/medium bugs across 5 crates (wave 7)
  • Replace silent error suppression with proper logging and remove debug artifacts
  • acp: Resolve lock-across-await concurrency bugs in prompt handler and LLM integration
  • Resolve 22 clippy warnings across 20 files (round 5)
  • acp: Update test to match empty messages behavior
  • Remove unused sha2 imports after crypto centralization
  • Resolve 4 resource leaks — zombie processes, missing Drop, unenforced timeout
  • Replace 15+ scattered len()/4 token heuristics with canonical estimate_tokens()
  • security: Reject ALL tools on approval timeout, not just dangerous ones
  • security: Harden input validation — LSP Content-Length cap + env injection allowlist
  • robustness: Eliminate .expect() panics in signal handlers and runtime creation
  • concurrency: Async correctness — cancellation, timeout abort, deadlock elimination
  • orchestration: Tool error status, stream handling, phase transitions, input logic
  • llm: Repair broken references after type extraction to types/
  • llm: Real Sigv4 signing, random route hash, task contracts
  • leaks: Eliminate zombie child processes and fire-and-forget spawns
  • concurrency: Eliminate TOCTOU races and await_holding_lock
  • io: Add file size guards to prevent OOM from unbounded reads
  • medium: Preserve error context, use async I/O, remove unnecessary clone
  • tests: Enable structured output in git test helper and fix edit assertion
  • tui: Version display, Ctrl+W behavior, and tool name assertion
  • llm: Guard Gemini toolConfig when no tools present, add startup perf instrumentation
  • llm: Strip SSE data: prefix before protocol parsing
  • llm,tui: Gemini provider reliability — safety stops, rate limit retry, and SSE tracing
  • agent-runtime: Classify rate limit errors as Transient, not Fatal
  • llm: Omit tools/toolConfig keys from Gemini JSON when absent

Refactor

  • Remove dead code from todo.rs
  • tui: Remove dead code, disabled examples, and unused helpers
  • tools: Remove dead code and unused plan template methods
  • core: Remove dead code, unused variants, and retry prompts
  • Remove unnecessary lint suppressions and fix pre-existing clippy issues
  • llm: Replace map_or with idiomatic is_none_or/is_some_and
  • Apply clippy-driven cleanups across 7 files
  • Session-scoped worktree CWD + tracing instrumentation
  • Remove CLI bench command, relocate headless utils, clean gitignore
  • Consolidate duplicate test_ctx() helpers into shared module
  • Replace verbose empty JSON object pattern with json!({})
  • Deduplicate format_timestamp between CLI and storage
  • Consolidate strip_ansi_escapes into rustycode-protocol
  • Consolidate truncate, normalize_tool_name, and format_duration duplicates
  • llm: Add empty types/ module skeleton for provider.rs extraction
  • llm: Extract message types to types/message.rs
  • llm: Migrate Cohere provider from inline serialization to Route+CohereProtocol
  • llm: Migrate Ollama provider from inline serialization to Route abstraction
  • llm: Migrate Gemini provider from inline serialization to Route+GeminiProtocol
  • llm: Migrate Bedrock provider from inline serialization to Route+BedrockProtocol
  • llm: Migrate OpenAI provider from inline serialization to Route+Protocol
  • runtime: Migrate String errors to thiserror typed enums

Documentation

  • Add production hardening plan, status, and agent-bench CI workflow
  • Add Plan 2 (Provider Migration) implementation plan
  • Update production hardening status with audit results

Chores

  • bench: Add native CI benchmark task definitions
  • Consolidate dependency versions to workspace references
  • Upgrade bollard 0.17→0.21, jsonschema 0.26→0.46, consolidate more deps
  • Add release profile optimizations (59MB → 35MB binary)
  • Upgrade bollard 0.17→0.21 API, add sha2 to protocol, optimize hash formatting
  • Remove unused sha2 dep from orchestration crate
  • Fmt fix for text.rs test assertion
  • Bump version to 0.4.0