v0.5.1

Adds CI-built release binaries (Linux x86_64, macOS x86_64 + aarch64, Windows x86_64). See CHANGELOG.md.

v0.5.0

See CHANGELOG.md for details.

TyCode v0.4.0

Changelog

All notable changes to TyCode are documented here.

[0.4.0] — 2026-05-30

GitHub release: v0.4.0. (Versions 0.0.0–0.3.0 were consumed by earlier
publish/unpublish cycles on npm and cannot be reused, so this release lands on
0.4.0.)

Model interaction & request handling

• Interruptible agent: Esc (or Ctrl+C) now aborts the in-flight model
  stream mid-token and stops the agent loop immediately. Partially streamed text
  is finalized and remaining tool calls get synthetic results so the transcript
  stays valid for the next turn.
• Automatic retry with backoff: transient provider failures (HTTP 429/5xx,
  rate limits, network/connection/timeout errors) retry up to 4 times with
  exponential backoff and jitter instead of killing the run. Each retry is shown
  inline and is itself cancellable.
• Concurrent tools: independent read-only tool calls (file_read, grep,
  glob_search, directory_tree, web_fetch, …) in a single batch run in parallel —
  a multi-read fan-out now completes in roughly the time of the slowest call.
  Mutating tools and dangerous shell commands stay sequential and gated on
  confirmation, preserving correctness and ordering.
• Streaming efficiency: SSE/NDJSON parsing in the Anthropic, OpenAI, and
  Ollama providers no longer reallocates the whole buffer per event
  (O(n²) → O(n) via drain); the OpenAI path also no longer risks
  double-parsing a record split across chunk boundaries.

Tools (0.2.0 line, also in 0.1.0 → 0.2.0)

• todo_write / todo_read: structured, session-persistent task list.
• multi_edit: multiple exact-string edits to one file, applied atomically.
• web_fetch: fetch a URL and strip HTML to readable text.
• Project memory now auto-loads CLAUDE.md and AGENTS.md alongside
  TYCODE.md / README.md.

Fixes

• Ctrl+Enter tool-call expand/collapse was unreachable (shadowed by plain
  Enter) — now works.
• Cleared all outstanding compiler warnings; the build is warning-free.

[0.1.0] — 2026-05-10

Initial public release.

• Multi-provider agent: Ollama (local), Anthropic, OpenAI, Google Gemini.
• Tool suite: file read/write/edit, shell with real timeouts, grep, glob,
  HTTP, process management, system info, directory tree.
• Autonomous agent loop with streaming responses.
• TUI with markdown rendering, inline expand/collapse of tool output,
  mouse scroll, dangerous-command confirmation, and auto context compaction.

TyCode v0.2.0

Changelog

All notable changes to TyCode are documented here.

[0.2.0] — 2026-05-30

GitHub release: v0.2.0

Model interaction & request handling

• Interruptible agent: Esc (or Ctrl+C) now aborts the in-flight model
  stream mid-token and stops the agent loop immediately. Partially streamed text
  is finalized and remaining tool calls get synthetic results so the transcript
  stays valid for the next turn.
• Automatic retry with backoff: transient provider failures (HTTP 429/5xx,
  rate limits, network/connection/timeout errors) retry up to 4 times with
  exponential backoff and jitter instead of killing the run. Each retry is shown
  inline and is itself cancellable.
• Concurrent tools: independent read-only tool calls (file_read, grep,
  glob_search, directory_tree, web_fetch, …) in a single batch run in parallel —
  a multi-read fan-out now completes in roughly the time of the slowest call.
  Mutating tools and dangerous shell commands stay sequential and gated on
  confirmation, preserving correctness and ordering.
• Streaming efficiency: SSE/NDJSON parsing in the Anthropic, OpenAI, and
  Ollama providers no longer reallocates the whole buffer per event
  (O(n²) → O(n) via drain); the OpenAI path also no longer risks
  double-parsing a record split across chunk boundaries.

Tools (0.2.0 line, also in 0.1.0 → 0.2.0)

• todo_write / todo_read: structured, session-persistent task list.
• multi_edit: multiple exact-string edits to one file, applied atomically.
• web_fetch: fetch a URL and strip HTML to readable text.
• Project memory now auto-loads CLAUDE.md and AGENTS.md alongside
  TYCODE.md / README.md.

Fixes

• Ctrl+Enter tool-call expand/collapse was unreachable (shadowed by plain
  Enter) — now works.
• Cleared all outstanding compiler warnings; the build is warning-free.

[0.1.0] — 2026-05-10

Initial public release.

• Multi-provider agent: Ollama (local), Anthropic, OpenAI, Google Gemini.
• Tool suite: file read/write/edit, shell with real timeouts, grep, glob,
  HTTP, process management, system info, directory tree.
• Autonomous agent loop with streaming responses.
• TUI with markdown rendering, inline expand/collapse of tool output,
  mouse scroll, dangerous-command confirmation, and auto context compaction.

0.1.0

TyCode v0.1.0: Complete overhaul with visual redesign, new tools, enh…

v0.0.1

TyCode v0.0.1 — Initial release

Full-featured Rust TUI AI agent CLI with multi-provider support.

Core features:
- Multi-provider: Anthropic, OpenAI, Ollama, Gemini
- Autonomous agent loop with tool use (file ops, shell, HTTP, search, process)
- Live streaming output with token-by-token display
- Multiline input (Shift+Enter), word deletion (Ctrl+Backspace)
- Smart Ctrl+C: cancel queue on first press, quit on second within 2 s
- Message queue while agent is processing (Enter to queue, Esc to clear)
- TYCODE.md + README auto-inject into agent context on startup
- Dangerous command confirmation overlay (rm -rf, git push --force, etc.)
- Real token usage display from API response metadata (all 4 providers)
- Context token estimate with colour-coded pressure indicator
- Auto-compact: summarises conversation when context exceeds 80K chars
- Settings editor, model/provider selector overlays, input history
- Native terminal text selection (no mouse capture)
- /clear, /cache, /import, /system, /model, /provider slash commands

v0.0.0

Initial commit